Ever since the recent Middle Eastern conflict broke out, I’ve been fascinated by real estate in the region and how it would react. Markets usually answer questions like that slowly and vaguely; Dubai answers them in public, because every transaction in the city lands in an open registry. So I’ve been building models on that data and following it closely, and publishing the work on a separate publication, Paimio Research, to keep the thematic dispersion down.

If the region interests you at all, I’d recommend two writeups from there.

The first is a deep dive on Emaar Properties, the listed developer behind the Burj Khalifa and Dubai Mall, a company with margins that resemble Nvidia’s, earned on buildings, trading at six times earnings. The report goes through how that’s even possible, customers funding the construction, the state feeding the land, and what the machine is actually worth at this point of the cycle:

https://www.paimioresearch.com/emaar-the-desert-machine/

The second looks at what the Dubai market is doing right now, transaction by transaction: volumes in the premium brackets have fallen by roughly two thirds since the war began and haven’t recovered with the ceasefire, while the cheapest segment is running at records. A market splitting in two, visible in the registry before it shows up anywhere else:

https://www.paimioresearch.com/dubai-real-estate-is-splitting-in-two/

Both are free to read, and if you want the ongoing work, the monthly registry updates and further company reports, you can subscribe at paimioresearch.com.

Emil

Subscribe now

Much is happening under the hood at Canatu, and today we once again received more clues to what is going on.

Four releases in a little over 24 hours. Yesterday a second reactor order from FST, paired with a preliminary outlook cut and the H1 revenue number. This morning a new strategy with updated long term targets, change negotiations for up to 17 people, and the half year report itself. That is not a coincidence. The new CEO wanted the bad number, the plan and the repeat order to land as one package, so nobody gets to read the revenue line without the FST order sitting next to it.

I will not dwell on the numbers. Revenue fell, the loss widened, opex is up because the headcount added under the previous CEO is now fully in the cost base, and cash is still ample. That was all expected. What matters is what was said. Here is the recap.

Not financial advice. This post reflects my personal views and analysis only and is provided for informational purposes. I may hold positions in the securities discussed and may change them without notice. Do your own research before making any investment decisions.

The first 100 days

Maximilian Slawinski used his first shareholder letter to say the obvious quietly: he has stopped everything that was not on a path to scalable, profitable growth, and he is accepting the top line hit from doing so. The engineering services business that actually grew in H1 is being discontinued. The executive team has been rebuilt around five people, with a new CTO and CMO brought in from outside. Business development has been folded into the business units rather than living as a separate exploration function. The whole thing has a name, Polaris, which is what you call a programme when you want the board and the organisation to treat it as a break with the past.

The layoffs fit the same logic. They cover Operations, Robotics, Mobility and Defence, and Medical Diagnostics. They do not touch Semiconductor. Priorities do not get much clearer than that.

The business model is now settled

For years there was ambiguity about whether Canatu would make pellicle membranes itself or sell the equipment to others. That is over. Canatu is an equipment and IP vendor for EUV pellicles. It sells reactors, consumables and licences, and collects royalties when the customer produces. Inspection membranes stay as a product business made in Finland. The CEO letter says it in plain words: “technology leadership in the CNT pellicle market as an equipment and IP vendor.”

This is the capital light choice and the right one given the balance sheet. It also means Canatu’s revenue depends on its customers getting their pellicles qualified at the chipmakers, and Canatu does not control that. The risk section still says the productivity claim for CNT pellicles has not been verified because no reactor is in production use yet. Read that sentence twice. FST has had a production licence since October 2025 and there are still no royalties.

FST doubles down anyway

That makes the FST order the most important item in the package. FST is ordering a second reactor before the first one has generated recurring revenue for Canatu. That is a customer paying for capacity ahead of qualification, which is what you do when you believe qualification is coming. The March long lead item order that some of us read as a possible new customer turned out to be the front end of this FST order. Delivery is 2027, so this is next year’s revenue, not this year’s.

The second reactor, shipped in December 2024 to a still unnamed customer, is still in customer acceptance testing. Twenty months. The report describes it as ongoing and nothing more.

The meat of the report

If you only read one sentence in the half year report, read this one, from the Semiconductor business overview on page 6:

“The first CNT100 SEMI reactor, delivered to FST (Fine Semitech), operates under a commercial production license. The remaining steps toward mass production are the end customers’ product approvals in pilot and risk production.”

Most readers skip that page. It sits between the key figures table and the financial overview and reads like filler. It is not. Three things in the sentence do a lot of work.

First, “end customers’.” Plural, apostrophe after the s. FST has more than one chipmaker qualifying its CNT pellicle. Canatu will never name them, but the set of chipmakers who could be qualifying a pellicle from a Korean supplier is not large.

Second, “pilot and risk production.” That is fab vocabulary, not supplier vocabulary. A pellicle does not have a pilot line or a risk production phase. A fab ramping a new node does. The sentence is telling you FST’s pellicle is inside a chipmaker’s ramp sequence right now, not sitting in a qualification lab waiting for a slot.

Third, “the remaining steps.” Not next steps, not future steps. Canatu is saying there is nothing left on its side or on FST’s side. What remains is the chipmaker’s approval clock, and that clock is not Canatu’s to read.

Now put that sentence next to what the Korean side has been saying, because the Finnish coverage has not done it.

Samsung invested in FST in 2021 as part of a programme to localise EUV pellicle supply, after investing in S&Stech the year before. In September 2025 Korean trade press reported that FST was in price negotiations with Samsung on EUV pellicles it would supply, that the prototype had met Samsung’s requirements on transmittance, uniformity, thermal resistance and hydrogen plasma durability with some particle issues left to solve, and that Canatu had supplied the membrane synthesis equipment. Price talks happen at the end of a qualification, not the beginning.

Then in January 2026 the piece that matters. Samsung placed an order with FST worth about 25 billion won for EUV pellicle mounting, demounting and inspection equipment for its Taylor fab in Texas. The reporting was explicit that this effectively confirmed Samsung would use EUV pellicles at Taylor for the first time. The equipment handles both CNT pellicles and conventional metal silicide ones, so it is not proof of CNT alone. But the sequence after it is.

January: Samsung orders pellicle handling tools from FST for Taylor. March: FST places a long lead item order with Canatu for a second reactor. August: FST converts it into a full reactor order worth over five million euros, before the first reactor has paid a single euro of royalty. FST is a small KOSDAQ company. It does not spend that kind of money on capacity for a product with no recurring revenue yet unless its customer and shareholder has told it to. The second reactor is Samsung’s pellicle localisation programme buying capacity through its supplier. Canatu’s press release describes it, accurately and quietly, as “expanding the customer’s manufacturing capacity of CNT pellicles.”

Two bullets above the FST sentence, in the market commentary on the same page, Canatu notes that Samsung reported mass production of its first generation 2nm technology. Taylor is the fab ramping that node, in exactly the pilot and risk production phases the FST bullet describes. Canatu put the two facts on the same page and left the join to the reader.

This reframes three things.

The end customer for CNT pellicles at scale is now identifiable. It is Samsung, and the first insertion point is Taylor, which is also where Samsung’s large US foundry contracts from last year are slated to run. When Canatu writes “end customers” in the plural, the natural reading is Samsung foundry plus Samsung memory, with SK Hynix as the other candidate through the same supplier.

The memory milestone in the new strategy is not a new market discovery. Samsung is the largest DRAM maker in the world and already runs EUV on its leading DRAM nodes. Once Samsung foundry qualifies FST’s pellicle, Samsung memory is the same supplier, the same pellicle, and the same handling equipment. Memory was promoted to a strategic milestone this week because the customer that is furthest along makes memory.

And for those of us who have been building a case around the second, unnamed reactor and the TSMC ecosystem, the FST chain is the mirror image. TSMC has been introducing Mitsui Chemicals’ metal silicide pellicles since 2019. Samsung chose to localise through FST and Canatu instead. Whoever the reactor 2 customer is, the FST precedent is now the case study they will be watching, and the report gives us nothing new on them beyond “ongoing.”

Why has nobody read the sentence this way? Because Canatu’s disclosures are written to be minimal and the company will not say Samsung. Because the Korean trade press is not in the feed of a Nordic small cap analyst. Because the March long lead order landed the same day as the withdrawal of the old targets and was read as a footnote to a bad news day, not as the first visible consequence of the January Samsung order. And because most people go straight from the key figures table to the outlook and treat the page in between as filler.

Subscribe now

Where the pellicles go, and what a reactor is for

Everything in this section is speculation built on public arithmetic. Canatu has never disclosed reactor throughput, royalty rates or consumable pricing, and Samsung has never said the word Canatu. I am laying out the assumptions so you can swap in your own.

What Taylor consumes

Pellicle demand is wafer starts times EUV layers that carry a pellicle, divided by how many wafers a pellicle survives, plus an allowance for new masks and swaps.

Taylor phase one is a 2nm GAA fab. A plausible ramp is 10 thousand wafers a month in 2027, 25 thousand in 2028, 30 to 40 thousand at steady state. A 2nm flow runs roughly 25 EUV exposures per wafer. Samsung has said it will use pellicles in certain core EUV processes first, so assume 40% of layers pelliclised initially, rising toward full coverage. Canatu’s own 2030 assumption is 15,000 wafers per pellicle; a new fab on a new material will replace more conservatively, so 10,000 is the prudent case.

Run those numbers. At 25 thousand wafers a month, 10 layers, 10,000 wafer lifetime, Taylor needs about 300 pellicles a year, call it 400 with mask turnover. At full 25 layer coverage it is 750, call it 900 to 1,000. At 40 thousand wafers a month and Canatu’s 15,000 lifetime it is again around 1,000. So a mature Taylor consumes somewhere between 400 and 1,000 CNT pellicles a year. That is the whole fab.

What one reactor makes

Three ways to estimate it, and they agree.

From Canatu’s targets: 10 to 20 reactors by 2030 supporting half of global EUV wafers at 15,000 wafers per pellicle works out to roughly 10,000 CNT pellicles a year across the fleet, or 500 to 1,000 per reactor. From the revenue target: 100 million, mostly recurring, from 10 to 20 reactors is 4 to 5 million per producing reactor per year; at 2,000 to 5,000 euros of royalty and consumables per pellicle that is 1,000 to 2,500 pellicles per reactor. From the physics: floating catalyst CVD deposits a membrane in minutes to tens of minutes and the reactor runs continuously, so raw output could exceed 10,000 membranes a year; what limits finished output is yield through transfer, framing, coating at FST and inspection, and on a brand new process that is not high.

Call it 1,000 to 2,500 finished pellicles per reactor per year at mature yield, with 1,000 as the conservative planning number. A reactor is demand limited, not capacity limited.

So reactor two is not for Taylor

One reactor covers Taylor several times over. FST ordering a second one, before the first has shipped a commercial pellicle and before a euro of royalty has been paid, only makes sense if the end customer intends to pull CNT pellicles into Korea.

Samsung’s Korean footprint is a different scale. Foundry at 3nm and 2nm in Pyeongtaek and Hwaseong is perhaps 20 to 40 thousand wafers a month at 25 layers, which is 600 to 1,200 pellicles a year at Canatu’s lifetime assumption. DRAM is the bigger number. Samsung runs EUV on its leading DRAM nodes across something like 200 to 300 thousand wafers a month with about five EUV layers each, which is 12 to 18 million passes a year and 800 to 1,800 pellicles depending on lifetime. Whole of Samsung at full adoption is 2,000 to 4,000 CNT pellicles a year. Two reactors is exactly the right count for that. It is also why “memory” was promoted to a strategic milestone this week: the customer that is furthest along makes more DRAM wafers than anyone.

What it means in euros

The FSB 2025 confirmed that royalties are charged per CNT pellicle sold by FST to its end customers, so the model is per unit. Nobody outside Canatu and FST knows the rate. Bounding it from the top, Canatu’s 2030 targets against its own market size imply a take of around 2,000 euros per pellicle all in. Bounding it from the bottom, a 10 to 20% royalty on a pellicle priced in the low tens of thousands plus consumables is 3,000 to 6,000. Use 2,000 to 5,000.

Taylor alone at 400 to 1,000 pellicles is then 1 to 5 million a year to Canatu at maturity, most likely 2 to 3. Samsung wide at 2,000 to 4,000 pellicles is 4 to 20 million. That is a real business, and it is the first recurring revenue Canatu will ever have from pellicles. It is not 100 million.

The reactor count is the whole story

Samsung at full adoption is two to four reactors. To get to 15 producing reactors by 2030, which is what the targets need, Canatu has to be inside the TSMC ecosystem, where leading edge EUV wafer volume is three to four times Samsung’s. That is the unnamed second reactor, shipped in December 2024 to what the FSB called “a global semiconductor leader” rather than the “Korean semiconductor company” it uses for FST. Different phrasing, and I do not think it is accidental. The report gives us nothing new on that reactor beyond “ongoing,” twenty months after shipment.

The targets

Over 100 million euros of revenue in 2030, at least 20% annual growth from there to 2035, revenue per employee above 400 thousand, and capex under 6 million a year from 2027. The March range of 100 to 150 million is gone at the top, and the 25 to 30% EBIT margin target is gone entirely. The company says it prioritises growth in this phase and profitability will follow scale.

Two things to say about this. One, dropping the upper bound is either the new CEO knowing the number is closer to 100 than 150, or the visibility on the upside being worse than the previous management let on. Two, the margin target is not really gone, it is hidden. At 100 million and 400 thousand per employee you are capped at around 250 people, and if you run that cost base against a normal gross margin you land in the 20 to 25% EBIT range anyway. They just do not want to promise it.

The company has also stopped publishing annual operative targets. The stated reason is negotiating position with customers and suppliers. The unstated reason is that the previous set of operative targets was being missed in public and had become a stick to beat the stock with. You can dislike the loss of transparency and still understand the choice.

Memory shows up

The strategy release lists “expansion of CNT based EUV pellicles to memory chip production” as a named milestone, separate from qualification for device manufacturing generally. Memory was in the March material as a small slice of the pellicle market with upside. Now it is a priority. The reason is not mysterious: FST is Korean and its end customers are Korean, and the report notes that the remaining steps toward mass production are the end customers’ approvals in pilot and risk production, plural. The path that actually has a reactor, a licence and a repeat order runs through Korea and memory. The strategy now says so.

Robotics, Mobility and Defence is a different business than it was in March

The Automotive unit was renamed in June and the report makes clear why. The market commentary is about robotaxis, Chinese L3 licensing, Waymo volumes and defence budgets. The CEO letter talks about “safe 24/7 autonomous driving.” And the outlook contains a new risk factor that was not there in March: the approval of operating autonomous vehicles depends on regulations implemented by local governments.

That sentence tells you the ADAS camera heater lead customer’s volumes are tied to autonomous fleets, not to mass market passenger cars. Full windshield heaters and solar cells, which were priority applications in March, no longer appear in the strategic milestones. The new milestones are ADAS heater ramp, defence and robotics expansion, and a 3D heater product. The unit’s own leader is now the Semiconductor SVP on an interim basis, and it is one of the three areas covered by the change negotiations. This is a unit being narrowed to the things that can ship, with the moonshots parked.

Medical Diagnostics keeps its milestones

Proof of concept for total testosterone was reached, an alpha prototype system is targeted for end 2026, ISO 13485 was obtained, and the strategy names the start of an FDA pathway and a joint development agreement with a leading diagnostics company as milestones. It also got layoffs. Milestone driven development with a tighter budget is the phrase.

The membrane customer

The Semiconductor revenue collapse in H1 was not only about reactors. The report says a major inspection membrane customer was working down high inventory levels during the half. Inspection membranes have been Canatu’s most reliable revenue line, and this customer is very likely the single largest account in the company. The sentence is written in the past tense, which is meant to tell you the destocking is done and H2 normalises. Whether it does is one of the two or three things that will decide the full year number.

What was not said

No problems with the CNT technology itself. No slippage in the FST relationship, the opposite. No mention of a third reactor customer beyond “opportunities with existing and potential new customers have progressed.” No change to the balance sheet story. And nothing at all, in either direction, about who the second reactor customer is.

Next up: my thoughts, and the key information said in the webcast.

Read more

The company has new management in place, several catalysts sitting just around the corner, and a technology that potentially puts it at a genuine chokehold in the chip supply chain. If you’ve only caught my scattered posts, this is the piece that pulls it all together from the ground up, what Canatu does, how it makes money, where the real risks are, and why the timing is interesting right now.

No speculation-for-its-own-sake here. Just the full picture, laid out plainly.

Quick foreword: This is a longer deep dive, around a 45-minute read. Be sure to subscribe if you want to come back and read it later.

Hope you find it valuable!

NFA.

Subscribe now

Background: what Canatu actually is

Let’s start from zero, because Canatu sits in a corner of the market most people have never looked at.

The material: carbon nanotubes (CNT).

A carbon nanotube is exactly what it sounds like, a tube made of carbon, rolled up from a single sheet of graphene, about 100,000 times thinner than a human hair. Despite being almost unimaginably small and light, CNTs are extraordinary: stronger than steel, more conductive than copper, able to withstand temperatures up to 1,500°C, and nearly transparent. For decades they’ve been a “material of the future” that struggled to find applications where their cost was justified. The hard part was never imagining what CNTs could do, it was manufacturing them cleanly, consistently, and at scale.

That manufacturing problem is Canatu’s whole reason for existing.

The company.

Canatu is a Finnish nanotechnology company, spun out of Aalto University in 2004, based in Vantaa. Over roughly two decades and more than €80 million of investment, it developed and patented a distinctive way of producing CNT films, a “dry deposition” process. Without getting technical yet, the short version is that most competitors make CNT films using a wet, multi-step chemical process, while Canatu grows and deposits the nanotubes dry, in far fewer steps. Canatu’s claim is that this produces cleaner, stronger, higher-quality films, and does so more simply. That process, and the reactors that run it, are the crown jewels: protected by hundreds of patents and a set of trade secrets.

Canatu makes money in three broad areas, but one matters far more than the others: semiconductors. The other two, automotive (heaters for self-driving-car cameras) and medical diagnostics (early-stage biosensors), are real but secondary. The semiconductor business is the reason anyone is paying attention, and it’s where this piece will spend most of its time.

What they make for the chip industry.

Here’s the one concept you need, and I’ll explain it properly later: a pellicle. When the most advanced chips are made, the pattern is projected through a photomask, an intricate, extraordinarily expensive template. If even a single speck of dust lands on that mask, it ruins every chip printed with it. A pellicle is an ultra-thin protective membrane, thinner than a soap bubble, stretched over the mask to keep contamination away.

The problem: as chipmaking machines get more powerful, the conditions inside them get so extreme, intense heat and a corrosive hydrogen environment, that conventional pellicle materials start to fail. Carbon nanotubes are one of the very few materials that can survive there. And Canatu doesn’t just sell the finished film; it sells the reactor that produces CNT pellicle membranes, along with a license and a royalty on every membrane that reactor makes. Sell the machine once, then earn a cut of everything it produces. We’ll come back to why that business model matters enormously.

Share Emil Hartela Investing

The story: how Canatu got public.

For most of its life Canatu was a private, venture-backed company, the kind of long-horizon deep-tech bet that takes twenty years to mature. It reached the public market in September 2024 through a slightly unusual route: a merger with Lifeline SPAC I.

A SPAC (special purpose acquisition company) is essentially a pool of money raised on the stock exchange with no business of its own, its sole purpose is to find a promising private company and merge with it, taking that company public without a traditional IPO. Lifeline SPAC I was a Finnish blank-check vehicle backed by a circle of well-known Finnish investors. It raised around €100 million, went looking for a target, and in 2024 combined with Canatu. Through that deal, Canatu listed on Nasdaq First North in Helsinki (a lighter-touch growth market) and raised roughly €100 million to scale its business.

Two things worth knowing about that history, because they matter later. First, the SPAC structure left behind some quirks, warrants and other instruments that affect the share count as the stock rises, which we’ll cover. Second, the company came public with ambitious targets that it has since had to push out, part of why the stock has had a rough ride.

Where things stand today.

Canatu is a small company, roughly €250 million market capitalization, with about €90 million of net cash on the balance sheet, and it’s still early in commercializing its semiconductor business. Revenue actually fell in 2025 as reactor deliveries were delayed. Management has been substantially rebuilt over the past year, with a new CEO, CTO, and commercial lead brought in from the semiconductor industry. And the company is sitting on two installed reactors at two chip customers, one publicly known, one not, whose qualification is the hinge on which the entire investment case turns.

That’s the shape of it: a two-decade-old Finnish nanomaterials company, recently public via a SPAC, with a niche but potentially critical position in how the world’s most advanced chips get made. The rest of this piece unpacks each part, the technology, the business model, the market, the customers, the money, and the risks, and explains why the next few months are unusually important.

The technology, the business model, and the one thing it all hinges on

How advanced chips get made

To understand why Canatu might matter, you need a rough picture of how the most advanced chips are made. Stay with me, it’s simpler than it sounds.

Every cutting-edge chip, the processors in your phone, Nvidia’s AI accelerators, is built by printing unimaginably tiny patterns onto silicon wafers. The printing is done with light. You shine light through a patterned template (a “photomask” or “reticle”) and project that pattern, shrunken down, onto the wafer. Do this hundreds of times, layer on layer, and you build a chip with billions of transistors.

The catch is resolution. To print smaller features, you need shorter-wavelength light. The industry’s current frontier is EUV, extreme ultraviolet lithography, which uses light at a 13.5-nanometer wavelength, far beyond visible light. EUV is what makes today’s most advanced chips possible.

And here’s a fact that matters enormously for this whole story: there is exactly one company on earth that makes EUV machines. ASML, based in the Netherlands. Every advanced fab, TSMC, Intel, Samsung, SK Hynix, Micron, buys its EUV scanners from ASML. These machines cost well over €150 million each, and the entire leading edge of the semiconductor industry depends on them. When you hear about the chip supply chain being a chokepoint, ASML is the ultimate example. Canatu, as we’ll see, is trying to occupy a much smaller but conceptually similar position, one narrow spot that everyone has to go through.

Subscribe now

Why the machines are getting harsher

EUV machines aren’t static, they get more powerful every generation, along two separate axes. Both matter, and people constantly confuse them, so let’s separate them cleanly.

Axis one: power. An EUV machine’s light comes from a source measured in watts. More power means the machine can print more wafers per hour, which is what makes it economical. Today’s machines run around 250 watts. The roadmap pushes this up hard, toward 400W, then 600W, then beyond. Rising power is the key trend to watch, and it’s the one this whole story hinges on.

Axis two: resolution, or “NA” (numerical aperture). This is about the optics, how wide a cone of light the lens can gather. A bigger lens (higher NA) resolves smaller features. The current machines are “Low-NA” (0.33). The next generation, “High-NA” (0.55), uses much bigger, much more expensive optics (around €350–400 million each) to print even finer detail. Intel has started using High-NA first; TSMC has said it will stay on Low-NA for its next several nodes on cost grounds, and Samsung has pushed High-NA out toward the end of the decade.

The crucial point, and this is where a lot of commentary gets it wrong: whether a fab goes High-NA or stays on Low-NA, source power is climbing toward 600W either way. High-NA machines are themselves a 600W-plus regime. And Low-NA machines are being pushed to higher power to squeeze out more throughput. So the harsh conditions are coming down both roads. You don’t need to bet on the High-NA timeline to believe power is rising, it’s rising regardless.

Subscribe now

Why old pellicles stop working

Now back to the pellicle, the thin protective membrane stretched over the photomask to keep dust off. Remember, a single particle on the mask ruins every chip printed with it, so at these prices (a mask can cost hundreds of thousands of euros, a wafer tens of thousands), protecting it is essential.

The problem is that the inside of an EUV machine is a brutal environment, and it gets worse as power rises. Two things attack the pellicle:

Heat. The membrane absorbs some of the light’s energy as heat. At 250W it’s manageable. As power climbs toward 600W, conventional pellicle materials, thin films of polysilicon or similar, simply can’t take the thermal load. They deform, degrade, or rupture.

Hydrogen. EUV machines are filled with hydrogen to keep the optics clean. But that hydrogen, in this energized environment, chemically eats away at many materials over time, slowly destroying the pellicle.

So as the industry marches toward 600W and beyond, the old pellicle materials stop being good enough. The membrane needs to survive intense heat and resist hydrogen and stay almost perfectly transparent to EUV light, all at once. Very few materials can do all three.

Carbon nanotubes are one of them. Strong, heat-resistant, chemically durable, and highly transparent, exactly the profile the job demands. This is why the entire industry, ASML and imec (the leading research institute) included, has converged on CNT as the leading candidate for next-generation pellicles. And it’s why a small Finnish company that spent twenty years learning to manufacture clean CNT films suddenly finds itself potentially sitting on something the whole leading edge needs.

The business model: sell the machine, then tax the output

Here’s where Canatu’s setup gets genuinely interesting, and widely misunderstood.

Canatu is not primarily trying to sell finished pellicles. Its stated model is to sell the reactor, the machine that produces CNT pellicle membranes, to the chipmaker (or its pellicle supplier), and then earn two recurring streams on top:

1. Royalties on every pellicle membrane that reactor produces, and

2. Consumables, the proprietary materials the reactor needs to keep running.

Think of it like selling a printer cheaply and then earning on every page it prints, forever, except the “pages” are consumed continuously by a customer who can’t run their multi-hundred-million-euro machines without them. You sell the reactor once (a one-time payment of several million euros), and then you collect a small, high-margin cut on a stream of output that never stops as long as the fab is producing chips.

Why does the model matter so much? Because it turns a piece of equipment into an annuity. A single reactor placed at a major fab isn’t a one-off sale, it’s the beginning of a recurring, high-margin royalty stream that can run for years, scaling up as the fab consumes more pellicles. And a large fab runs dozens or hundreds of EUV machines, each burning through pellicles continuously. The math on that, which we’ll get to, is what makes the current valuation look strange.

There’s also a strategic elegance to it. A giant fab like TSMC does not want to be dependent on a single outside supplier for finished pellicles, it likes to control its own production. Canatu’s model fits that perfectly: it sells the fab the machine to make its own membranes and keeps the downstream steps (coating, framing) in the customer’s hands. Canatu owns the one hard layer, the membrane synthesis, and lets the customer own the rest. That’s exactly the arrangement a control-obsessed fab is willing to accept.

Subscribe now

So here’s the whole thing, made simple

Strip away every detail and it comes down to this:

The world’s most advanced chips are made by machines that are getting more powerful. As they cross a certain power level, the old protective membranes fail, and carbon nanotube membranes are one of the only replacements that work. Canatu makes the machine that produces those membranes, and gets paid a royalty on every one it makes.

So everything, the entire investment case, hinges on a single question:

Do the fabs adopt Canatu’s reactors, qualify them, and start paying royalties, or not?

Where things actually stand: two customers, two clocks

We ended the last section on the one question everything hinges on: do the fabs adopt Canatu’s reactors and start paying royalties? So let’s look at where that actually stands right now, because this is the live, unfolding part of the story.

Canatu has two reactors installed at two customers. They are at very different stages, and understanding both is the key to understanding the stock.

Customer one: FST (the known one)

The first reactor went to FST, a Korean company. This one is public and relatively well understood. FST is a pellicle maker, and its end customer, per Korean reporting, is Samsung, feeding Samsung’s advanced (2nm-class) chip production, including the big new fab in Taylor, Texas.

Here’s the encouraging part. FST’s reactor has already passed its major qualification hurdle. Remember from earlier that a reactor has to clear a series of acceptance tests, factory acceptance, then site acceptance (”SAT”), before it can move into real production. FST’s reactor completed SAT in mid-2025, roughly ten months after delivery, and Canatu even earned an early commercial license fee from it. FST is now moving toward pilot and volume production.

What’s the FST leg waiting on now? The final steps: full qualification of the finished pellicle in the actual production line, and the ramp to volume, at which point the royalties and consumables revenue we described start flowing. Management has publicly indicated that the first license payments from this production are expected to begin in 2026. So the FST leg is the nearer-term, more visible half of the story, it’s the one that proves the business model actually works and starts producing real recurring revenue. Watch it as the leading indicator.

Customer two: the unnamed “FOUNDRY” (the one that matters most)

The second reactor is where it gets interesting, and where most of the value, and most of the mystery, lives.

This reactor shipped from Finland in December 2024 to a customer Canatu will only describe as “a global semiconductor leader.” Its own investor materials label this customer “FOUNDRY.” The CEO confirmed on camera that it’s “a big semiconductor fab.” For over eighteen months, the company has never named it.

The realistic candidates are a very short list: TSMC, Intel, or Samsung’s foundry arm. And based on a fair amount of digging, which I’ve laid out in detail in a separate companion piece, I think the most likely answer is TSMC. I won’t rerun the full case here, but the short version is that the evidence points that way from several independent directions: the patent timing lines up almost too neatly, TSMC is publicly building an in-house pellicle facility that would need exactly this kind of machine, a Korean competitor’s CEO named TSMC out loud in an interview, and the just-retired Canatu CEO has spent his retirement publicly enthusing about one specific fab. None of it is proof. All of it leans the same way. (If you want the full evidence chain, read the companion piece; here, just hold it as “probably TSMC, not confirmed.”)

Now here’s why the identity matters so much, and why I keep saying this second customer can make or break the whole thesis.

FST/Samsung is a real, valuable leg, but on its own it’s a good business, not a transformational one. The second customer is the one whose fleet is large enough to turn Canatu’s royalty model into something enormous. If it’s TSMC, we’re talking about the company that runs over half the world’s advanced chipmaking capacity. A royalty on the pellicles feeding a fleet that size is a different order of magnitude entirely. So the FST leg proves the model; the second customer is what determines the scale. One is the proof of concept, the other is the prize.

Two customers, two different clocks

The most useful way to hold all this in your head is that these two relationships run on different timelines and get resolved by different kinds of news. Let me lay both out.

The FST / Korea clock (near-term, visible):

• SAT: done (mid-2025)

• Now: moving to pilot and volume production

• Next milestone: royalty and consumables revenue starting to appear, expected to begin 2026

• Where you’ll see it: in Canatu’s own financial reports, and often in Korean trade press first, which has consistently reported this chain ahead of anything out of Helsinki

The FOUNDRY / (probably TSMC) clock (longer-term, hidden):

• Shipped: December 2024

• Now: still in qualification, the SAT has been open around eighteen-plus months, notably longer than FST’s ten

• Why the delay? Two innocent explanations compete. If it’s TSMC, it fits their famously rigorous, slow qualification culture. If it’s Intel, it fits a company that has been cutting capital spending and slowing its roadmap. We genuinely can’t tell which yet, and the two readings have very different implications.

• Next milestone: completion of that second SAT, which Canatu has targeted for 2026

• Where you’ll see it: the SAT completion would likely come from Canatu (still probably without a name attached), while the identity most likely leaks first through Taiwanese trade press, the way TSMC supply-chain stories always break, as the customer starts ordering the surrounding equipment for a production line

What we’re actually waiting for

So, concretely, here’s the state of play as this is written.

The FST leg has cleared its big hurdle and we’re waiting for the royalty revenue to start showing up, the proof that the model pays. The second, larger leg is still in qualification, and we’re waiting for two separate things: the SAT to complete (the technical green light), and, eventually, the name to surface (the event that would let the market attach a fleet size, and a value, to the relationship).

That second SAT is the single most important pending event in the entire story. If it completes and the relationship converts toward production, and especially if the customer turns out to be who I think it is, the stock is a fundamentally different animal than the one trading today. If it stalls, or fails, or the customer turns out to be a smaller or slower player than hoped, much of the upside case deflates.

That’s what “the second customer can make or break the thesis” means in practice. Everything we’ve covered so far, the technology, the business model, the annuity math, is real and interesting. But it’s this one qualification, at this one hidden customer, that turns “interesting” into either “transformational” or “disappointing.” The next few reports are where we start to find out.

Competition: who else is in this race

A fair question hangs over any “chokehold” thesis: if this position is so valuable, why can’t someone else just take it? So let’s look honestly at who Canatu is up against. The competition sits at three different levels, and they matter in very different ways.

The incumbent: Mitsui / the ASML-imec route

The established pellicle supply chain runs through ASML and imec (the leading semiconductor research institute), with Mitsui Chemicals handling assembly and distribution. This is the incumbent path, and it’s not going away. Mitsui has supplied EUV pellicles to the industry for years, and TSMC has historically used metal-silicide pellicles from this lineage.

But here’s the key limitation, and it’s central to Canatu’s whole pitch. The incumbent pellicles are made using older materials and, in some cases, uncoated or differently-constructed CNT membranes that don’t hold up as well at the highest power levels. According to Canatu’s own management, Mitsui’s approach uses an uncoated CNT pellicle whose durability doesn’t match Canatu’s, and Mitsui reportedly missed its own production-facility timelines. The incumbent is real and entrenched, but it’s competing with a solution that may not survive the 600W-and-beyond future that the whole thesis is built around. Incumbency protects the near term; it doesn’t obviously win the next generation.

The fabs’ own in-house ambitions

This is the subtler competitive threat, and arguably the most important one. The biggest customers, TSMC above all, don’t necessarily want to depend on any outside supplier for something this critical. They have the resources and the expertise to build their own.

And they’re not starting from zero. TSMC holds its own pellicle patents, and even patents related to nanotube reactors, meaning it has at least explored building its own membrane-production capability. This is the real ceiling on the bull case: even if a fab adopts Canatu’s reactor today, it could, over time, develop an in-house alternative and reduce its dependence. Fabs also deliberately avoid single-sourcing critical components; they dual-source and keep fallbacks precisely so no supplier can hold them captive.

The counterpoint, and it’s a real one, is revealed preference. Building a production-grade CNT membrane process is extraordinarily hard, that’s the whole reason it took Canatu two decades. The fact that sophisticated players have chosen to buy Canatu’s reactor rather than build their own tells you something. If in-house were easy, they’d have done it. So the in-house threat is best understood as a long-term ceiling on how dependent any fab becomes on Canatu, rather than a near-term reason they won’t adopt at all. The realistic outcome is “Canatu is the lead or first source for the new CNT route,” not “Canatu owns them forever.”

The challenger: a Korean startup with a different method

The most interesting new entrant is aweXome Ray, a Korean startup pursuing a fundamentally different way of making CNT membranes, a continuous “roll-to-roll” process rather than Canatu’s batch dry-deposition. Its CEO has publicly claimed the approach is a step more advanced than Canatu’s, and (in the same breath, amusingly) is the person who named TSMC as Canatu’s customer in an interview. So the challenger is simultaneously a rival and one of our better sources.

How worried should you be? For now, not very, but keep an eye on it. On close inspection, aweXome Ray is small and early: it recently raised only a modest bridge round (~$5M), it’s pre-revenue, its samples are still under customer evaluation, and its heritage is actually in CNT-fiber X-ray tubes for air purification, EUV pellicles are a recent pivot. Its entire lifetime funding is roughly the price of a single Canatu reactor. It’s backed by Korean pellicle-ecosystem money (an S&S Tech affiliate), which tells you the Korean supply chain is hedging into an alternative, but it’s years behind Canatu on the qualification path.

The honest read: roll-to-roll is a genuinely credible alternative architecture that could, if it works at pellicle grade, attack Canatu exactly where it’s most vulnerable (the claim to be the only volume-capable process). But “if it works” is doing heavy lifting, no one has qualified it, and it’s far behind. Treat it as a long-term watch item and a validator of the category (serious money keeps flowing into CNT pellicles), not a near-term threat to Canatu’s current position.

The bigger picture on competition

Put it together and here’s the fair summary. Canatu has a real, buyer-validated lead in the one architecture (dry deposition) that the industrializers have actually chosen so far. The incumbent route may not survive the high-power future. The fabs’ in-house ambitions are a genuine long-term ceiling but haven’t stopped them buying. And the most credible challenger is real but years behind. Nobody in this field has definitively crossed the finish line, so Canatu’s position is best described as a strong first-mover lead in an unfinished race, not an unassailable monopoly. That’s a valuable place to be, but it’s a lead to be defended, not a moat to be taken for granted.

Management: a deliberate rebuild

One of the most striking things about Canatu over the past year is how thoroughly the leadership has been rebuilt. On the surface that looks like turmoil. Look closer and it reads more like a company deliberately re-tooling itself for a different phase, swapping the team that invented the technology for a team that knows how to manufacture it at scale.

The old guard

For most of its history Canatu was run by its founder-era leadership, the people who spent two decades turning a university idea into a working, patented technology. The long-time CEO, Juha Kokkonen, led the company through its SPAC listing and the signing of those first two reactor contracts. The founder-CTO and the original technical team built the science.

That team did the hard early work. But inventing a technology and industrializing it are different jobs requiring different people, and over the past year the board clearly decided it was time for the second kind.

Worth noting, and this matters for how you read the transition, is how Kokkonen left. It wasn’t an abrupt, awkward exit. He was given a warm, public send-off, stayed on as an advisor through year-end, kept his shares, and has spent his retirement publicly enthusing about the semiconductor industry his old company sells into. People who are fleeing a sinking ship don’t usually leave to applause and a standing advisory role. It reads as a planned, orderly handover rather than a rupture, which is a small but genuine point of reassurance.

The new team

Here’s who the board brought in, and the pattern in their backgrounds is the whole point.

CEO: Dr. Maximilian Slawinski, who joined in May 2026 from Soitec. This is a meaningful signal. Soitec is a company whose entire business is scaling engineered semiconductor materials into high-volume production, which is exactly the transition Canatu is trying to make: from delivering a handful of reactors and passing acceptance tests, to full commercial qualification, volume ramp, and a recurring royalty business. Slawinski also relocated to Finland to take the role, a person managing a decline doesn’t usually move countries for the job.

CTO: from the semiconductor equipment world, with a career built specifically on industrial ramp-ups, taking a lab-proven process and turning it into repeatable, high-yield manufacturing. Again, exactly the skill the current phase demands.

Commercial lead (CMO): from the semiconductor and industrial sector (Infineon, Teradyne lineage), brought in to drive sales, marketing, and business development as the company moves from “we have a promising technology” to “we are selling and scaling it.”

Put the three together and the message is unmistakable. This is not the crew you assemble to wind a company down or manage a stall. It’s the crew you assemble when the hard part is no longer the science but the manufacturing and the commercial ramp, when you’re moving from “does this work?” to “can we build a lot of it and sell it?” The board hired people who scale things. That, in itself, is a statement of intent, and the kind of hiring that tends to bring confidence to a management transition rather than fear.

There’s also a supporting detail worth mentioning: around the same time as the CEO announcement, the board quietly re-struck the company’s internal incentive plans. Re-basing incentive targets when you bring in a new leader is what boards do when they’re resetting for a new phase and want the new team’s equity to be motivating rather than underwater, not what they do when they’re giving up.

On ownership: weak on paper, but aligned

Now the honest part, because this is where a skeptic would push, and rightly.

If you look purely at direct insider ownership, it’s not impressive. The operating managers below the top hold relatively small share counts, on the order of a few hundred to a couple thousand shares each. The new CEO, as of the latest disclosures, owns essentially none, though he’s only been in the seat a matter of months and his equity will come through incentive plans that vest over time. So if you were hoping to see executives with huge personal fortunes riding on the stock, that’s not the picture at the individual-manager level.

Two things soften that, though.

First, alignment through incentives. The management team’s upside comes largely through performance share plans, options, and incentive structures tied to the stock rising, which the board just re-based for the new phase. That’s a genuine form of alignment; it’s just alignment via future equity rather than shares bought with personal cash today. It’s weaker than “the CEO put his own money in at these prices,” but it’s not nothing, the team does well if, and largely only if, the stock does well.

Second, and more importantly, the people who own the most haven’t gone anywhere. The largest insider-aligned holder, the chairman, holds a double-digit percentage of the company through his investment vehicle, and that stake has sat completely unmoved through the entire decline in the share price. When the person with the most information and the most money on the line doesn’t sell a single share while the stock falls, that’s a more meaningful signal than any individual manager’s modest holding. The informed, aligned money at the top is sitting still, which is quietly reassuring.

So the honest summary on ownership: direct executive holdings are unremarkable, and you shouldn’t pretend otherwise. But the team is aligned through incentives that were just reset for the new phase, and the big aligned holders at the top have shown zero inclination to sell. It’s not a “management is backing up the truck with personal cash” story, and anyone who tells you it is hasn’t checked. It’s a “the team is incentivized to win and the biggest holders are holding” story, which is a reasonable, if less dramatic, thing to be able to say.

Why the rebuild matters

Step back and the management change fits the larger arc of this piece. Canatu is at an inflection: the technology is proven, the first reactor has passed qualification, and the question now is execution, can they get the second customer over the line, ramp production, and turn reactors into a recurring royalty machine? That is a manufacturing-and-commercial challenge, not a science challenge. And the board has spent the past year assembling a team whose entire collective background is exactly that: scaling engineered materials into volume semiconductor production.

You can read the churn as instability if you want. But the more natural reading, given who left, how they left, and who arrived, is a company that knows precisely what phase it’s entering and has staffed for it.

Share

The numbers: guidance, and how to build it up yourself

Let’s talk about what this could actually look like financially, because the business model we described only becomes real once you attach numbers to it.

What management has told us

Canatu’s official long-term guidance is €100–150 million in revenue by 2030, at a 25–30% adjusted EBIT margin. That’s the anchor. Everything else is a question of whether the company hits, misses, or exceeds it.

For context on the starting point: revenue today is small, roughly €20 million-ish and, awkwardly, it actually fell in 2025 as reactor deliveries slipped. So the guidance implies a large ramp over the second half of the decade, and management has been explicit that the engine of that ramp is the reactor business and the recurring royalty-and-consumables revenue it generates. This is not a company promising to grow its existing product lines steadily. It’s a company saying: the pellicle royalty machine switches on, and that’s what gets us there.

Worth flagging honestly, and we’ll return to this in the risks section, that these targets have already moved once. At the time of the SPAC listing, similar revenue figures were pinned to 2027. They’ve since been pushed to 2030. So treat the year with some skepticism; the shape of the business is more reliable than the date.

How to rebuild the guidance from the ground up

Here’s the useful exercise: rather than just trusting the guidance, let’s roughly reconstruct it from the business model, so you can see what has to be true for the numbers to work. This is back-of-envelope, not a model, but it makes the logic tangible.

Recall the three revenue streams:

1. Reactor sales. Each reactor is a one-time sale of several million euros, call it €5–8 million. These are lumpy: a handful sold over the coming years. If Canatu sells, say, a few reactors across the period plus follow-ons to existing customers, that’s a few tens of millions in periodic equipment revenue. Meaningful, but not the main event, and not recurring.

2. Royalties. This is the heart of it. The way to think about royalties is: number of pellicles produced × royalty per pellicle. Analyst estimates put the royalty in the range of roughly €5,000–10,000 per pellicle, on a finished pellicle that sells for €30,000+. Now, how many pellicles? A single EUV scanner in full production consumes a pellicle every few days, so on the order of 100–180 a year, per machine. A fab runs many machines. So even a single customer, once ramped across a meaningful part of its fleet, consumes thousands of pellicles a year. Do the arithmetic: a few thousand pellicles at several thousand euros of royalty each gets you to tens of millions of euros of royalty revenue, from one customer, recurring, growing as the fleet converts.

3. Consumables. On top of royalties, the reactor needs proprietary materials to keep running, another recurring, high-margin stream that scales with production volume.

Stack those up and you can see how the €100–150M target is reachable without anything heroic: the FST/Samsung leg ramping to volume royalties, plus one more customer converting, plus reactor sales, plus consumables. The guidance isn’t a moonshot; it’s roughly “the two reactors we already have installed reach real production, and we sell a few more.” That’s the base case, and it’s the company’s own.

The upside case is simply: the second customer is very large (a major fleet), adoption goes broader than one line, and the royalty stream compounds. In that world you’re not talking about hitting €100–150M, you’re talking about pushing well past it, with the royalty line alone potentially rivaling the entire guided figure. We’ll keep the scenarios out of this section (they live in the companion valuation work), but the point is that the structure of the model means the upside isn’t linear, a bigger customer or broader adoption doesn’t add revenue, it multiplies it.

Why the margin profile is the quietly exciting part

Here’s the piece that’s easy to overlook, and it may be the most important thing in this whole section.

Canatu already runs gross margins around 72% on its licensing-heavy mix. That’s high, and it tells you something: this is fundamentally an IP-and-consumables business, not a hardware business. And IP-and-consumables businesses have a beautiful property, their margins improve as they scale, because the incremental euro of royalty revenue costs almost nothing to produce.

Think about it. Once a reactor is installed and qualified, every additional pellicle it produces generates a royalty that is nearly pure profit. There’s no factory to expand, no new capital to deploy, no cost of goods to speak of, the customer runs the machine, buys the consumables, and pays the royalty. So as the royalty base grows through the 2030s, the cost structure barely moves while the high-margin revenue compounds on top of it.

This is why the guided 25–30% EBIT margin by 2030 is likely a waypoint, not a ceiling. 2030 is roughly when the pellicle ramp begins in earnest, not when it matures. As adoption broadens and the recurring royalty mix grows relative to the lumpier reactor sales, the margin profile should get better, not worse, over the following years. A business earning a high-margin royalty on a consumable that the world’s most advanced fabs cannot operate without, and whose incremental margin approaches pure profit, is exactly the kind of financial profile that, if it materializes, gets valued richly. Recurring, high-margin, IP-protected royalty streams are among the most prized business models in the market, and Canatu is trying to become one.

The caveat

All of the above is conditional. It assumes the reactors get qualified, the customers adopt, and the royalties flow, the binary from earlier. None of these numbers mean anything until the second SAT completes and real royalty revenue starts printing. The FST leg’s royalties, expected to begin in 2026, are the first genuine test of whether the model produces the numbers the model should produce. Until then, the guidance is a well-reasoned promise, not a fact.

But the reason the setup is interesting is precisely this: if the model works even roughly as designed, the combination of a large recurring royalty base and a margin profile that improves as it scales is a genuinely powerful financial engine, sitting today inside a company the market values as though none of it is going to happen.

Subscribe now

The risks: everything that could go wrong

I’ve spent this piece explaining why Canatu is interesting. Now the other side, because a setup like this has real ways to disappoint, and some of them are serious. If you take nothing else from this section, take this: this is a binary, high-risk situation, and it should be sized accordingly. Here’s what can go wrong, roughly from least to most worrying.

Dilution, but with a cushion

Because Canatu came public through a SPAC, there are leftover instruments, warrants and earn-out shares, that increase the share count as the stock rises. In plain terms: if the stock does well, the number of shares grows, so the market capitalization you need to reach any given share price is higher than a naive calculation suggests. Anyone modeling the upside has to use the larger, fully-diluted share count, not today’s.

This is a real drag, but it’s the least worrying item here, for two reasons. First, some of that dilution is partly self-funding, several of those instruments bring cash into the company when exercised, which is convenient for a business that may need capital to scale if it lands a big deal. Second, and more importantly, Canatu is not a company at risk of running out of money: it holds roughly €90 million of net cash. That’s a genuine cushion. It’s also why the enterprise value, the market cap minus that cash, is only around €150 million, which is what makes the valuation look low relative to the opportunity. The dilution matters for the upside math; the cash pile limits the downside. Of all the risks here, this is the one I’d lose the least sleep over.

The core technology might not clear the bar, yet

Beneath the customer question sits a more fundamental risk that’s easy to overlook: it is not yet fully proven that CNT pellicles can survive high-volume production at the highest power levels for long enough to be economical.

The problem is lifetime. Inside an EUV machine, hydrogen radicals (present to keep the optics clean) chemically attack the carbon nanotubes over time, degrading the membrane. The industry’s answer is a protective coating on the film, which extends its life, and real progress has been made: the CNT pellicle has flown successfully in a scanner, and imec, the leading research institute, has stated they are now “working to extend the lifetime.” So the technology works; the open question is whether coated CNT pellicles can last long enough, at 600W and above, to be viable in mass production.

That question is genuinely unresolved industry-wide. No one has publicly demonstrated a fully-qualified, high-power, high-volume CNT pellicle with production-grade lifetime. If the answer turns out to be “not for several years,” then the entire timeline pushes right, qualification stalls, adoption slips, and the near-term thesis weakens considerably, regardless of who the customer is.

Two things soften this, without removing it. First, the lifetime challenge is a shared industry problem and an active engineering race, not a Canatu-specific failure, and the coating that addresses it is often applied by the customer or integrator, not Canatu itself (Canatu’s job is the best possible membrane substrate, and its patents show it’s engineering the film specifically to receive these coatings). Second, pellicles and membranes are already qualified and in use at current power levels for some applications, it’s specifically the hardest future case (600W+) that remains open.

But it should be stated plainly: whether coated CNT pellicles achieve mass-production lifetime at high power is the single biggest technical risk in this thesis, distinct from the customer-identity question, and it is not something to wave away. It’s the physics that everything else is ultimately betting on.

The binary: this either works or it doesn’t

Here’s the first serious one. As we’ve said throughout, this is fundamentally a binary business. Either the fabs qualify Canatu’s reactors and start paying royalties, or they don’t. There isn’t a comfortable middle outcome where the company muddles along as a modest success, the pellicle royalty model either switches on and becomes a powerful recurring engine, or it stalls and Canatu remains a clever company with a great story and little volume.

That means the downside scenario is real. If the second reactor’s qualification fails, or the customer walks, or CNT pellicles simply don’t clear the bar at the power levels that matter, a large part of the investment case evaporates. The cash cushion means the stock wouldn’t go to zero, but it could fall a long way from here. A binary with a fat upside also has a real left tail, and you should hold both in view at once.

Adoption risk: even if it’s TSMC, that’s not a guarantee

This one is subtle and important, and it’s easy to get carried away past it. Even if the second customer is TSMC, and even if the reactor passes qualification, that does not automatically mean TSMC ends up relying heavily on Canatu.

A few reasons. TSMC is the most capable manufacturing company on earth; it has its own pellicle patents and even its own reactor-related IP, so it is perfectly capable of eventually building its own solution rather than depending on an outside supplier. Big fabs also deliberately avoid being captive to a single vendor for a critical component, they dual-source, they keep fallbacks, they qualify alternatives. So the realistic bull case is “Canatu becomes the lead or first source for the new CNT pellicle route,” not “Canatu becomes TSMC’s permanent, exclusive supplier.” Adoption could also be gradual and layer-by-layer rather than a sudden fleet-wide switch, which would make the royalty ramp slower and smaller than the most optimistic math implies. The identity being confirmed would be a big deal, but it would not, by itself, guarantee the scale of the prize.

A brand-new team we don’t really know yet

The management rebuild is, on balance, encouraging, the incoming team’s background is exactly right for the scaling phase. But let’s be honest about the flip side: they are all new, and we have very little track record with them at this company.

A CEO a few months into the job, a CTO and commercial lead who’ve barely started, however strong their résumés, are unproven in this specific, difficult transition. Scaling a nanomaterials process into high-volume semiconductor production, while managing a demanding fab customer through qualification, is genuinely hard, and we simply don’t yet know how this particular team executes it. Good backgrounds raise the odds; they don’t guarantee the outcome. The confidence the new team brings is partly a bet on pedigree, and pedigree can disappoint.

Delays, delays, delays

If there’s one thing Canatu’s history reliably teaches, it’s that timelines slip. The company came public with targets pinned to 2027; those targets are now 2030. Reactor deliveries have been delayed. The second reactor’s qualification has already run far longer than the first’s. The whole advanced-lithography industry, in fact, has a strong tendency to run later than hoped, the power ramps, the node transitions, the pellicle qualifications, all of it tends to arrive behind schedule.

This matters especially because of the warrants mentioned above: they have an expiry date. A thesis that is right but late is a very different investment from one that’s right on time, and given this company’s record, “later than expected” should be treated as the base case, not the exception. Patience isn’t optional here; it’s a requirement, and even patient holders can be tested.

Competition is real and moving

Canatu is not alone. The pellicle world has several competitors at different layers, and the CNT-membrane approach specifically is attracting new entrants.

The traditional pellicle makers (Mitsui, and others working with ASML/imec) are established and won’t simply concede the field. And notably, there’s a Korean startup, aweXome Ray, pursuing a different CNT membrane manufacturing method (a continuous “roll-to-roll” process rather than Canatu’s batch approach), which its CEO has claimed is more advanced. (Amusingly, that same CEO is the one who named TSMC as Canatu’s customer in an interview, so the competitor is a source and a rival.) For now, that startup is small, pre-revenue, and earlier-stage than Canatu, and the fact that sophisticated buyers keep choosing Canatu’s reactor is evidence its process lead is real. But it’s a reminder that Canatu’s “only game in town” position is a lead, not an unassailable monopoly. If an alternative process qualifies in a couple of years, the moat narrows. This is a field where nobody has definitively crossed the finish line yet, and being first is not the same as being permanent.

Taiwan risk

Finally, the risk that sits over the entire semiconductor sector and therefore over this thesis specifically: Taiwan. If the most likely customer is TSMC, then a large part of Canatu’s upside is tied, directly or indirectly, to the health and stability of Taiwanese chip manufacturing. Any serious geopolitical disruption involving Taiwan and China would be catastrophic for the global chip supply chain, and Canatu, as a would-be supplier into it, would not be spared. This is a low-probability, high-severity risk that no analysis of a TSMC-linked company can honestly ignore. It’s not specific to Canatu, but it’s real, and it’s part of the package.

Putting the risks together

None of these individually breaks the thesis, but together they define what kind of investment this is. It’s a binary, illiquid, timeline-prone microcap, dependent on a qualification that could fail, at a customer that isn’t confirmed and might not adopt heavily even if it’s the giant we suspect, run by a promising but unproven new team, in a field with real and emerging competition, exposed to the single largest geopolitical risk in the global economy.

The reason it’s still interesting despite all that is the asymmetry we’ve laid out: a large cash cushion under the price, an enterprise value that prices in almost none of the upside, and a genuinely transformational outcome if the binary breaks the right way. But the honest framing is that this is a high-risk bet with a wide range of outcomes, and the left tail is a real place the stock can go. Size it like what it is.

Share

Catalysts, and what the future might hold

So where does this leave us, and what should you actually be watching?

After all the technology, the business model, the customers, and the risks, the appeal of Canatu comes down to a rare combination: a company with genuinely large potential, sitting in front of a series of concrete, dated events that will start to resolve the uncertainty, one way or the other, soon. You don’t have to wait years in the dark. The next several quarters are full of signposts.

The near-term catalysts

Here’s what’s actually coming, roughly in order.

The FST royalty ramp. This is the nearest and, in some ways, most important. The FST/Samsung reactor has passed qualification and is moving toward volume production, and management has indicated the first royalty payments should begin in 2026. When that recurring royalty and consumables revenue actually starts showing up in the financials, it will be the first hard proof that the business model works, that a reactor really does convert into an annuity. That moment turns the whole thesis from theory into evidence. Watch the financial reports, and watch Korean trade press, which has consistently reported this chain first.

The second SAT. The single most important pending event. Canatu has targeted completion of the second reactor’s site acceptance test in 2026. If that qualification completes, especially with language suggesting the relationship is moving toward production, it’s a major de-risking of the larger, more valuable leg. The way it’s described will matter, technical progress versus schedule caveats will tell you a lot about which customer it is and how healthy the relationship is.

The name. At some point, the identity of that second customer is likely to surface. If it’s who the evidence suggests, that revelation would let the market attach a real fleet size, and a real number, to the opportunity for the first time. Based on how these things go, the name will most likely leak through Taiwanese supply-chain press before either company ever confirms it, quite possibly as the customer starts ordering the surrounding equipment for a production line.

The mechanical seller finishing. More technical, but real: much of the stock’s recent weakness has come from one old fund winding down and mechanically selling its position into a thin market. That selling has an end. When it’s done, a significant source of downward pressure simply lifts, and in a thinly-traded stock, that alone can change the price behavior.

And further out, the structural driver: as EUV machines push toward 600W and beyond over the next few years, conventional pellicles increasingly can’t cope, and CNT pellicles move from “promising option” to “necessary.” That’s the demand wave the whole thesis is ultimately riding, and it builds through the second half of the decade.

The potential

Let’s be direct about the size of the prize, because it’s the reason any of this is worth the risk.

If the pieces fall into place, if the second customer is as large as the evidence suggests, if the qualification passes, if adoption broadens, and if the royalty model scales the way its economics allow, then Canatu is not a €250 million company. A recurring, high-margin royalty stream sitting at a genuine bottleneck in the world’s most important supply chain, with margins that improve as it scales, is the kind of business the market values in the billions. The honest, if-it-all-works outcome here has genuine multi-billion-euro potential. That’s not a wild claim; it falls out of the business model and the scale of the customers involved once you follow the math.

But, and this is the essential counterweight, that outcome is far from given. Every “if” in that sentence is a real condition that might not be met. The qualification could fail. The customer might adopt only lightly, or build its own. The timelines, true to form, could slip well past what anyone hopes. The competition could catch up. This is not a company where the big outcome is the likely case, it’s a company where the big outcome is possible, and currently priced as though it’s nearly impossible. The opportunity is the gap between those two things, not a promise that the good scenario arrives.

Light at the end of the tunnel

Here’s the note to end on, though. After a genuinely rough stretch, delayed deliveries, falling revenue, withdrawn targets, a sliding share price, there does finally seem to be light at the end of the tunnel.

The technology is proven and increasingly validated by the wider industry. The first reactor has passed qualification, and the first royalties are reportedly close. A capable, purpose-built management team is now in place for exactly the scaling phase the company is entering. The mechanical selling that has weighed on the stock is nearing its end. The structural demand, rising EUV power making CNT pellicles necessary, is building. And a series of concrete catalysts sits just ahead, each one capable of moving the story from speculation toward proof.

None of that guarantees the happy ending. The risks in the previous section are real, and a reader who walks away thinking this is a sure thing has misread me, it isn’t, and it’s precisely the kind of situation where you size a position to survive being wrong. But for a company that spent the past year in the doghouse, the setup going into the next few reports is meaningfully better than the share price suggests. A proven technology, a rebuilt team, a fading forced seller, a building demand wave, and a genuine, dated chance to find out whether the biggest chipmaker on earth is quietly becoming a customer.

That combination, large asymmetric upside, bounded downside, and near-term events that start to resolve the question, is why Canatu is, to me, one of the more interesting situations on the market right now. Not one of the safest. One of the most interesting. There’s a difference, and this whole piece has been an attempt to explain exactly what it is.

Subscribe now

How I think about it (and a note before you go)

Let me end with my honest, personal take, separate from the analysis above.

Canatu is one of the more genuinely interesting situations I’ve come across. You have a real technology at a real bottleneck, a business model that turns a single machine into a recurring royalty stream, a rebuilt management team purpose-built for the scaling phase, a mechanical seller who has been artificially depressing the price and is nearly done, and a stack of concrete catalysts sitting just ahead, any of which could move the story from speculation toward proof. And underneath it all, the genuine possibility that the biggest chipmaker on earth is quietly becoming a customer. When it aligns, the upside is not a nice double, it’s the kind of re-rating that turns a microcap into something institutions have to own.

But I want to be equally clear about the other side, because I’ve tried to be honest throughout this piece and I’m not going to drop that at the finish line. The risk here is on the higher end. This is a binary. The qualification could fail. The customer might adopt only lightly, or build its own. The core technology still has to prove it can hit production lifetime at high power. The timelines, true to this company’s form, will very likely slip. And it sits under the largest geopolitical risk in the global economy. This is not a core holding or a sleep-well-at-night position. It’s a high-risk, high-reward bet that belongs, if it belongs in a portfolio at all, sized as the speculative slice you can afford to be wrong on.

Subscribe now

If you found this useful

This is the kind of work I do, deep, primary-source, investigative research on a small set of overlooked names, the sort of digging most people don’t have the time to do themselves. If this piece was valuable to you, consider subscribing, and if you want the full depth, the complete investigations, the models, the companion evidence pieces, and the work on the other stocks I cover, consider becoming a paid subscriber. The free posts give you the thesis; the paid work gives you the whole thing, every source, every number, every counter-argument, on multiple names, not just this one.

Share

Feel free to comment on what you think, and if you spot any inconsistencies, don’t hesitate to correct them.

None of this is investment advice. I'm long Canatu, including through the warrants, so I'm not a neutral party, read accordingly. This piece is my own analysis built from public information, and parts of it, especially the customer's identity, are inference, not confirmed fact. I may be wrong, and I've tried to be honest throughout about exactly where and how. Do your own work, and never size a speculative, binary position like this as anything other than what it is.

APR Technologies reported its half year on 20 August. I called it a dull report publicly and I stand by that. Nothing broke, nothing landed, the loss widened as expected.

Then I watched the presentation, and it is a substantially more bullish document than the report it accompanies. Not in tone, in content. There are concrete, checkable facts in the recording that appear nowhere in the eighteen pages the company filed the same morning.

That gap turns out to be the most useful thing I learned this quarter, both for what it says about the business and for how to follow this company from here.

Webcast quotes are from an automatic transcript, translated and lightly cleaned.

Part One: The Numbers, Briefly

Net sales came in at 7.1 MSEK for the quarter against 4.7 a year ago, EBIT at minus 11.1 MSEK, cash at 117.5 MSEK with no debt. The 53 percent growth deserves a shrug: Note 2 shows only about 1.9 MSEK was genuinely invoiced, with the rest grants and movement in unbilled accrued revenue. Q2 also redefined net sales to explicitly include grant funded projects, which Q1 did not. The loss widened because they chose to widen it.

One line matters. Tangible fixed assets went from 1.8 MSEK to 12.5 MSEK. They spent 10.4 MSEK on production equipment in three months, roughly seven times their entire previous equipment base, in a quarter where they invoiced under 2 MSEK.

My August piece ended by saying the binding variable in this investment had stopped being demand and become execution speed of the factory. That is the line where execution speed shows up in the accounts.

What the accounts do not tell you is whether it works. The webcast does.

Part Two: What the Recording Says That the Report Does Not

The factory is running, not being built

The report says only that capacity build continues. Lars, on the Enköping line:

“We have also built a robot line that assembles our product and does all the visual vision inspections. It handles the welding of the product itself, everything fully automatic. The capacity is very high, and it can be run around the clock if we wanted to. So it works now.”

The Uppsala cleanroom equipment is installed and in startup. So the 10.4 MSEK has already converted into working, automated capacity rather than purchase orders.

Yield, answered directly

Someone asked the right question: is the bottleneck material, manufacturing precision, or getting repeatable quality between units?

“It has gone surprisingly well during the spring and we have good yield, so now we can raise the tempo and build up larger capacity. So we see no direct bottlenecks that are difficult to handle.”

He volunteered that he had been worried in the spring about the process steps that were new to them, and that it resolved better than he expected. For a company whose entire thesis rests on making hundreds of thousands of units, an unprompted “we have good yield” from the CEO is not a small thing, and it is not in the report.

A pipeline the report does not admit exists

“Then we have new dialogues with a number of additional larger and smaller global tech customers, with whom we are having similar discussions.”

Similar discussions. Not vague interest, not booth conversations. Similar to the relationship that is the entire investment case.

And on inbound:

“Quite a few significant customers have contacted us wanting to talk specifically about Rheion. The Rheion brand has already established itself.”

The written report mentions exactly one customer, three times, and never suggests there is a second.

Geography, which the report scrubbed entirely

“We are soon heading off to a number of trade shows and customer meetings in China, the USA, Italy, France, Germany and others, and that is just what is planned for August and September.”

The Q2 report contains no country names at all. Not one.

The anchor relationship is direct

“We want to work directly with the customers. Zoppas is a partner now, but in many cases, including the customer we have, we deliver directly, and then we also have to match the requirements that are placed on us. They place quite high demands on all suppliers.”

No distributor, no partner in between. APR is a direct supplier to the anchor customer and has to clear that customer’s supplier qualification. That is the context that makes the China production constraint make sense: it is a supplier requirement, not a preference.

Zoppas is already working, before it is signed

Peter describes Zoppas as a family company outside Venice, global, active from China to the USA, supplying heating elements to essentially all satellites, plus defence, data centers, battery systems, EVs and consumer appliances. Three thousand Zoppas heating units flew in the Artemis capsule.

Then:

“They have already introduced us to the big players in Italy. Leonardo, for example.”

APR exhibited inside the Zoppas booth at Farnborough and will do so again at BEX in Rimini. The framework agreement guided for Q3 is paperwork on a partnership that is already running trade shows and making customer introductions. That changes it from a binary event into a formality, and makes a delay more informative than a signature.

And one that cuts the other way

Asked how far the 55 MSEK from the warrant exercise goes:

“It lasts a good twelve months ahead.”

117.5 MSEK across twelve months is roughly 29 MSEK per quarter against an actual Q2 burn of about 21 MSEK, so he is guiding to a real step up in spend. Twelve months from now is August 2027. The stated volume target is 2028. There are no warrants left after TO 1, so any future raise is a fresh issue, and the AGM handed the board an issuance mandate in May.

On the CEO’s own framing, the balance sheet does not reach the volume year. Either 2027 revenue arrives on schedule or there is another raise in between. That belongs in everyone’s model and it is the single most important thing said all day.

Part Three: Why the Report Is So Thin

Here is the part that makes the gap interesting rather than just annoying. The report is not thin because the business got worse. It is thin because someone went through it with a redline.

Q1’s outlook, published 20 May, opened with this:

“APR Technologies har kommit väsentligt närmare ett kommersiellt genombrott med Bolagets mikropumpar för kylning av elektronik, vilka har väckt ett stort intresse hos flera av de allra största techbolagen i världen.”

“APR Technologies has come materially closer to a commercial breakthrough with the Company’s micropumps for cooling electronics, which have attracted great interest from several of the very largest tech companies in the world.”

And later on the same page:

“Bolagets besök på mässor och företag i Europa, Asien och USA har skapat intresse hos flera namnkunniga techgiganter, med vilka Bolaget nu för djupare tekniska diskussioner med.”

“The Company’s visits to trade fairs and companies in Europe, Asia and the USA have created interest among several renowned tech giants, with whom the Company is now conducting deeper technical discussions.”

None of that survives into Q2. No commercial breakthrough, no largest tech companies, no renowned tech giants, no geography.

But look at what did survive, word for word. Q1:

“Bolaget fortsätter arbetet med det globala techbolag som testar och anpassar APR:s mikropumpar i sina produkter.”

Q2:

“Inom Chip Cooling fortsätter arbetet med det globala techbolag som testar och anpassar APR:s Rheion®-teknologi i sina produkter.”

“Work continues with the global tech company that is testing and adapting APR’s technology in its products.” Identical apart from the product name.

Everything deleted between the two reports sat around that sentence. The sentence itself was untouched. And every deleted phrase has one thing in common: it is a claim about a third party’s state of mind that APR cannot evidence. The one relationship they can evidence, because there is a signed order and physical deliveries against it, was preserved verbatim. That is what a redline looks like. A deteriorating customer relationship produces the opposite pattern.

The rest of the document agrees. Q2 added a risk section that did not exist in Q1. The related party note went from “no other material transactions identified” to naming four consulting entities plus supplier transactions with Shanghai Chillwind and the TO 1 underwriting arrangement. Net sales got a formal definition and grant income was broken out for the first time. And the MAR footer changed from the vague legacy formulation to the wording Spotlight’s own guidance prescribes, asserting outright that the report contains inside information. Every one of those increases scrutiny of the company rather than reducing it.

The honest complication is that Q2 does not only delete, it inserts. The CEO letter now says large customers have high requirements, long decision processes and extensive test programs, and that APR’s job is to reduce the customer’s risk step by step. That was not in Q1 in any form, and a compliance review explains removals, not new sentences. My best reading is that Q1 described the stage more flatteringly than warranted and the same review corrected it. It is the only explanation that fits both the deletions and the insertion without requiring the program to have gone backwards, which matters because the hard facts moved forward in the same report.

And the practical conclusion, which is the useful one. The discipline is document level, not company level. Both written reports and the written business update were tightened. The live presentation was not, which is why the funnel, the countries and the working factory all appear there and nowhere else.

So from here: the reports will be sparse and the webcasts will carry the actual news. If you skipped the Q2 presentation because the report was dull, you missed a running production line, a yield answer, a pipeline of additional customers and the runway statement.

Part Four: The Unnamed Big Tech Customer

As you might remember, I have already speculated that the customer could be Amazon, and that the application is somehow linked to BBUs, sidecars, PSUs and immersion cooling.

The presentation gives a mixed message on that front.

On one side, immersion. The entire product demo is immersion cooling: chip, micro fin, dielectric fluid, Rheion on top, roughly 95 degrees down to about 60 when the unit switches on, drawing under half a watt. Lars calls it next generation immersion cooling and says the big players all have immersion in the pipeline even though it is not yet common. That sits well with a good part of what I wrote in August.

On the other side, the global tech customer barely appears. Two mentions in 32 minutes. The deliveries are working very, very well, and the customer is testing and building the units into their own equipment and lines. That is the whole of it. No volumes, no timing, no colour.

Compare that to Q1, when they were essentially shouting that they had a tech customer, plural even, several of the very largest tech companies in the world. In Q2 there is moderation. Some of that is natural, the listing excitement has worn off and the third report is always calmer than the first. But there is a lingering feeling that either the big unnamed customer is further out on the timeline than I assumed earlier, which is very likely because these things always get delayed, or they simply made the presentation more balanced. And there is of course the tail risk that they already know the relationship is slipping in some respect.

The key item, and the one I keep coming back to, is that both the report and the webcast describe this customer as demanding. The report says large customers have high requirements, long decision processes and extensive test programs, and that APR’s job is to reduce the customer’s risk step by step. On the webcast Lars says they need a production apparatus mature enough to meet the demands of large tech buyers, that those buyers place high demands on all suppliers, and that APR delivers to this customer directly rather than through a partner.

The way I read it, they very likely have the technology the customer needs. What they are struggling with is convincing a company of that size that a company of this size can deliver.

And that is where the twelve month runway stings. Put yourself in the customer’s procurement seat. Your prospective supplier has twenty employees, one production line that just started up, and on their own CEO’s account about a year of cash. You are being asked to design their part into something you will ship in volume for years. Seen that way, everything APR is doing right now, the machines, the yield work, the hiring, the Chinese ownership separation, is not selling. It is supplier qualification.

So the team is very likely racing at full speed to reassure this customer. What at some point looked like a sure thing now looks more like a bloody battle.

Still, I think the story is alive and well. The interest is clearly there, the funnel is growing, and they are clearly scaling well.

What to Watch

Late 2026. The patent families APR filed during 2025, inside the collaboration, begin publishing as the 18 month secrecy window expires. Their claims language remains the best discriminator the public record will ever offer, and it is two to four months away.

Q3 2026. The Zoppas framework agreement. Since the partnership is already running trade shows and customer introductions, treat a delay as more informative than a signature.

Autumn. BEX in Rimini and a drone show in London in September, plus customer meetings in China, the USA, Italy, France and Germany. On this quarter’s evidence, whatever comes out of those will appear in a webcast before it appears in a report.

19 November. Q3 report, and more importantly the Q3 presentation. Watch whether the capex line continues or was a one off, and listen for any change in the twelve month cash language.

Q4 2026 to Q1 2027. The validation window closes and the customer decision either arrives or visibly does not. On the CEO’s own runway statement, this is also roughly when the funding question becomes live.

Spring 2027. Customer side patent filings from the collaboration period begin surfacing. The first large company filing that cites APR Technologies as prior art is a fingerprint.

Closing Thoughts

For me this report reads slightly bullish leaning.

The webcast is full of excitement, especially around the market finally understanding and needing the product. Peter’s version of it is the part that stuck with me: they saw this coming when they started the company, but back then it was not urgent. Now the customers are sitting with the problem in front of them, cold plates have maybe five percent left in them, and a new generation arrives every quarter instead of every year. The technology did not change. The world caught up to it.

More importantly, the webcast gives us a glimpse into a very successful start of the production ramp. For a company at this stage, going from lab to volume is the single biggest thing that can go wrong, and it is where most hardware microcaps quietly die. A robot line that assembles and inspects fully automatically, a cleanroom in startup, and a CEO who volunteers that yield is good and that there are no bottlenecks which are hard to handle, is about as good as this phase gets. There are no massive issues anywhere in this quarter.

The open question is the big tech customer, and the report was light on it. The webcast was not much better. Two mentions and no real colour.

But at the same time, we now know others are interested too. Additional larger and smaller global tech customers in similar discussions, inbound contacts asking about Rheion by name, dialogues turning into projects. A year ago this was a single counterparty story. It is less binary now.

So the worst case is that the main big tech customer turns out to be more challenging than anticipated, for one reason or another, while the product itself still has enormous demand behind it.

I did a tiny add today.

Disclaimer: I hold a position in APRTEC and my view is accordingly biased. Everything above is built from public documents, the company’s own presentation, and IR correspondence. The parts about customer identity are speculation, clearly labelled as such. Nothing here is financial advice, and nothing here should be read as a claim of fact about any company named. Do your own research.

At the time I was confused about why the narrative was so explosive. Why does a deep-tech company with almost zero public chatter have such a devoted following in the shadows?

Back in May I only wrote a brief post. Never a proper thesis. It felt like I was missing a piece of the puzzle. I also couldn’t explain why the stock kept grinding lower on no news, and I more or less concluded that the market knew something I didn’t.

After a lot of digging, I think I finally see the picture. Or at least a clearer version of it. There’s still a great deal of speculation and reading between the lines here, and I want to be honest about that upfront. But if the speculation is even remotely right, Canatu is one of the most interesting things I’ve come across.

So let’s jump straight into the rabbit hole.

Read more

I just finished ingesting Haypp's Q2. It was much like what I expected from my revenue estimate post. I will begin by recapping the earnings and then share my commentary on different aspects.

The Numbers

Net sales came in at 1 176,7 MSEK, up 27,7% reported and 27,1% at constant currency. That is the fastest growth since the IPO five years ago. Volume grew 28% and nicotine pouch volume grew 45%, the fastest NP growth in the history of the group. NP is now 71% of total volume. Active consumers hit an all time high of 667 thousand, up 24,4%.

Gross profit was 229,2 MSEK, up 29,2%, at a gross margin of 19,5%. That is the highest margin Haypp has ever reported. Media & Insights added 0,9 percentage points to that margin, but reported margin only improved 0,2 points; the rest was reinvested into targeted pricing. M&I is now just over 10% of group revenue.

Profitability went the other way, by design. Adjusted EBITDA was 55,4 MSEK at 4,7% against 6,5%. Adjusted EBIT was 28,6 MSEK at 2,4% against 4,2%. Net profit was 0,0 MSEK. Overheads rose to 14,8% of sales from 12,7%, split as marketing 1,8%, fulfilment 1,6% and G&A 11,3%. Marketing spend inside other external costs went from 3,5 MSEK to 21,0 MSEK, roughly six times higher, concentrated in the US and UK.

Core did 825,3 MSEK, up 18,2% reported and 15,8% organic, an acceleration from 14,0% in Q1. Gross margin rose 2,1 points to 19,7% on Media & Insights, and EBITDA grew 33% to 87,9 MSEK at a 10,7% margin, an all time high for the segment. NP volume up 19%. The snus decline kept moderating, and Haypp is now taking share in traditional snus, which management describes as a side effect of building a better household offer rather than a strategic goal.

Growth did 351,4 MSEK, up 57,3% reported and 62,4% organic, now 30% of group sales. Active consumers up 74,2%, orders up 72,3%, total volume up 80% and NP volume up 91%. US volume up 125% in a market growing 15% to 20%; UK volume up 113%. Gross margin fell 5,5 points to 18,9% and EBITDA was -32,5 MSEK against -5,6 MSEK. AOV fell 8% to 785 SEK, which management puts down to new consumers starting with small trial orders.

The gap between 80% volume growth and 57% sales growth matters: it is manufacturer-financed price reductions in the US, which management says are margin-neutral with an offset in COGS.

Balance sheet is the soft spot. Net debt rose to 221,6 MSEK from 40,1 MSEK at end of March, and cash fell to 16,4 MSEK from 127,1 MSEK. Leverage went from 0,2 to 1,0 times. The driver is working capital, mainly opportunistic inventory builds and the new Swiss warehouse, plus lease commitments. Six month operating cash flow was 54,9 MSEK against 137,1 MSEK. After quarter end Haypp signed a 400 MSEK committed facility with its house bank.

Other items. Saudi Arabia launched in July, which management calculates as the second largest NP market in the world. Austria was exited on 1 July, having contributed under 9 MSEK in Q2. The UK warehouse was automated and a Swiss warehouse opened. The FDA’s May enforcement discretion is expected to roughly double US SKU count over the next year. In the UK, the ultra strong segment is expected to be banned, removing an assortment disadvantage.

Guidance. EBIT margin stays around current levels for the rest of 2026, G&A intensity does not rise further in H2, and margin expansion resumes in 2027 and 2028. The 2028 targets are unchanged and management says they are on track.

Read more

Read more

APRTEC is a more recent holding of mine, one I found by going through countless names in the liquid cooling race. Understanding the company is trickier than it first appears: the website paints a picture of direct chip cooling, while the patents point toward BBUs and power related products. That gap is not a red flag, it is the normal distance between a platform company’s vision and its first products, but it means the real story has to be assembled from primary sources rather than read off the homepage.

In this piece I dive into the deepest part of the story: the rabbit hole of their unnamed global tech customer, the counterparty behind the development deal, the pre-series orders, and the 150 000 to 500 000 unit volume scenario that the entire investment case hangs on. I have gone through patents, grant databases, hiring pages, earnings calls and one very helpful local newspaper, and while certainty is not on offer, the clues stack in interesting directions.

I expect APRTEC to become a regular series here, where I update and comment on events as they land, and there will be plenty of them: the Q2 report, the patent publications coming late this year, and the customer’s volume decision. The name is severely under covered, essentially no one writes about it, and that is exactly where digging really deep can still produce alpha. So that is what I have done.

Who Is the Unnamed Global Tech Customer?

A rabbit hole in public sources

If you read my APR Technologies deep dive, you know the spine of the whole thesis is one relationship: an undisclosed customer that paid roughly 14 MSEK in spring 2025 to design APR’s EHD pumps into its own products, took first delivery of custom units in December 2025, ordered a pre-series on volume tooling in January 2026, and has a disclosed follow-on scenario of 150 000 to 500 000 pumps per year, which would mean hundreds of MSEK in annual revenue from a single counterparty. Everything else in the APR story orbits this deal.

So naturally I wanted to know who it is. I have spent an unreasonable amount of time going through patents, grant databases, job postings, earnings calls, local newspapers and corporate filings, and I want to say this clearly at the start: I do not know who the customer is, and neither does anyone else outside the companies involved. High certainty is not available here. The customer is running this program in complete public silence, which itself tells you something about what kind of company it is. But the public record contains multiple genuinely compelling clues, and when you stack them, some candidates fit far better than others.

This piece walks through the case for each candidate, from strongest to weakest. Everything below is built from public sources, nothing more. Treat it as an educated rabbit hole, not as information.

The dossier: what we actually know

Before the candidates, the facts. APR has described the customer three different ways across its documents: “ett globalt teknikbolag” (a global tech company), “ett ledande globalt elektronikföretag” (a leading global electronics company), and, in the listing memorandum, the most specific version, “ett multinationellt tech-infrastrukturbolag” (a multinational tech infrastructure company). The application is described as dielectric liquid cooling of circuit boards in IT equipment, and the annual report says APR’s “kylmodul”, cooling module, was integrated into the customer’s systems. In an interview, CEO Lars Almhem described the customer as testing APR’s immersion cooling solution “in a redesigned system,” on 6 to 9 month validation cycles, with revenue potential of 15 to 50 miljoner euro per year. The pre-series is validated in the customer’s “production processes and authentic end-use environments.” The customer demanded that APR divest its Chinese ownership links, which drove the spinoff of the battery subsidiary Chillwind. And the annual report notes the relationship existed “a longer time” before the 2025 agreement. In Q1 2026, APR management traveled to Texas and California.

Two more details come from outside the company, and they are the best clues in the whole record. First, the local newspaper Enköpings-Posten visited APR’s new factory in late December 2025 and described the customer as “a global tech customer of similar stature as the giants Microsoft, Apple or Google,” adding, almost in passing, that the customer had been at the factory and collected a first delivery in person the very same day the paper visited. A local journalist standing inside the building reached for the biggest names in tech as the comparison class. Second, APR’s own press boilerplate has, since at least 2021, mentioned an unnamed “American partner located in Seattle, Washington.”

What the patents say APR is building

This part matters more than people realize, because APR’s patent estate quietly describes the product even while the customer stays hidden.

Read APR’s filings in chronological order and you watch a company pivot in real time. The 2013 to 2016 families are pure space: satellites, RF amplifiers, lab manufacturing methods like screen-printed platinum sintered at 800 degrees. In the 2019 filings, “computing systems” moves to the front of the application lists and direct-contact chip immersion appears. A 2021 filing introduces, for the first time, the phrase “power back up in stationary applications,” which is UPS and battery-backup language. Then 2023 is the pivot year: “telecommunication equipment” and “ICT” enter the vocabulary, electrode gaps widen tenfold (a higher-flow pump), the manufacturing methods switch from lab techniques to stamping and plating, which are volume production methods, and a new family appears whose entire purpose is a lifetime-hermetic, non-pressurized sealed loop using a multilayer foil gas bag to absorb the fluid’s thermal expansion.

And then the 2024 and 2025 filings put hard numbers on it for the first time anywhere in the portfolio: channels under 3 millimeters, rows of flat components held within 2 to 3 degrees of each other by a flow-distribution plate with graded hole sizes, a casing described as “liquid tight but not pressure tight,” an external radiator loop under temperature setpoint control. Put together, the recent filings describe one specific object: a factory-sealed, maintenance-free immersion cassette holding a row of flat heat-generating units, pumped by a stamped, volume-manufacturable solid-state pump, thermally balanced to within two degrees, sealed for the life of the product. The wording leans battery, but the claims are deliberately drafted to also cover “micro-processors and printed circuit boards,” and, tellingly, the patent examiners prosecuted the newest family against Fujitsu and Tyco server immersion tank prior art. The examiners, reading the engineering, saw servers.

Read more

That name is APR Technologies, ticker APRTEC, and this is the deep dive I promised in Execution Memo No. 014.

A few words on what this piece is and isn’t. Coverage of this company is close to nonexistent, so I have tried to write the piece I wish had existed when I started digging: thorough enough to actually understand the technology, the people and the economics, but not a 200 page initiation report. Everything performance related that comes from the company is labeled as such; there is, as you will see, an important distinction between what has been claimed and what has been proven. And the standard warning applies with unusual force here: this is a venture stage company on Spotlight, the position is risky and extremely illiquid, I own shares, and nothing here is financial advice.

With that, let’s follow the heat.

Subscribe now

Part 1: The problem

AI compute is thermally gated

Every generation of AI silicon runs hotter than the last. A GB200 package draws up to 2 700W with heat flux above 50 W/cm² on the cold plates. The Rubin generation pushes toward 2 300W per package in its Max configurations, with variants reportedly reaching 3 700W. Racks have gone from tens of kilowatts to 120kW+, with 600kW rack architectures already announced for 2027.

Air cooling died somewhere around Hopper. The industry’s answer is direct to chip liquid cooling: water glycol pumped through copper cold plates pressed onto the chips. It works, it’s deployed at massive scale, and it is being pushed visibly toward its limits.

The stack, and why every layer is a tax

To understand where the opportunity sits, you have to follow the heat from the transistor to the coolant, because heat doesn’t jump into the liquid. It crawls through layers, and every layer resists.

In a classic packaged chip, a desktop CPU say, the journey is: silicon die, then a thermal interface material filling microscopic gaps, then a metal lid, then a second interface layer, then the cold plate’s base, and only then liquid. Nvidia has already stripped this as far as the current architecture allows; H100, B200 and GB200 modules are bare die packages where the cold plate presses onto the silicon through a single interface layer. Even that last layer is a nuisance: it’s a consumable, replaced every time a cold plate comes off, and one of the two dominant resistances in the chain.

The other dominant resistance is the one most people never hear about. Even when heat reaches the liquid, the liquid touching the metal barely moves. Friction pins a thin film of near stationary fluid against the wall, the boundary layer, and heat must conduct through this sluggish film before flowing coolant can carry it away. Conduction through liquid is slow. In a modern direct to chip system, the interface material and the boundary layer together make up the bulk of total thermal resistance.

So all of chip cooling reduces to two questions: how do you get rid of the layers, and how do you kill the boundary layer.

Watch what the industry does, not what it says

Single phase answers both with force. The recipe never changes: push more liquid through finer channels, accept the rising pressure, let the pumps pay in energy and wear. Vera Rubin is this recipe at its logical extreme; trays go fully fanless and the racks run nearly double the coolant flow at the same CDU pressure. For Rubin Ultra, supply chain reporting points to microchannel cold plates where coolant runs directly across the die surface. Microsoft has demonstrated channels etched into the silicon itself, cutting peak temperature rise by 65 percent versus the best cold plates.

Two phase answers the boundary layer question with boiling, which moves heat far more effectively than conduction through a stagnant film. I have written about two phase before and I still believe in it long term. But it leaves the interface layers intact, chip makers haven’t blessed it, questions about maintenance and reliability at hyperscale remain open, and managing boiling vapor inside a chip package is a problem nobody has solved. It is, in my view, further away than the demos suggest.

Put the datapoints in a row and the direction is unambiguous: coolant is migrating toward the silicon. Room, to rack, to plate on the lid, to bare die, and next to channels at and inside the chip. Every step strips one more layer of the tax.

The endpoint has a specification

Follow the migration to its endpoint and write down what it demands:

A fluid that can touch live silicon without destroying it. Water is disqualified on chemistry.

A loop at near zero pressure, because a bonded chip stack cannot take mechanical stress.

A pump that works at chip scale. Mechanical pumps do not miniaturize; bearings, seals and impellers don’t shrink.

Control fine enough to chase hotspots that move around the die in milliseconds, which passive channels, sized once at manufacture for the worst case, fundamentally cannot do.

Neither of today’s camps is built for this. The endpoint belongs to a third kind of solution, and almost nobody is building it.

Subscribe now

Part 2: The company

Fifteen years of preparation

APR Technologies was founded in 2011 in Enköping by microelectronics researchers, Are Björneklett, Peter Nilsson and Robert Thorslund, with backgrounds in industrial research including 3D chip packaging; the founders were inventors or co-inventors of, among other things, XiVia and Bluetooth. Their founding insight, stated in 2011, was that cooling would become an ever larger problem and that new thermal solutions would be needed.

For its first decade the company built thermal control hardware for the space industry. It won development projects from the Swedish space agency, sold temperature regulation systems to ground stations, won (and then walked away from, over price pressure) a component procurement for the NASA/ESA Orion crew vehicle, delivered parts to an instrument on the ISS, and built long-running relationships with ESA and Airbus, who formally approved APR’s pilot production line, manufacturing processes and quality system in 2022. In 2024 the company was selected into NATO’s DIANA accelerator and won a 19 MSEK grant from the Swedish Energy Agency to demonstrate high volume production. It ISO9001 certified its operations based on the quality system from the space division.

Why does this heritage matter for an AI cooling thesis? Because space is the one customer base where a failed cooling component ends the mission. Fifteen years of surviving NASA, ESA and Airbus qualification is the strongest reliability evidence a company this size can hold, and reliability, as we’ll see, is the heart of the product.

One more detail from the company’s own history that I find telling: in 2020 APR built a lab prototype for super efficient cooling of Nvidia chips, judged the market not quite ready, and predicted the need would grow. That was five years before the market agreed.

The company listed on Spotlight Stock Market on January 13, 2026 at a valuation of roughly 330 MSEK, alongside a warrant series (TO 1) that was exercised to 99,85 percent in June, adding roughly 55 MSEK before costs.

Rheion: the technology

APR’s core technology, branded Rheion this spring, answers both of the questions Part 1 reduced cooling to.

The pump is electrohydrodynamic (EHD). Electric fields push a dielectric fluid through microchannels, with no mechanical pump anywhere and essentially no moving parts in the loop. The physics is old and solid; charges injected into a weakly conducting dielectric liquid get dragged by the Coulomb force between electrodes, transferring momentum to the fluid. NASA Goddard has patented pumps in the same family for space thermal control, and EHD hardware has flown on the ISS. What’s remarkable is not that it works; it’s that essentially nobody else has commercialized it. Reviews of the field classify EHD pumping as early stage as a category. APR has spent fifteen years in a discipline that barely has other occupants.

Because the fluid is dielectric, it can sit in direct contact with live silicon. The interface layers can simply go; the layer problem is solved by removing the layers. And because the flow is generated by electrode structures inside the channels themselves, the boundary layer is disrupted actively by electric body forces rather than by brute pumping power. The company claims a 5 to 10x improvement in heat transfer coefficient, practical cooling capacity of 100 to 300 W/cm² versus 20 to 50 for conventional approaches, chip heat loads handled up to the kW range including hotspots, all at 0,1 to 1W of pump power. Multiple independently controlled units can be tiled across a chip, steering fresh coolant to hotspots in real time as they appear. The system runs sealed at near zero pressure, avoiding both the pressure stress of pumped water loops and the vapor management complexity of two phase.

The whole architecture, the pump, the microchannels and the fluid, is patented as one integrated system: 74 granted patents plus 31 applications across 16 families, most of the portfolio’s costs already taken.

Two deeper points worth internalizing.

First, the pump and the fluid are not separable choices. An EHD pump physically cannot pump water; water is far too conductive. And water cannot touch a die. So the thermally handicapped fluid class, dielectrics conduct heat far worse than water, is the only class that works at the endpoint of the coolant migration, and APR’s entire architecture, micro fins multiplying surface area, micrometer proximity to the die, active boundary layer disruption, is a fifteen year old engineered answer to exactly that handicap. They aren’t fighting the fluid’s weakness; the design exists to cancel it.

Second, EHD gets better as it shrinks, which is the opposite of mechanical pumping. Electric field strength is voltage divided by electrode gap, so smaller channels mean stronger pumping at constant voltage, and the electrodes are patterned lithographically, meaning the technology rides the semiconductor industry’s own manufacturing curve. The published record suggests today’s claims sit at perhaps 10 to 20 percent of what the physics permits; academic microchannel work has demonstrated 790 W/cm² with water in 50 µm silicon channels as far back as 1981, and DARPA programs have pushed embedded cooling toward 1 kW/cm². The scaling laws cooperate, which is historically the difference between a product and a paradigm.

The honesty section: claims versus proof

Everything performance related above is a company claim. There is no published thermal test vehicle result, no third party benchmark, nothing comparable to Accelsius’s public 4 500W socket demonstration or the Corintis and Microsoft in chip data. The published academic EHD record demonstrates flow control recovery at 77,5 W/cm² and roughly 2 to 5x heat transfer enhancement, meaning APR’s claims sit above the public state of the art. There are two readings: either fifteen years of proprietary work in a near empty field has put them ahead of a literature that barely exists (ESA’s own workshop material describes the EHD technology as mastered by APR, with deterministic, fast flow regulation), or the claims are marketing forward extrapolations awaiting proof.

What exists instead of published benchmarks is a customer, and this is the spine of the whole thesis.

The anchor customer

In spring 2025 a leading global tech company signed a key agreement with APR: an initial product order of roughly 1 MSEK, plus roughly 14 MSEK paid to fund a year of development, designing APR’s pumps into the customer’s own end products and building APR’s solutions into their complete system toward their customers. In December 2025 the first custom designed products were delivered. In January 2026 a follow on pre-series order of roughly 1,5 MSEK arrived, covering on the order of a hundred pumps built on volume tooling; the company states the order confirms the technology meets its calculated functionality. Live testing at the customer has been running for over six months with, per the company, positive results, and per the Q2 operational update the pre-series units are now delivered and under test.

The disclosed follow on if tests convert: an inquiry for large scale production of 150 000 to 500 000 units per year, which the company states would mean revenues of hundreds of MSEK per year, from one customer. Behind it sits a separately disclosed inquiry from larger data center manufacturers for roughly 100 000 pump units per year with revenue potential of 50 to 100 MSEK annually. Long term, the company estimates volumes of 500 000 to 1 000 000 units per year within electronics, notes licensing negotiations with international companies, and states that over 50 large American tech companies fit its product market.

Who is the customer? Undisclosed. The memorandum calls it a leading global electronics company; the spin off of the China linked battery subsidiary Chillwind was done partly to meet customer requirements on the ownership picture, which tells you the customer is Western and security conscious. On X there has been light speculation about various candidates; none of it is confirmed and I won’t add to it. What matters analytically is the structure: customer funded development is the strongest validation signal available at this stage. A global company does not spend 14 MSEK of its own money designing a Swedish microcap’s component into its products for fun. It is also, for now, the only validation, and everything between “meets calculated functionality” and “runs in volume for years” is precisely what the ongoing tests exist to establish.

The use cases beyond the data center

The company organized itself this spring into three business areas, and the breadth here matters for the downside case.

Chip Cooling is the spearhead: data centers, AI infrastructure and advanced electronics, where heat directly limits performance, energy use and system design. This is where the anchor customer sits and where the volume scenarios live.

Space & Defence is the heritage business: recurring revenue from the space and nuclear industries (a new order from an established nuclear industry customer was announced in May), ESA and Airbus as repeat customers, NATO DIANA funding for test and verification activities, and, most interestingly, an evaluation agreement with Zoppas Industries Heating Element Technologies (ZIHET), a global Italian thermal group. The intended structure: APR contributes core technology, know how and technical support; Zoppas contributes industrial capacity, established sales channels and takes the system integrator role toward customers. Per the Q2 update the evaluation is in its final stage with an agreement expected during Q3. This is the licensing model in miniature, and I’ll come back to why it matters for the economics.

Energy covers battery related applications: energy storage, mobility, drones. The company exhibited at The Battery Show Europe in Stuttgart with a battery cell partner and reports early dialogues with drone companies in Eastern Europe. The battery angle exploits something unique in the architecture: Rheion is bidirectional. It’s not just a cooler; it’s a thermal controller that can heat as well as cool, hold a setpoint, and respond in milliseconds. For batteries that means both high load cooling and cold climate conditioning (lithium plating avoidance, winter readiness), which no cold plate and no boiling system can offer.

And then there is the use case the market hasn’t touched at all: photonics. Co-packaged optics is coming to AI hardware, and thermal is its central engineering obstacle. Laser wavelength drifts roughly 0,1 nm per degree, photonic components need control within about ±1°C while sitting next to 500W+ ASICs, and the industry’s current solution is essentially to give up: exile the lasers to external modules (the ELSFP standard) and stabilize them with thermoelectric coolers, which burn one to three watts for every watt of heat they move. Nvidia’s chosen modulator technology for CPO, micro rings, is the thermally fragile one, requiring continuous active thermal tuning inside the package. A sub 1W, bidirectional, millisecond setpoint controller at component scale is aimed at precisely the problem the industry is currently routing around rather than solving. APR has zero disclosed photonics customers; this is architectural fit, not pipeline. But it’s a real, documented unmet need, and I count it as a free option on the position.

Subscribe now

Part 3: The people

For a company this size with this little coverage, the diligence has to be on the people as much as the product. The good news is that the people are unusually legible, because the board was visibly rebuilt in December 2025, right before listing, around one specific problem: industrialization.

The CEO: Lars Almhem

Almhem, born 1961, took over as CEO in December 2025. His background: SAAB Automobil and Volvo Cars, 25 years running his own business in quality, efficiency and supply chain globally for Ericsson, NOTE, IKEA and Volvo, production chief at Plasman, and most recently CEO of listed SHT Smart High Tech from mid 2022 to February 2025.

The SHT chapter deserves scrutiny because it’s a dress rehearsal for this job, and I’ve looked at it closely. SHT sells graphene enhanced thermal interface materials, one layer of the thermal stack away from APR. Under Almhem, SHT landed development orders from AMD (including for future AI processors), a delivery approval from another large American electronics company, global consumer distribution through Thermal Grizzly, and the partner strategy that matured into the Henkel relationship, built on the insight that giant customers demand size and stability from suppliers, so a microcap must scale through partners. That is item for item the playbook now running at APR: customer funded development from a global giant, partner led scaling via ZIHET.

The honest counterweight: revenue never converted during his SHT tenure. Full year 2023 net sales were under 3 MSEK against a 38 MSEK operating loss, and the actual breakthroughs, first deliveries toward a world leading AI hardware company, advance payments through the Henkel channel, came after his departure, seeded during his time but harvested by his successor. His departure itself was an amicable, planned succession with warm words from the founder. So the fair summary: proven at building blue chip validation and partner rails at a thermal microcap; unproven at converting them to volume. For the phase APR is in right now, that is arguably exactly the right profile. The phase after is where the rest of the board comes in.

The board

Peter Jansson (joined December 2025) is the quiet key recruit. Co-owner of Lintron Elektronik, which became the foundation of NOTE AB, one of Sweden’s larger contract electronics manufacturers; COO and senior adviser at NOTE until 2009; integration lead for nine acquired units and operationally responsible for three factories in Eastern Europe. He has actually done the thing APR must now do: run volume electronics production at industrial quality and cost. Notably, he and Almhem sit together on another board (HEAD Engineering), meaning the CEO recruited a factory veteran he already knew and trusted.

Johan Claesson (joined August 2025) is the balance sheet, and worth a proper look because he’s more than a financier. Born 1951, Handelshögskolan educated, he owns Claesson & Anderzén with his brother: fourth generation family capital founded 1912 in Kalmar, today a group with property assets valued in the 12 miljarder SEK class, operations across southern Sweden, Germany and Estonia, roughly 46 000 hectares of Ukrainian farmland via Grain Alliance, and a history chairing listed Catella. Swedish wealth rankings have placed him around 2,5 miljarder SEK personally, with the family sphere well above that. His investment style, per an old profile literally titled “He waits for the fall”: hold liquidity, wait for dislocation, buy big.

Through his vehicles Movenio Fastigheter and Bremia he put over 90 MSEK into APR during 2025 and controls roughly 30 percent. And here’s the detail I find most interesting, sitting in plain sight in the memorandum’s directorship tables: during 2025 Claesson took the chairman seat of newly formed entities including CA Datacenter Holding AB and a family of CA Batteribalans (battery energy storage) companies. The anchor investor’s own group is concurrently building data center and BESS ventures, two of APR’s three business areas. This is strategic capital with potential first customer channels attached, not a passive allocation. One more connection worth knowing: Movenio participated in SHT’s December 2024 raise, which is very likely where Claesson met Almhem before recruiting him.

Peter Nilsson and Robert Thorslund, two of the three founders, remain on the board (Nilsson as chairman, having stepped up from the CEO role he held since 2012). Together with co-founder Are Björneklett they hold roughly 35 percent. The technical memory of fifteen years lives here.

Mats Ekberg rounds out the board with small cap financial process experience; for completeness, his firm was paid to produce the listing memorandum and he chaired a dairy company through a 2024 bankruptcy. Process support rather than industrial competence.

Ownership and alignment

Insiders control over 60 percent: Claesson’s vehicles around 30 percent including Bremia, founders around 35 percent, all under lock up agreements limiting sales to 20 percent of holdings in year one. The CEO holds about 120 000 shares. Alignment reads well: the people who know the technology stayed, the people who know factories took the wheel, and the balance sheet has a named backstop with real skin in the game.

The gap, stated plainly

Nobody in the building has worked at a hyperscaler, a chipmaker, or a data center operator. The technical chip knowledge exists in the founders (3D packaging researchers) and the commercial adjacency in Almhem (selling TIM into the AMD ecosystem), but the domain where the value is supposed to be realized has no native representative. The company’s implicit answer is structural: the customer brings the domain knowledge, funds the development, and carries the product to market inside its own systems; ZIHET plays that role for space and defense. Component companies have always scaled this way, Bluetooth itself did, but it means APR’s understanding of its most important end market is mediated through counterparties. A data center or semiconductor name joining the board would be the single most bullish governance signal this company could send, and its absence is the first thing I would fix.

The SHT connection, and why the two aren’t competitors

Since Almhem connects the two companies, a word on SHT Smart High Tech, because the relationship is instructive. SHT sells a better version of the interface layer (graphene enhanced TIM); APR’s endgame abolishes the interface layer at the chip. On the surface that sounds like competition. In practice they occupy opposite ends of the same timeline. Every improvement in cold plates and boiling makes the TIM relatively more of the remaining bottleneck, so SHT’s product gains value throughout the transition, and especially if two phase converts; APR’s architecture wins at the terminal state where fluid meets silicon directly, which arrives socket by socket, years out, at the highest value packages first. Almhem has said publicly he sees the companies as complementary rather than competing, and structurally he’s right: they are opposite sides of the TIM elimination trade, separated by time. For a portfolio builder, the pair is closer to a hedge on architecture timing than a redundancy.

Subscribe now

Part 4: The economics

What the financials actually look like

Let’s be blunt about the starting point. Nine month 2025 net revenue was 12,5 MSEK against an operating loss of 6,7 MSEK; full year 2024 was 13,1 MSEK of revenue against a 10,4 MSEK loss. Of the nine month revenue, 7,1 MSEK was service sales to the tech infrastructure key customer, meaning most current revenue is one relationship’s development fees, with the rest coming from ESA project sales, a battery customer in Asia, the nuclear industry, and public grants. Headcount is around 17 to 20 people. This is a venture stage company with a ticker.

The balance sheet, however, was rebuilt for the task. Roughly 115 MSEK was raised across 2025, another 55 MSEK came in via the warrant exercise in June, and cash after the exercise sits in the neighborhood of 140 MSEK. The board’s stated view is that this funds current operations and planned expansion for at least twelve months. The 2026 industrialization program is roughly 40 MSEK of CAPEX, largely equipment for pilot production and controlled environment assembly, partly offset by the 19 MSEK state grant.

Unit economics: component pricing, not metal bending

Here’s where it gets interesting, because APR’s product has electronic component economics rather than cooling hardware economics.

Back out the average selling price from the company’s own disclosed scenarios: 100 000 units generating 50 to 100 MSEK implies roughly 500 to 1 000 SEK per unit, and the anchor customer scenario implies the same range. A Rheion unit is a sealed ceramic device with printed electrodes, microchannels and proprietary fluid; at volume on automated tooling, the bill of materials plausibly lands around 200 to 400 SEK. That’s gross margin potential in the 50 to 65 percent range in theory, and I’d underwrite 45 to 55 percent in practice, because a customer taking hundreds of thousands of units per year will extract volume pricing (the company’s own history contains a lesson here: APR walked away from the Orion space procurement in 2016 over excessive price pressure, which tells you management has margin discipline and tells you what negotiating with giants does to prices).

Why component pricing holds at all: the product is a sole source, spec’d in part that customers design their products around. The business model, in the company’s own words, ties each customer relatively hard to APR as the supplier of cooling for that customer’s product. Add 74 patents and the absence of any functional substitute, and you price on value, not on BOM. Compare commodity cold plates, bent copper in a brutal supply chain running 25 to 35 percent gross, and you see why the distinction is worth 20 margin points.

At a few hundred MSEK of revenue, with a 60 to 90 person organization, patent costs already largely taken, and depreciation on the modest capex base, that translates to EBITDA margins plausibly in the 20 to 30 percent range. Not software. But rare for hardware.

The factory, its ceiling, and the royalty escape hatch

The production plan: from artisanal short series today to roughly 10 000 units per year starting spring 2026 (about 1 000 units a month), scaling toward 100 000 units per month, one million per year, during 2028. Cleanroom process steps run at the Ångström laboratory in Uppsala; assembly, quality control and flows are being built with high automation in Enköping. The capital efficiency is striking: at the implied unit pricing, a fully utilized million unit line generates 500 MSEK to 1 miljard SEK of revenue against a build cost measured in low hundreds of MSEK, meaning the asset pays itself back in roughly a year of gross profit at full utilization.

But notice the ceiling: at their unit pricing, one million units caps manufactured revenue around a miljard SEK. Scaling beyond it means either a much bigger factory, and the company itself flags that a large volume order would likely require expanding production capacity beyond current plans, with capital needs beyond the warrants, or the more elegant path: licensing.

The licensing channel already exists in embryo. Chillwind, the divested battery business, pays running royalties on sales using APR’s technology. The ZIHET structure has the partner manufacturing and selling while APR contributes technology and collects on it. And management explicitly contemplates selling manufacturing licenses against a per unit royalty as a future scenario, alongside an aftermarket in service on installed systems. Every krona arriving through that channel comes in at something like software margins, and it is the channel the company must lean on anyway once its own walls top out. High margin hardware with a royalty escape hatch is a rare structure, and it exists here because the defensible asset is IP and fifteen years of process know how, not the factory.

The financial goal management has put on paper: 500 MSEK of revenue by 2030, with an ambition to have the growth plan fully financed.

The market, sized honestly

The headline numbers are enormous and mostly irrelevant. Data center liquid cooling passing 15 miljarder dollars by 2028; direct to chip alone growing toward 5,6 miljarder dollars by 2030 at roughly 20 percent annually. APR sells none of the racks, CDUs and plumbing that make up most of that.

The honest addressable slice is component level thermal control: chip level devices, precision temperature stages, pack level battery systems, the layer where solid state flow steering is the actual differentiator. I put that at 3 to 4 miljarder dollars by 2030 across data center, photonics, defense, space and batteries. Against that, the realistic base case is simply the company’s own disclosed pipeline: the anchor customer scenario (hundreds of MSEK), the data center manufacturer inquiry (50 to 100 MSEK), and growing space, defense and energy revenue. That lands somewhere around 300 to 800 MSEK by 2030, bracketing management’s own 500 MSEK target, and requiring roughly one percent of the addressable slice. One percent, anchored by a customer already paying, for the only commercialized option in a discipline the rest of the industry hasn’t entered.

Subscribe now

Part 5: Liquidity, Spotlight, and the cash question

This section exists because it will determine more of your practical experience owning this stock than anything in Part 2.

The venue. APRTEC trades on Spotlight Stock Market, an MTF, not a regulated exchange. Disclosure standards are lighter, institutional participation is minimal, there is no liquidity guarantor agreement, and research coverage is essentially zero. This is both the opportunity (nobody is looking) and the tax (nobody is looking).

The float. With insiders above 60 percent under lock ups, the effective free float is small. The stock moves violently on modest volume; it roughly doubled from the 13,50 SEK listing level within months, ran hard into early June, and swings of 10 percent plus on no news are routine. Position sizing and limit orders are not optional here. If you cannot tolerate a 50 percent drawdown on sentiment alone, this is not the name.

The cash math. Roughly 140 MSEK of cash, a burn that will rise with the industrialization program (40 MSEK CAPEX plus growing opex), and a stated runway of at least twelve months. Read that plainly: a further raise in 2027, possibly late 2026, is close to certain, and here is the paradox worth internalizing in advance: good news accelerates it. A volume order from the anchor customer would require production capacity well beyond the current program, and the company says openly that such a scenario could require additional capital, with loans and other financing examined first but further issues not excluded. So model dilution into your expected return now, and decide in advance whether you’d participate. The mitigating factor is the anchor owner: raising is a different exercise when a multi miljard family sphere with a stated preference for dry powder sits at 30 percent and has already followed on repeatedly. The dilution risk is real; the financing risk, the risk that money simply isn’t available, is much smaller than for the average Spotlight microcap.

Part 6: Valuation, and the M&A floor

I’m not going to pretend a DCF on a pre volume company means much. The more honest frame is the one the industry itself has been writing in transaction prices for two years, because the liquid cooling sector has been consolidating at a furious pace, and it gives us marks at every maturity stage.

At the bottom of the ladder: Flex acquired JetCool in November 2024 for roughly 53 miljoner dollars total. JetCool was a venture stage company with clever chip level cooling technology (microconvective jets), modest revenue, and a team. In the middle: Schneider Electric paid 850 miljoner dollars for 75 percent of Motivair, implying roughly 1,1 miljard for a 150 person company with real CDU and cold plate revenue. At the top: Ecolab agreed this spring to buy CoolIT for 4,75 miljarder dollars at 29 times forward EBITDA, explicitly to align with Nvidia’s cooling roadmap. Vertiv took CoolTera along the way, and Boyd has been reported as a sale candidate. In under two years the strategics have swept the shelf; what remains independent is a handful of venture darlings, the two phase players, and APR.

Now place APR on that ladder. Today, at roughly 27,6 miljoner shares and recent prices, the market cap sits around 700 to 800 MSEK, roughly 65 to 80 miljoner dollars depending on the day. That is about one JetCool above the JetCool price, for a company with a deeper patent estate (74 granted versus JetCool’s much smaller portfolio), its own proprietary fluid, fifteen years of flight heritage, NATO selection, and a paying global customer in live validation, none of which JetCool had. I think of the JetCool mark as establishing something close to a floor logic: this is roughly what a strategic pays for a differentiated cooling technology team and IP estate before commercial proof. At worst, the bear case where the anchor customer walks, you own a competent team, 74 patents, a proprietary fluid and fifteen years of progress in a sector where that exact bundle has a demonstrated clearing price not far below the current market cap, plus a real, revenue generating space and defense business underneath it. The floor isn’t zero, and in a consolidating sector it isn’t even particularly far down.

The upside frame is the other end of the ladder. If the volume order lands and APR becomes a company shipping hundreds of thousands of units into a global customer’s products, the comp set jumps from JetCool to the Motivair and CoolIT axis, where strategics have paid roughly 4 to 8 times revenue and up to 29 times EBITDA for proven, roadmap aligned cooling businesses. On the base case revenue range that supports a valuation of 1,5 to 4 miljarder SEK; in the bull case where licensing carries volume past the factory ceiling, meaningfully more. That is the honest shape of the asymmetry: roughly 1 to 2x of downside protection from the tuck in floor against 4 to 10x in the scenarios where execution matches the disclosed pipeline. And critically, the rerating mechanism doesn’t require years of compounding; the CoolIT arc, bought by KKR pre volume, sold three years later for multiples of entry, shows that in this sector a validation event moves a company from one rung of the ladder to another in a single step. If things start rolling, I believe this stock could rerate by an order of magnitude in a fairly short time frame, because that is quite literally what the industry currently pays at the next rung.

Two caveats on the M&A frame. Control sits with insiders who have every incentive to wait for the higher rung; the floor supports the stock, but no near term bid should be expected. And the company itself flags that its products could come to be classified as strategically significant or dual use, which, combined with Swedish FDI screening and the NATO DIANA involvement, trims the list of possible acquirers even as it validates the asset.

Subscribe now

Part 7: Risks

Let me be explicit, because the risk profile here is closer to a venture investment that happens to have a ticker.

Single customer concentration. The thesis spine is one undisclosed customer. If their tests disappoint, or their own product fails commercially, or they redesign around APR, the volume scenario evaporates. The company’s own risk section names customer concentration explicitly.

Claims without public proof. All performance figures are company claims. No thermal test vehicle results, no third party benchmarks. The pre-series testing may prove them; it may not.

The 100x scaling problem. From artisanal production to a million units a year during 2028, executed by an organization of about twenty people. Yield on sealed fluidic ceramic devices at volume is unproven by anyone, anywhere. The company itself rates the production scale up risk as medium probability, high impact.

Dilution. Near certain, as covered in Part 5. Success makes it arrive sooner.

The fluid question. The fluid chemistry is undisclosed. Fluorinated fluids are the canonical working liquids in EHD research, and if APR’s fluid is fluorinated, EU PFAS regulation sits directly on the thesis. Conversely, the company recently advertised for a fluid and coolant specialist to evaluate coolants on thermal, electrical, chemical and environmental requirements, suggesting fluid selection is active engineering. I have not been able to resolve this question from public material and consider it the single most important unanswered item.

Competition converging on the neighborhood. Corintis has raised over 50 miljoner dollars, ships microfluidic cold plates at real volume, and partners with Microsoft on in chip cooling. AVC reportedly builds Nvidia’s coming microchannel plates. None of them do solid state pumping or active flow control, that discipline remains essentially empty, but if passive geometry plus bigger central pumps stays good enough for long enough, APR’s differentiator gets commoditized around. This, more than two phase, is the real competitive risk: the bet is intelligent thermal control versus brute force, and brute force has a lot of budget behind it.

Timeline slippage, the Sivers lesson. Swedish deep tech microcaps with world class niche technology, blue chip engagements, and imminent commercial breakthroughs have a well documented base rate: qualification cycles run longer than retail patience, financing needs compound, and the stock round trips before anything converts. Almhem’s own SHT tenure demonstrated exactly this pattern. Assume every timeline in this piece slips 12 to 24 months and check whether you’d still hold; I did, and I would, but only at the position size I actually have.

Liquidity. Covered in Part 5. It bears repeating: this stock can fall 30 percent because one holder needed a boat.

Subscribe now

Part 8: Recent developments and what to watch

The recent tape

The past quarter has been unusually eventful for a company this size:

The operational update in early July was, for me, the entry catalyst. It brought a visibly sharpened strategy: three clean business areas with Chip Cooling aimed squarely at data centers and AI infrastructure, the Rheion branding and a rebuilt website that finally explains the offering, and confirmation that the pre-series units ordered by the global tech company have begun delivery and are being tested by the customer. Earlier in the spring the story was harder to grade, a jumble of batteries, space and electronics all at once, right as the stock ran roughly 400 percent in short order; the update resolved the jumble in exactly the direction the thesis wants.

The TO 1 warrant exercise landed at 99,85 percent, adding roughly 55 MSEK and removing the near term financing question.

The ZIHET evaluation is in its final stage, with an agreement expected during Q3. If signed on the intended structure, it’s the first live demonstration of the license and royalty model.

A nuclear industry order in May added another qualified demanding environment customer to the heritage business.

A test rig has been built for demonstrating and benchmarking the technology against relevant cooling scenarios, described as a step toward giving industrial customers concrete decision material. Read: the company is preparing to answer the claims versus proof gap.

Hiring signals: recruitment of a PhD level thermal design engineer and a fluid and coolant specialist within weeks of each other, plus a consultancy partnership with Mint Engineering to add engineering capacity. Companies at this size hire ahead of expected demand, not behind it; the roles read like industrialization and qualification, not research. A small insider purchase was also registered recently, modest in size but pointing the right way.

Coverage: still nearly nothing. A handful of X threads and one thoughtful independent write up focusing on the precision control angle. The discovery hasn’t happened, which for current holders is the point.

The catalyst calendar

Q2 report, August 20. First read on post listing momentum, cash position after the warrant money, and any commentary on customer testing.

ZIHET final agreement, expected Q3. Watch the economics: license fees, royalty structure, any committed volumes.

The anchor customer’s volume decision. The big one, and the timeline belongs to the customer, not to APR. The memorandum framed final tests running during 2026 in the customer’s facility; the honest expectation is a decision somewhere between late 2026 and 2027, with slippage more likely than acceleration. This is the binary that moves the company between rungs of the valuation ladder.

Production milestones. The 10 000 unit per year line entering operation; any announcement of expanded capacity plans would itself signal what management is hearing from customers.

Board or advisory additions. A data center or semiconductor name joining would be the strongest possible tell that the volume phase is real.

Q3 report, November 19, and, realistically at some point, a capital raise, which given the anchor owner is more likely to be an event to participate in than to fear.

Subscribe now

Closing

Let me compress the whole piece into its essential shape.

The industry has agreed, in actions if not in words, on where chip cooling is going: coolant migrating step by step toward the silicon. The endpoint of that migration has a physical specification that disqualifies water, mechanical pumps and boiling. One listed company has spent fifteen years building the only commercialized technology that matches the specification, hardened it in space where failure isn’t an option, fenced it with 74 patents including the fluid itself, and convinced a global tech company to pay for the privilege of designing it into their products, a validation now running live, with a disclosed volume scenario behind it that would transform the company’s scale.

Around it, the sector is consolidating at prices that make the current market cap look like a floor: the last comparable technology tuck in cleared a majority of APR’s entire valuation, and the proven names go for miljarder. Against that stands venture grade risk in every direction: one customer, unproven claims, a 100x manufacturing ramp, certain dilution, an unanswered fluid question, and liquidity that punishes impatience.

I own it small, sized like the venture bet it is, watching three things: the volume order, the fluid, and the boardroom. The asymmetry is the position; the patience is the price of admission.

Thank you for reading. Questions welcome, as always.

Disclaimer: This is not financial advice. I hold a position in APRTEC and my view is accordingly biased. The company is a venture stage microcap trading on Spotlight Stock Market with very limited liquidity; the stock can move violently on small volume and the risk of substantial or total loss is real. All performance figures originate from the company unless otherwise stated. Do your own research.

The liquid cooling scene

First, let’s take a look at the liquid cooling scene. It is vital for understanding the bet here. As more and more money is funneled into AI and data centers running state of the art chips, several bottlenecks emerge. The bottleneck I am interested in is heat removal.

With each generation of chips the trend has remained the same: higher temperatures and thus a larger need for cooling. The core problem is that current single phase solutions become less efficient as temperatures rise.

I have previously written quite a bit about two phase cooling, and I still think two phase cooling looks like a pretty clear winner in the long run. However, I think there might still be room for other technologies that bridge the gap.

Subscribe now

The pitch

Currently the market is dominated by traditional single phase cooling; it is the obvious solution. It works well, it is not too complicated and it is reliable enough. However, it likely won’t be very long before chips produce heat flux that can’t be handled by single phase. The main issue is that single phase doesn’t scale linearly in energy use and wear and tear. When scaled to handle more heat and higher heat flux, current single phase solutions start consuming far more energy via pumping power before eventually running into a complete wall, and they also wear out faster as you pump liquid through them at higher rates.

Two phase has the potential to fix this. Instead of having to pump a bunch of liquid at the chip’s metal encasing, two phase relies on the radically more effective heat transfer achieved by the change of state in the pumped liquid. In many ways two phase cooling seems like the obvious play. It doesn’t hit the same bottlenecks in heat transfer as single phase, and we already largely have the tech developed, with ZutaCore and Accelsius leading the push.

However, there are real problems with two phase cooling, and I think there might be space for other tech solutions to thrive in this rapidly expanding market. One of the main issues with two phase is complexity. Yes, we have two phase solutions in testing and even deployed, but questions about maintenance, scalability and reliability remain largely unanswered. Very few players have committed to two phase and chip producers haven’t given the nod yet. The tech is still years out from becoming the number one choice. And most importantly, heat transfer from metal to liquid, and the difference between boiling versus heating, isn’t the only path to radically improving heat removal.

Following the heat

To see where the gap is, you have to follow the heat from the transistor to the coolant. Heat doesn’t jump from the chip into the liquid; it crawls through a stack of layers, and every single layer adds resistance.

In a classic packaged chip, think a desktop CPU, the journey looks like this. Heat is generated in the silicon die, conducts through the silicon itself, passes through a thermal interface material filling the microscopic gaps between the die and the metal lid, then through the lid, then through a second interface layer between the lid and the cold plate, and finally through the cold plate’s base before it ever touches liquid.

Nvidia already understands that every one of these layers is a tax. Its flagship accelerators have stripped the stack down as far as the current architecture allows: H100, B200 and GB200 modules are bare die packages, with the cold plate pressing onto the silicon through a single layer of interface material. Even that last layer is a nuisance; it is a consumable that has to be replaced every time a cold plate comes off, and it remains one of the two dominant resistances in the whole chain.

The other dominant resistance is the one most people miss. Even once heat reaches the liquid, it isn’t home free. The liquid touching the metal barely moves; friction pins a thin film of near stationary fluid against the wall, the so called boundary layer. Heat has to conduct through this sluggish film before the flowing coolant can carry it away, and conduction through liquid is slow. In a modern direct to chip system, the interface material and the boundary layer together make up the bulk of the total thermal resistance.

So the entire game of chip cooling reduces to two questions: how do you get rid of the layers, and how do you kill the boundary layer.

Subscribe now

Watch what Nvidia does, not what it says

Single phase answers both questions with force. The cold plates on a GB200 are already sophisticated pieces of engineering, copper bases with skived microchannel fins, but the recipe for more cooling never changes: push more liquid through finer channels to thin out the boundary layer, accept the rising pressure, and let the pumps pay for it in energy and wear. Vera Rubin, arriving this year, is that recipe taken to its logical extreme; the trays go fully fanless and the racks run nearly double the coolant flow at the same CDU pressure. For Rubin Ultra, the supply chain reporting points to microchannel cold plates where coolant runs directly across the die surface. Meanwhile Microsoft has demonstrated channels etched into the silicon itself, cutting peak temperature rise by 65 percent versus the best cold plates.

Put those data points in a row and the direction is unambiguous: the coolant is migrating toward the silicon. It started in the room, moved to the rack, then to a plate on the lid, then the lid was removed, and now it is heading for channels at the die and eventually channels inside the package. Every step of that migration strips one more layer of the tax. The industry has, without quite announcing it, agreed on the destination.

Now follow the migration to its endpoint and notice what the endpoint demands. A fluid that can touch live silicon without destroying it, which disqualifies water. A loop running at near zero pressure, because a bonded chip stack cannot take mechanical stress. A pump that works at chip scale, which disqualifies mechanical pumping outright; bearings, seals and impellers do not miniaturize. And control fine enough to chase hotspots that move around the die in milliseconds, which passive channels, sized once at manufacture for the worst case, fundamentally cannot do.

Neither of today’s camps is built for that specification. Single phase water fails on chemistry. Two phase fails inside the package, because managing boiling vapor within a bonded stack is a nightmare nobody has solved. The endpoint of the migration belongs to a third kind of solution, and almost nobody is building it.

The land grab

Read more

In this quarter’s edition we have some mind boggling data to deal with. Saying that Haypp Group web traffic was strong in Q2 would be an understatement; their traffic seems to have disconnected from its past growth trajectory. I will begin by showing you the data. After that I will dissect the big elephant in the room, and finally I will share my take on how Q2 earnings will go down.

NFA. Nothing here is investment advice.

Backdrop

Before diving into the numbers it is worth setting the scene, since Q2 was an unusually busy quarter on the regulatory front. Most of what matters for Haypp happened around the edges of policy, so I will keep this part focused.

The first item is the French ban. France’s decree banning nicotine pouches took effect on 1 April 2026, right at the start of the quarter. The ban is the most comprehensive in Europe; it covers manufacture, sale, import, possession and use of all non medicinal oral nicotine products. France thus became the first European country to criminalise possession and use, going further than Belgium, the Netherlands or Germany. A partial legal challenge suspended the manufacturing and export provisions, but the import, possession and use ban all took effect as scheduled.

The second and more important item is the FDA enforcement guidance. On 8 May 2026 the FDA issued guidance titled “Enforcement Priorities for Certain New Tobacco Products Marketed Without Premarket Authorization.” In plain terms, the agency clarified that it does not intend to prioritise enforcement against nicotine pouch products whose manufacturers have a pending and accepted PMTA. Crucially, the FDA also said it will create and maintain a public list identifying which products fall under this policy. This is a big deal for the industry. For four years pouch manufacturers, suppliers and retailers had been stuck in a grey zone of uncertainty; Haypp’s own VP of regulatory affairs described the guidance as remarkably clear, essentially file a compliant PMTA, wait 180 days, get acceptance and then launch. Companies can now push new products with far less fear of being wiped off shelves. It is worth noting the products remain technically unauthorised until the FDA actually grants marketing authorisation, and some state attorneys general have pushed back; but the direction of travel is a more orderly, regulated market, which benefits a compliant retailer like Haypp.

The third item is the most recent, landing just a few days ago. On 30 June 2026 the FDA issued modified risk tobacco product orders for 20 Zyn products, allowing PMI/Swedish Match to market them with the claim that using Zyn instead of cigarettes lowers the risk of mouth cancer, heart disease, lung cancer, stroke, emphysema and chronic bronchitis. These are the first MRTP orders ever granted for nicotine pouches. The implication for Haypp is bullish; the entire category now has an official regulatory stamp saying pouches are less harmful than smoking. That is exactly the kind of legitimacy that pulls smokers into the category, expands the total addressable market, and makes online retailers of authorised products more valuable. It also strengthens the harm reduction narrative that Haypp leans on when arguing against bans like France’s.

The fourth item is on the hiring side. Q2 looks to have been a very active recruitment season for Haypp; their careers page has had a notable number of open positions through the quarter. This lines up with management’s stated plan to invest in growth, and it strongly suggests headcount is growing, which matters for the margin discussion later.

The final item is on the EU side. The bloc’s long awaited revision of the Tobacco Products Directive (TPD3), which would for the first time bring nicotine pouches formally into EU wide scope, has once again slipped. The Commission published its evaluation of the current directive on 2 April 2026, concluding that novel nicotine products including pouches are insufficiently covered, which is the trigger for a revision. But the actual legislative proposal, expected at various points to arrive in mid 2026, has effectively stalled in the evaluation and impact assessment phase due to sharp disagreements between member states. The proposal is now not expected to land in draft form until late 2026 at the earliest, with rules unlikely to take effect before roughly 2028. The implication is neutral to mildly positive for Haypp in the near term; the regulatory overhang that could eventually impose flavour restrictions or tighter standards across the EU keeps getting pushed further out, buying the category more time to grow under the current fragmented national patchwork.

Subscribe now

Web Traffic Breakdown · June 2026

Web traffic is the cleanest leading indicator we have for Haypp Group; a pure-online B2C retailer where every paying customer has to land on one of these domains, and where revenue is single-transaction, so visit volume maps almost mechanically to orders × AOV. In an in-house audit of the last 9 reported quarters, panel visits explained group revenue at Pearson r ≈ 0.94 on a YoY% basis.

Each domain card below leads with the June total, then the month-on-month move vs May, the gap to the Q1 2026 monthly baseline, and YoY.

Core, Mature Nordic

Looking at Haypp's core markets we see strength in the smaller sites, while the biggest core site, snusbolaget.se, is slightly down for the quarter at -6,5%. This was largely expected already at the end of last quarter, as net sales grew more strongly than active customers, a leading indicator for core. The equal weighted average across the Core cohort came in at +8,9% versus the prior quarter baseline, with snushjem.no the standout at +28,2%. See charts below.

snushjem.no — Norway

June 2026: 65K visits — +28.2% growth vs the Q1 2026 monthly average (51K). Month-on-month it’s up 15.0% on May’s 57K. On a YoY basis the site is +112.9%. This is the strongest grower across the segment so far this month.

snuslageret.no — Norway

June 2026: 152K visits — +13.5% growth vs the Q1 2026 monthly average (134K). Month-on-month it’s up 1.7% on May’s 149K. On a YoY basis the site is +32.2%. One of the strongest growers in the segment this month.

snus.com — Norway

June 2026: 45K visits — +6.6% growth vs the Q1 2026 monthly average (43K). Month-on-month it’s down 8.8% on May’s 50K. On a YoY basis the site is +21.0%. Growing in line with the segment composite.

nettotobak.com — Sweden

June 2026: 140K visits — +2.9% growth vs the Q1 2026 monthly average (136K). Month-on-month it’s down 11.4% on May’s 158K. On a YoY basis the site is +42.6%. Growing in line with the segment composite.

snusbolaget.se — Sweden

June 2026: 484K visits — -6.5% decline vs the Q1 2026 monthly average (517K). Month-on-month it’s down 1.6% on May’s 491K. On a YoY basis the site is +32.9%. The weakest performer in the segment, dragging the composite.

Subscribe now

Growth — UK · US · DE · CH (Vape + HNB)

For the growth segment we are seeing something completely unprecedented. In the history of Haypp there hasn't been this aggressive a web traffic inflow to their sites; more on that later. The equal weighted average across the Growth cohort came in at +61,0% versus the prior quarter baseline, driven by haypp.com at a staggering +158,0%. Not everything rose though; snusmarkt.ch was the weakest at -13,5%. See chart below.

haypp.com, (multi-market flagship)

June 2026: 832K visits — +158.0% growth vs the Q1 2026 monthly average (322K). Month-on-month it’s up 32.4% on May’s 629K. On a YoY basis the site is +196.9%. This is the strongest grower across the segment so far this month.

northerner.com — United States

June 2026: 998K visits — +130.2% growth vs the Q1 2026 monthly average (434K). Month-on-month it’s up 33.5% on May’s 748K. On a YoY basis the site is +261.6%. One of the strongest growers in the segment this month.

vapeglobe.de — Germany

June 2026: 31K visits — +25.3% growth vs the Q1 2026 monthly average (25K). Month-on-month it’s down 20.6% on May’s 39K. On a YoY basis the site is +225.8%. Growing in line with the segment composite.

nicokick.com — United States

June 2026: 449K visits — +5.3% growth vs the Q1 2026 monthly average (427K). Month-on-month it’s down 3.9% on May’s 467K. On a YoY basis the site is +0.8%. Growing in line with the segment composite.

snusmarkt.ch — Switzerland

June 2026: 15K visits — -13.5% decline vs the Q1 2026 monthly average (17K). Month-on-month it’s up 12.5% on May’s 13K. On a YoY basis the site is +23.2%. The weakest performer in the segment, dragging the composite.

Note

Although the Haypp and Northerner charts look like they are going parabolic, they seem to be coming down sharply in the last week of June.

Subscribe now

The Elephant in the Room

If you have read multiple Haypp Group web traffic reports by now you will immediately see that something is up with Northerner and Haypp.com. Never in the history of all of Haypp web traffic has there been such a violent move. Not only are the two sites completely synced; they both started rising at the same time and peaked at the same time. Even further, the size of the moves are very close. This does not feel organic. So the question is: why have two of Haypp’s biggest sites all of a sudden seen 200% surges in traffic in Q2.

First I will present some of the evidence so that you can get a look into some of the logic behind my prediction and what I tinker around.

The first piece of the evidence can be found in SimilarWeb. Looking at SimilarWeb data and where that traffic is coming from, you notice that Germany is dominating that traffic. Essentially in one quarter the amount of Germans visiting Haypp’s websites has increased over 400%.

The second piece of evidence can also be found in SimilarWeb. A big chunk of the new traffic seems to be coming from display ads, many of them targeted at adult sites.

Another piece of evidence, or perhaps more of a potential lead, is the fact that France’s ban of nicotine pouches went into effect at the start of the quarter.

Even further, many German gas stations and convenience stores have voluntarily stopped selling nicotine pouches.

Now piecing all this together, it seems like Haypp Group has decided to do a sort of blitz marketing campaign that particularly targets Germans. Essentially we likely have a mix of factors giving Haypp extraordinary growth here. First, the French ban likely has some small part of Frenchman ordering NP to Germany and then using it illegally in France. At the same time Germans also struggle to get NP as a lot of the logical places to buy it don’t sell it, pushing online penetration way up.

Haypp Group must have realized that now is their time to strike, and thus they started a strong ad campaign, likely the largest one they have ever done.

My data roughly shows that Haypp has gotten around 5 million extra visits to their site in Q2. Most of this can likely be attributed to them running ads. Now the cost of running these ads should be very low; we are talking CPC under 10 cents. So essentially the extra cost of their marketing push would likely be under 5 MSEK.

Now the bigger implication here is of course what do these new consumers order, if they order at all. It is almost always the case when there is a huge influx of customers that order value goes down per visit. The difficult part is knowing to what extent. We don’t have any data to estimate the elasticity of current events, but I will try to do my best to predict how well this spike translates into revenue in the next chapter.

Now before I end this section I want to say a word of warning. My interpretation of what is going on can be faulty. Web traffic can be manipulated and there are countless factors that could mislead in this type of situation. Germans all of a sudden ordering a bunch more nicotine pouches seems highly suspicious. It is possible that SimilarWeb’s panel data simply over counts Germany.

Share

What the street thinks

Before showing my own estimates, let’s take a look at what analysts and Twitter think. First the analyst consensus from Barclays, Deutsche Bank and Pareto Securities; they land at 1140 MSEK. Chart below. Here it is worth noting that the 1140 MSEK likely represents a time series estimate, that is roughly where revenue should land if they grow at the pace of management’s guidance.

And here is the result of Twitter, from my one week poll that got 34 votes. The result can be seen below and averaged shows 1158 MSEK. On average the X crowd has been really good at estimating revenues, with the exception of last year's Q2. Thus 1158 MSEK provides another valuable data point, as last year's Q2 miss was likely caused by a bias to the upside, since very few expected Haypp shrinking.

In total, both X and analysts seem to think that revenue will have grown roughly 25% from last year; a very reasonable estimate. Next, let's dive into my estimate.

The revenue prediction

Read more

To understand the significance of what I have been seeing in the post pandemic years, I will start by taking you back to how things looked roughly ten years ago in places like Capri and Porto Cervo.

Back then, the yachts you saw sat in the 30 to 40 meter range, and very rarely more than that. Spotting something larger happened maybe once a week. I still remember seeing Paul Allen’s Octopus at roughly 126 meters somewhere near Capri. At the time, that boat ranked among the ten largest yachts in the world.

When you hopped off your boat and took a stroll into one of the luxury marinas, things were also very different. There was clearly more moderation. Stuff did not have to sport a strong brand or be well known. Social media was still in its infancy, so the shape of luxury in many of these places was completely different. I will return to social media soon, because it plays a bigger role in this story than one might think.

Now to the present. I just returned from the Mediterranean once again, and it suddenly hit me that the change between now and ten years ago is massive. The first thing you notice is that those 30 to 40 meter yachts, which represented a sort of ceiling back in the day, are now closer to 100 meters. Instead of seeing one very long yacht like Octopus once a week, you now see several 100 meter boats daily. The 100 meter boat is essentially the new 40 meter boat. At first this might sound insane, but I will do my best to explain what is going on.

The same shift shows up across the marinas and the wider luxury ecosystem. Where you previously had a marina filled with nice but unknown restaurants and brands, you now have pretty much the opposite. The towns are filled with all the usual suspects like Stefano Ricci, Kiton, Brunello and Loro Piana. There are very few unknown names left. By unknown, I mean brands that do not already have a store in every major city around the world. Restaurants follow the same pattern. Many are now well known luxury chains. It feels like there is a Zuma or a Cipriani in pretty much every place now.

The thing I want you to keep in mind from this intro is that there has been a massive change over the last ten years, and few have had time to register its magnitude. Many simply hold the wrong perception of reality when it comes to the post pandemic world. This new reality is stark enough to raise an obvious question. What kind of investments will benefit from it? So I decided to map several winners in the field that I believe have the potential to do well. But before that, I will explain the underlying trends driving the stark difference we can now see.

Subscribe now

Why the Boats Got Bigger

As I mentioned earlier, the last ten years have brought a massive change in the luxury yachting scene. I will now explain my take on why the difference is so stark.

The first culprit is social media, and especially Instagram. It was not long ago that Instagram did not exist, let alone the subcultures that I believe are driving a lot of the change we are seeing. Both old money and the wealth and hustle crowd are thriving on Instagram, and this magnifies the whole yachting sector. It is now far easier to see that your boat is not as big as someone else’s. Your wife and children would rather go to Zuma for the pictures. Yachting already represents the top of luxury, and when you couple it with social media, the result is exactly what we see today. Bigger yachts, more exuberance, and importantly, less durable spending. I believe social media is a large part of the puzzle, but not all of it. There are two other factors that are equally important.

The second tailwind is tech, the wealth created from tech, and what it does to the buyer demographic. In the past, yachts could really only be afforded by 70 year olds. Now you have tech billionaires and millionaires who are under 40. The effect is fairly self explanatory. These younger people have more of a need to show off, since they grew up with social media. Many of them sit on more wealth than they know what to do with, and most importantly, they are young. Yachting is not always the most comfortable form of luxury for an older person, but for a 40 year old there is nothing better. On top of that, the fact that under 40 tech billionaires now own yachts and post about them online adds even more pressure on everyone else to get a taste of the lifestyle. That brings us to the third reason yachts are booming like never before.

The third reason ties back to the picture social media paints. If yachting looks common, and not partaking makes you an outsider, then the only solution for someone with less money is to charter. And charter they have. Chartering has grown explosively over the past ten years. A person earning 500k a year obviously cannot afford to buy a 50 meter yacht, but what they can do is gather ten friends, put in 30k each, and split a week on the water. Sustainable or not, it is happening. More and more people are willing to spend an outsized portion of their income and wealth on chartering bigger and bigger boats.

These three forces combined are what, in my opinion, has created this massive influx of money chasing a sector that ten years ago was far smaller. Now let us jump into some of the stocks that enable and profit from this changed reality.

Subscribe now

Sanlorenzo S.p.A. (SL.MI, Euronext Milan)

Of all the stocks I have looked at so far, Sanlorenzo is the one that strikes me as the safest and most reasonable bet. It trades at roughly 11 times earnings, which sits well below its own ten year average, and the business underneath that multiple looks genuinely healthy rather than cheap for a reason.

The story starts in 1958 in Ameglia, on the Ligurian coast. For most of its life Sanlorenzo was a small, almost artisanal builder. The modern chapter began when Massimo Perotti, still chairman and CEO today, took control in 2005 and slowly turned it into the reference name at the top of the market. It listed in Milan in 2019, and since the IPO it has hit every financial target it set, growing revenue, margins and order intake across very different market conditions.

What makes Sanlorenzo interesting for this thesis is how pure the bet is. The company holds its leadership in the 30 to 40 meter segment, which is exactly the part of the market that has migrated upward over the last decade, and through its Superyacht division it builds well beyond that. There is no mass-market boat dragging down the brand. The structure is clean: Bluegame for smaller utility yachts, the core Yacht division, and the Superyacht division at the top. Compare this to many peers who either stretch their range down into far smaller boats to chase volume, or sit buried inside a wider industrial group, or remain private and uninvestable. Sanlorenzo gives you close to a pure play on exactly the kind of large, expensive yacht that now fills Porto Cervo.

The thing they are genuinely good at is scarcity. Management talks openly about choosing value over volume and elegance over extravagance, and they mean it as a financial strategy, not just marketing. They deliberately keep waiting lists long rather than flooding the market, which protects pricing and resale value. Waiting lists currently extend as far out as 2029. That discipline shows up in the margins. EBITDA margin has climbed from 14.5% in 2019 to 18.8% in 2025, and net profit went from about 27 million euros in 2019 to 107 million in 2025.

The buyer base is worth dwelling on, because it ties directly into the trends I laid out earlier. Sanlorenzo sells to a sophisticated, repeat ultra high net worth clientele, and a meaningful share of growth is coming from exactly where you would expect given the demographic shift: the Americas and the Middle East. In Q1 2026, revenue from the Americas rose almost 19% and the Middle East and Africa rose 24%. These are younger, newer money buyers entering a market that used to belong to an older European crowd.

The backlog is the part that makes the whole thing feel solid rather than speculative. As of the end of March 2026 the order backlog stood at roughly 1.23 billion euros, with 90% already sold to final clients rather than to dealers. That backlog covers around 72% of 2026 revenue guidance, and order intake has now grown for seven consecutive quarters, with Q1 2026 up over 25% year on year. In a cyclical industry, that kind of forward visibility is rare and it is what lets you underwrite the earnings with some confidence.

None of this means it is risk free. The broader superyacht market actually normalized in 2025, with total sales down around 8% versus 2024, and Sanlorenzo outperformed rather than escaped that. The business is still cyclical, and a prolonged downturn in new builds would eventually pressure intake. But the combination of a clean balance sheet sitting in net cash, a long backlog, a disciplined scarcity model and a low double digit multiple is exactly why I see this as the anchor position of the basket rather than the lottery ticket.

Share

Ferretti S.p.A. (YACHT.MI, Euronext Milan)

If Sanlorenzo is the anchor, Ferretti is the close second, and arguably even slightly cheaper. It trades at a similar low double digit multiple while carrying something Sanlorenzo does not have: a stable of the most recognizable names in the entire industry.

That portfolio is the whole reason to own it. Ferretti owns and markets seven brands, including Ferretti Yachts, Riva, Pershing, Itama, CRN, Custom Line and Wally, distributed through roughly 60 dealers across more than 70 countries. Riva alone is arguably the most iconic name in yachting, the brand of the classic wooden runabout that has meant glamour on the water since the 1950s. And for me personally the jewel is Wally, my favorite brand of the lot, the maker of those radical, almost spaceship-like sailing and motor yachts that look like nothing else on the water. Owning Ferretti means owning the heritage and the design language that a large part of the buyer base is actually chasing.

Strategically Ferretti is broader than Sanlorenzo. Where Sanlorenzo deliberately stays narrow and scarce, Ferretti spans everything from serial production boats up through full custom superyachts built at CRN, covering far more of the market. The group runs both a serial market and a superyacht segment, giving it a transversal presence across most segments of yachting. That breadth is a double edged sword. It gives diversification and scale, but it also means the brand isn’t quite the pure top-end play that Sanlorenzo is. You are buying a portfolio rather than a single rifle shot at the largest boats.

The superyacht end, which is the part this thesis cares about most, is in good shape. Production capacity in the superyacht segment is virtually fully utilised out to 2029, which tells you the demand for the genuinely large, expensive boats is still there even as the cheaper end wobbles.

Now the honesty section, because Ferretti has more hair on it than Sanlorenzo. The first quarter of 2026 was clearly softer. New yacht turnover fell about 8% year on year, but the bigger worry was order intake, which dropped almost 34% to roughly 180 million euros, driven largely by geopolitical tension in the Middle East delaying orders and deliveries. The net order backlog slipped to 722 million euros from 829 million at the end of 2025. Margins held up, which is the saving grace. Adjusted EBITDA came in at 48.7 million euros versus 52.5 million a year earlier, so profitability proved resilient even as the top line softened.

The second issue is governance, and it is messier. China’s Weichai Group won a shareholder vote that ended the twelve year tenure of long-time CEO Alberto Galassi, while Czech investor KKCG appealed the validity of that decision, leaving the governance situation unresolved. Italy has also been weighing KKCG’s request for golden power scrutiny, partly because Ferretti owns a small non-core patrol boat division. A new global CEO, Stassi Anastassov, is now in place, but the ownership picture is genuinely in flux, and that is the single biggest reason this sits a notch below Sanlorenzo on the risk scale rather than at the same level.

So the trade-off is clean. Ferretti gives you the better brands, a broader footprint, and a slightly cheaper entry, in exchange for a softer recent quarter and real governance uncertainty. For a basket built around this theme, it earns its place precisely because of Riva, Wally and CRN, names that the post pandemic buyer actually wants, even if you size it a little smaller than the anchor to account for the noise.

Subscribe now

MarineMax, Inc. (HZO, NYSE)

MarineMax is the stock that gives you a completely different way to play this theme, and that is exactly what makes it interesting. The two Italian names are builders. MarineMax is everything around the boat. It bills itself as the world’s largest recreational boat and yacht retailer, marina operator and superyacht services company, with over 120 locations worldwide, including more than 70 dealerships and over 65 marina and storage facilities. If the Italians sell you the boat, MarineMax wants to dock it, broker it, finance it, insure it and service it.

The pieces that matter for this thesis are the ones at the top of the chain. Its superyacht services run through Fraser Yachts Group and Northrop & Johnson, two of the best known names in superyacht brokerage, both acquired over the past decade as part of a deliberate move upmarket. On the infrastructure side, its IGY Marinas operate luxury facilities across the Caribbean, the Mediterranean and major US yachting destinations. This is the part I find most attractive. Marinas and brokerage are the toll booths of the yachting world. As boats get bigger and more numerous, somebody has to berth them and somebody has to handle the sale, and MarineMax owns both functions. It is a way to bet on the activity around the boom rather than on any single hull selling.

Here is where my reservation comes in, and it is a real one. The financials are messy and the headline numbers are going the wrong way. Second quarter fiscal 2026 revenue came in at 527 million dollars, down more than 15% from a record 631 million a year earlier, with same-store sales down 15% against an 11% increase in the prior-year period. The company swung to a net loss of 2.6 million dollars, versus net income of 3.3 million the year before. Used boat sales were the weakest line, falling 44% year over year. This is not a clean compounder. It is a cyclical retailer having a tough stretch as the broader boat market normalizes.

The counterargument, and the reason it still earns a place, is the mix shift underneath the decline. The higher-margin businesses are exactly the ones tied to this thesis, and they are growing while retail shrinks. The superyacht services, marina, finance, insurance and parts operations grew both as a share of revenue and in absolute dollars year over year, even as boat revenue fell around them. That shift shows up in profitability. Gross margin actually expanded to 34.4% from 30% a year earlier, driven by the increasing contribution from those higher-margin businesses. In other words, the part of MarineMax that this whole article is about is the part that is working. The drag is the commodity retail boat business, which is the part you care about least.

There is also an outside chance of a corporate catalyst. The company received an unsolicited, non-binding indication of interest from The Donerail Group to acquire all outstanding shares for 35 dollars per share in cash, and that approach drew further reported interest from large private equity names. Whether anything comes of it is unknowable, but it puts a degree of attention on the valuation that a sleepier name would not have.

So my honest framing is this. MarineMax is not the pure play, and the income statement makes that obvious. You are buying a diversified marine conglomerate with a cyclical retail anchor dragging on a genuinely attractive luxury services and marina core. But it is one of the only listed ways to own the toll booths rather than the hulls, and in a field this thin on options, that uniqueness is worth something. I would treat it as the more speculative, mix-shift bet of the basket rather than a name you underwrite on current earnings.

Subscribe now

The Italian Sea Group S.p.A. (TISG.MI, Euronext Milan)

I want to be completely upfront here. The Italian Sea Group belongs in this article on paper, because it owns some of the most storied names in the business, but the company is currently in the middle of a serious crisis that I do not yet understand well enough to underwrite. I am including it for completeness and transparency, not as a recommendation, and I would rather tell you plainly what I do and do not know than pretend I have it figured out.

The brand portfolio is the reason anyone looks at it in the first place. TISG’s brands include Admiral, Tecnomar, Perini Navi, Picchiotti, NCA Refit and Celi 1920. Perini Navi in particular is a legendary name in large sailing yachts, which ties directly into that trend I mentioned earlier about big sailing yachts being the ultimate flex. In a normal world, this would be a pure-play builder at the very top of the market.

But this is not a normal situation, and the timeline of the last few months is alarming. In early 2026 the company launched an independent forensic investigation into unauthorised overspending, two weeks after taking a 25 million euro shareholder loan to shore up its cash position. Workers staged a two-hour strike in February after wages were delayed by eight days due to liquidity problems, and the board acknowledged the company had incurred significant extra-budget costs in executing orders. The chair and vice chair resigned, with founder and CEO Giovanni Costantino taking over as chair.

It got worse from there. The Italian Stock Exchange ordered the company’s exclusion from STAR status, a complaint was filed against former executives, and the company initiated a negotiated settlement procedure under Italian insolvency law. Then the real shock. In May 2026 the company announced that losses had reduced its share capital below the legal minimum, a material event under Article 2447 of the Italian Civil Code, and the shares fell more than 37% in a single day. The exact amount of the losses is still being determined, pending ongoing accounting reviews, and the forensic due diligence by KPMG has experienced delays. The stock, which traded above 8 euros at its 52-week high, has collapsed to roughly 1.2 euros.

There is also a darker backdrop that I am still trying to make sense of. TISG was the builder of the superyacht Bayesian, which sank off Italy in August 2024. There is now litigation swirling around that event, with the company reportedly pursuing legal action while a UK investigation raises unresolved questions about the yacht’s design and stability. How much of the financial crisis is connected to that, versus separate accounting and order-management problems, is exactly the kind of dynamic I have not yet untangled.

So my honest position is this. I do not currently hold it and I am not telling you to. The brands are real and the underlying market is the same booming market the rest of this article describes, which is precisely why a distressed, restructuring situation like this can become interesting to a certain kind of investor once the facts are clearer. But right now it is a forensic-investigation, capital-below-minimum, court-supervised-restructuring story, and anyone who tells you they know how it resolves is guessing. I am flagging it so you have it on your radar, and I will only form a real view once KPMG’s review is complete and the true scale of the losses is known. Treat this section as a transparent placeholder rather than a thesis.

Subscribe now

Grand Harbour Marina p.l.c. (GHM, Malta Stock Exchange)

The last name on the list is more of an honorable mention than a core idea, but it is worth including because it is about as pure a play as you can find anywhere in this space. Where MarineMax bundles marinas inside a sprawling retail conglomerate, Grand Harbour Marina is essentially just the marina, full stop.

The business is simple to understand, which is part of the appeal. The company owns and operates marina facilities providing berthing and ancillary services for yachts and superyachts, running through two segments: the Grand Harbour Marina in Malta and a stake in the IC Cesme Marina in Turkey. The Malta asset sits in the historic Grand Harbour next to Valletta, and it is built for exactly the kind of boat this whole article is about. The marina offers over 250 berths and 26 superyacht berths able to take vessels up to 135 metres. It is owned and managed in association with Camper & Nicholsons Marinas, a recognized name in the luxury berthing world.

This is the toll-booth idea in its purest form. As the trend pushes the Mediterranean toward more and bigger boats, the constraint that does not move is space in good harbours. Prime superyacht berths in desirable locations are genuinely scarce, the supply cannot be conjured up quickly, and the operator collects on occupancy and berth sales regardless of which shipyard built the hull. Malta has actively leaned into this, with the government pursuing a long-standing policy to make the island a superyacht-friendly destination. So in theory you are buying a slice of the dock itself, the chokepoint that every one of those 100 metre boats eventually needs.

Now the caveats, and they are significant. This is a micro cap in the truest sense. As of May 2026 the stock traded around 1.19 and the entire market capitalization was roughly 24 million dollars, on trailing twelve-month revenue of about 9.4 million. That tininess brings everything you would expect: the stock is thinly traded, listed on the Malta Stock Exchange rather than a deep liquid market, and any position is hard to build or exit without moving the price. The financial performance also partly depends on the timing and number of berth sales, which makes results lumpy from year to year. One offsetting positive worth noting is the income. The shares have carried a high dividend yield, recently in the double digits, though with a micro cap you always have to ask how durable a payout like that really is.

So I would not build a thesis around it, and I am not pretending it belongs in the same tier as the Italian builders. But if what you want is the single most direct, undiluted exposure to the act of parking a superyacht somewhere expensive in the Mediterranean, this is close to the only listed way to get it. Treat it as a curiosity and an honorable mention: a tiny, illiquid, pure-play bet on the scarcity of the dock, suitable only for someone who understands exactly what a micro cap entails and sizes it accordingly.

Subscribe now

Misc.

Beyond those names, I also looked into all of these:

Close Brothers Group plc (CBG.L, London Stock Exchange)
Garmin Ltd. (GRMN, NASDAQ)
Chugoku Marine Paints, Ltd. (4617.T, Tokyo Stock Exchange)
AkzoNobel (AKZA.AS, or OTC: AKZOY, Euronext Amsterdam)
PPG Industries (PPG, NYSE)
Gurit Holding AG (GURN.SW, SIX Swiss Exchange)
Hexcel Corporation (HXL, NYSE)
Toray Industries (TRYIY, or 3402.T, Tokyo)
Rolls-Royce Holdings (RR.L, London Stock Exchange)
OneWater Marine (ONEW, NASDAQ)
Catana Group (CATG.PA, Euronext Paris)
Fountaine Pajot (ALPAJ.PA, Euronext Paris)
Grand Banks Yachts (G50.SI, Singapore Exchange)

They are all related to the trend in some way, but most of them either fall outside the large yacht end of the market or are not pure play enough to earn a full write up. I am leaving them here in case anyone wants to go through them themselves. Think of them as an extension of the same thesis rather than the heart of it.

The connecting logic is worth spelling out, because it is really the second half of the story. The trend does not stop at billionaires buying mega yachts. It cascades downward. As social media makes the water look like the place everyone is supposed to be, the middle class and upper middle class start spending outsized chunks of their income to get a taste too. That is where most of these names come in. Instead of a 100 metre custom build, it is a financed boat, a chartered catamaran, a fractional share, a week split between friends. The money is smaller per head but the population is vastly larger.

You can see how each of these slots into that wider funnel. Catana and Fountaine Pajot build the cruising catamarans that increasingly fill the charter fleets, which is exactly the rental democratization I described earlier. OneWater Marine is another US boat retailer in the same vein as MarineMax, riding the broader ownership wave. Close Brothers touches marine lending and finance, the plumbing that lets people buy boats they could not pay for outright. Garmin sells the marine electronics that go on almost everything that floats. Rolls-Royce, through its MTU engines, sits in the propulsion of large yachts among many other things. Grand Banks builds at the more traditional motor yacht end.

Then there is a whole cluster tied to one specific input: advanced materials and marine coatings. Chugoku Marine Paints, AkzoNobel and PPG all supply the paints and coatings that every hull needs. Gurit, Hexcel and Toray supply the composites and carbon fibre that make large, light, structurally demanding builds possible, and this is the same carbon fibre story behind those big sailing yachts I mentioned earlier, where the structural engineering is the whole flex. These are the picks-and-shovels names, the trouble being that for almost all of them, marine is only a sliver of a much larger and unrelated business, which is precisely why they did not make the main list.

So treat this section as a research starting point rather than a set of recommendations. The pure-play, large-yacht-exposed names are the four I covered in depth. Everything here is a wider, more diluted way to bet that the entire pyramid, from the billionaire at the top to the upper-middle-class family chartering a catamaran at the bottom, keeps growing.

Subscribe now

The Thesis

So now the real question. What actually makes this opportunity interesting? Stating that the number of mega yachts and the amount of luxury on the water has increased over the past ten years does not, on its own, warrant a thesis. It is what happens from here that matters.

Right now the prevailing narrative in the field goes roughly like this. Things are going well, but we already saw the cyclical peak. Much of the recent spending gets attributed to covid and the liquidity that came with it. And the Iran crisis is read as a material dent in demand. Several of these points are true. But none of them actually weaken the real tailwinds driving the structural increase in demand.

Start with social media. Instagram is not going anywhere. If anything, as more and more younger, Instagram-native people start inheriting and earning serious money, the amount of capital flowing toward yachts increases rather than fades. Then layer on the demographic point from earlier. The number of tech-related millionaires and billionaires is still rising. And here is the part most people miss: essentially every yacht bought so far has been bought by pre-AI-boom wealth. The people getting rich off AI, or off something like SpaceX, have not even had time to start looking at yachts yet. That cohort is coming. Finally, as the industry matures, chartering becomes easier and more accessible than ever, pulling in an entirely new layer of demand from below.

Now look beyond the current tailwinds to what AI and robotics are likely to do to the global economy. I think two things are very probable. First, physical goods like yachts get cheaper to produce. If you have humanoids working around the clock on production lines, output stops being constrained by human labour capacity and becomes constrained mainly by energy. Second, and just as important for the long run, wealth keeps polarizing. As network effects and data moats compound across an AI-driven economy, a slice of society gets richer and richer. I am personally not very optimistic about the middle class in relative terms. So focusing on the very top of the pyramid, the mega yacht end, feels like the smarter approach rather than betting on the broad base.

To summarize what I think makes this opportunity real, there are two things happening at once. Most investors believe the sector is just seeing a short cyclical boom, fuelled by covid, that will fade. I think that read is wrong. The mechanism, as I see it, runs more like this: the more pictures and cultural relevance yachting accumulates, the more people are willing to overspend and pour their excess wealth into it. Yes, there may well be a local cyclical top. But I am having a genuinely hard time seeing how all this recent AI-driven wealth does not, at some point, find its way to mega yachts.

Subscribe now

End Note

My personal view is that this sector is likely to prove very durable over the long term. Demand should continue climbing higher and higher. But at the same time, it is vital to stay focused on brand, because without a strong brand the margins are impossible to defend. On that front the current valuations look compelling.

The one thing to keep in mind is that the wealth driving this demand is highly sensitive to broader market moves. So while I believe this sector is a great place to park money over the long run, these stocks are also likely to take a beating in any serious market downturn. The two things are not in conflict. You can love the structural story and still expect a brutal drawdown if equities roll over, because the marginal yacht buyer feels the stock market in his bones.

If you liked this writeup, please share it. It helps surprisingly much. And if you have any questions, just ask.

NFA. This writeup is for informational and entertainment purposes only and reflects my personal opinions, not investment advice. I am not a licensed financial advisor. Everything here may contain errors, and nothing should be taken as a recommendation to buy or sell any security. The companies mentioned are ones I find interesting, and I may hold positions in some of them. Always do your own research and consult a qualified professional before making any investment decision. You alone are responsible for your money.

This write up maps the companies positioned to capture that volume. The goal is to identify the biggest winners of the coming IPO boom, the institutions sitting at each choke point as private wealth converts to public liquidity.

Subscribe now

The volume moves in predictable stages. Shares must list, then clear and settle, then the proceeds must be advised, allocated, and in many cases spent on status goods. Each stage has a narrow set of institutions already positioned at the bottleneck. The question for each name is whether it captures more revenue from the added flow or whether fee pressure strips the gains away before they arrive.

One mechanism makes the point. When a single SpaceX-style listing triggers index inclusion inside fifteen trading days, passive vehicles have no choice but to buy. That forced purchase does not depend on sentiment or analyst upgrades. It is mechanical. The same mechanical quality runs through custody hand-offs, margin calculations, and the advisory conversations that follow. Once shares sit inside an established platform, extracting them means rewriting permissions, tax lots, and reporting frameworks that took years to build. That friction is the moat.

The new shares do not arrive as a blank slate. They land inside a permissioned surface that already knows the client’s full picture, processing larger checks at the same negotiated margins. The rest of this piece works through that surface stage by stage and names the stock at each one.

The bear case that has kept multiples depressed

The prevailing fear is straightforward. Advisors already expect to charge less than 1 percent on accounts above five million dollars, with average fees for clients above ten million dollars sitting near 66 basis points. Commission revenue across the industry is projected to drop from 22.8 percent of total revenue in 2024 to 16.6 percent by 2026. If the same pattern holds after the largest IPOs in history, the incremental AUM may simply arrive at lower unit economics.

A second concern compounds the first. One marquee deal that slips or disappoints could trigger a broad de-rating across any name tied to capital markets activity. Financials have already experienced multiple compression even while earnings growth has remained resilient. Platforms that rely on listing fees or episodic trading spikes face the additional risk that the entire wave arrives later or smaller than current pricing assumes.

I have watched the same logic applied to earlier cycles. When valuations for anticipated mega-deals compress in advance, secondary platforms and wealth franchises often de-rate together even if their core custody or clearing volumes stay flat. The fear is not irrational; it simply assumes the new wealth will behave like ordinary inflows rather than concentrated liquidity events that force re-pricing across multiple product lines at once.

The compression numbers themselves come from advisor surveys that predate the current IPO calendar. They reflect client pushback on value delivered, not an absence of assets. That distinction matters when the assets arrive in sudden, outsized blocks rather than steady drips.

Why the same shift expands rather than destroys the revenue pools

History offers a clearer template. When online trading reached 37 percent of retail equity volume by 1998, many assumed full-service brokers would lose their clients. Merrill Lynch still held 1.5 trillion dollars in private-client assets by the end of 1999 while the exchanges recorded record trading volumes. The new channel did not replace the old one; it expanded total activity and left the largest incumbents with larger balance sheets.

The ETF transition followed the same pattern. Active managers were expected to lose share permanently. Instead the largest asset managers acquired ETF platforms early and now control 74 percent of U.S. equity ETF assets, with BlackRock’s iShares alone exceeding 5.4 trillion dollars. In both cases the infrastructure that moved the assets captured the economics of the shift rather than watching them migrate elsewhere.

The 1990s tech IPO wave produced the same result at even larger scale. Newly public founders routed liquidity into wealth-management relationships and alternative allocations. Exchanges collected elevated listing and trading fees while overall assets under management expanded across the system. The mechanism is identical today: the banks that underwrite the IPOs already sit inside the syndicates, the exchanges already run the index-inclusion rules, and the wealth platforms already hold the custody relationships.

What changed this time is speed and scale, and the first proof point has already printed. SpaceX priced on June 11 and began trading on the Nasdaq on June 12 under the ticker SPCX, raising 75 billion dollars at a 1.77 trillion dollar valuation — the largest listing in history, run through a syndicate of nearly every major bank with Goldman Sachs and Morgan Stanley as leads. Those two alone earned a combined 500 million dollars in underwriting commissions on the deal. The stock opened at 150, closed its first day up 19 percent, and has since climbed toward 200, carrying the market value above 2.5 trillion dollars within a week. OpenAI and Anthropic have both since filed confidential S-1s of their own — Anthropic on June 1, OpenAI on June 8 — queuing up the next two legs of the same wave. The fee pool from each deal lands inside institutions that also own the downstream wealth and alternatives franchises. The same entities that price the IPO stand to advise the proceeds.

The parallel that keeps surfacing is how regulatory gates concentrate rather than disperse the flow. Every new issuer must still file through established underwriters and list on venues with fast-entry index rules already in place. Those gates do not multiply with each wave; they simply process larger volumes at the same structural margins.

Subscribe now

Here is the map

The flow follows five linked clusters. Exchanges and listing venues collect fees at the moment of entry and index inclusion. Brokers and trading platforms monetize the subsequent order flow, custody, and margin. Wealth and private banks capture recurring advisory fees once the proceeds sit in client accounts. Alternative asset managers receive fresh allocations into credit, real estate, and private equity. Luxury houses capture the high-margin, discretionary purchases that follow large liquidity events.

A single founder who sells five percent of a newly public company can trigger activity across every cluster in sequence. The same institutions handle each step because the regulatory, compliance, and relationship layers required to move that capital at scale already sit inside them.

Permissioned surfaces offer a useful frame rather than products. An issuer cannot reach index inclusion without choosing a listing venue that already maintains fast-entry rules. A newly liquid shareholder cannot move large blocks without a prime broker that already holds the necessary regulatory permissions. Those surfaces do not multiply when new wealth arrives; they simply process higher volume at margins that were already negotiated years earlier.

The same surfaces also determine where value pools. Once a client relationship includes both private-share custody and public-market execution, the cost of switching rises because the new provider must replicate tax-lot history, margin agreements, and reporting frameworks simultaneously. That replication cost grows with the size of the liquidity event, not the number of providers.

Exchanges and listing venues

$NDAQ operates the primary listing venue and index ecosystem that captures both the initial listing fee and the data revenue that follows. The street narrative treats it as a straightforward beneficiary of any IPO wave, with recurring revenue from listings and market data already priced in. What actually happens is more structural. Nasdaq’s fast-entry rule now places any top-40 market-cap company into the Nasdaq-100 within fifteen trading days of listing. Passive funds must buy the shares. That forced buying raises switching costs for issuers who want index visibility and strengthens the exchange’s pricing power over data and listing packages.

$ICE runs the NYSE and a diversified set of trading and data businesses. The consensus view sees narrower direct exposure to pure-tech IPOs. The reality is steadier: listing fees and data subscriptions compound with any expansion in equity wealth, and the exchange already clears and settles the majority of U.S. securities volume. Both venues sit at the first mandatory choke point.

The parallel worth watching is how index inclusion turned a one-time listing fee into a permanent rebalancing flow. Once the shares enter the Nasdaq-100, every passive vehicle rebalances on a schedule that cannot be avoided. That recurring mechanical demand sits outside the control of either the issuer or competing venues.

The same rule set also limits substitution. An issuer seeking immediate passive ownership cannot simply choose a venue without equivalent fast-entry mechanics; the index methodology itself channels flow to the established player.

Brokers and trading platforms

$HOOD monetizes retail trading volume and new account formation. The prevailing narrative positions it as a high-beta play on IPO excitement. Volume-driven revenue does rise directly with wealth-rotation events, yet the platform’s low switching costs and thin differentiation leave durability lower than pure listing venues.

$IBKR scales commissions and margin lending with trading activity from newly wealthy clients. Its global reach and sophisticated order-routing technology give it a durable position in institutional flow that pure retail platforms lack. The same technology stack that routes retail orders also handles the more complex crosses required when large employee liquidity events hit the tape.

$SCHW combines retail brokerage scale with custody and now private-market access after acquiring Forge. The street sees it as a steady wealth winner. The acquisition embeds the firm deeper into pre-IPO flow, raising client stickiness at the exact moment liquidity events accelerate. Once a client’s private shares sit inside the same interface that already holds their public brokerage account, moving the relationship requires replicating both the private-market data feed and the integrated tax reporting. That integration raises switching costs precisely when the volume of such clients is about to increase.

The competitive force at work is substitution threat. A pure retail platform can lose flow to any lower-cost interface, while an integrated custody-plus-private platform forces any competitor to rebuild the entire data and compliance layer before the client can move.

Wealth and private banks

$MS wealth-management and investment-banking arms both stand to gain. The consensus story emphasizes direct exposure to IPO and private-market activity. Morgan Stanley’s acquisition of EquityZen extends that reach into secondary trading of pre-IPO shares, raising barriers around client relationships that competitors without similar platforms cannot easily replicate. The same client who once traded private shares on a standalone venue now does so inside an account that also offers margin, estate planning, and direct underwriting access.

$GS private-wealth franchise serves ultra-high-net-worth tech clients through the same relationships that lead the underwriting syndicates. The integrated model keeps the firm inside both the fee pool at issuance and the ongoing advisory relationship. When a founder liquidates a portion of a newly public stake, the conversation about allocation often begins with the banker who already knows the cap table.

$BAER focuses on Swiss UHNW advisory directly tied to liquidity events. Earnings pressure and regulatory overhang have created a valuation gap, yet the structural link to cross-border tech wealth remains intact. Swiss private banks have long served as the default venue for non-U.S. founders seeking stable, multi-currency wealth structuring; that preference does not disappear simply because the shares become public.

What strengthens these platforms is the accumulation of context. Each liquidity conversation adds tax-lot history, family-office preferences, and allocation patterns that later advisors must replicate from scratch. The barrier is not the product; it is the accumulated permissions and memory inside the relationship.

Alternative asset managers

$BX receives allocations into credit, real estate, and private-equity funds from newly liquid wealth. The consensus positions it as the prime vehicle for deployment. Scale advantages in raising and deploying large mandates offset fee compression elsewhere; the platform that can absorb the largest checks tends to keep them.

The same logic that drove 1990s founders into alternatives applies here at larger dollar amounts. A single liquidity event can generate nine-figure checks that only the largest alternative platforms can deploy without moving market prices against themselves. That capacity becomes a self-reinforcing barrier: larger checks flow to the managers already equipped to receive them.

The force at work is barriers to entry through deployment capacity. Smaller managers cannot absorb the same check size without market impact, so the largest platforms receive the flow by default. That advantage compounds with each successive wave rather than eroding.

Luxury status beneficiaries

$RMS, $MC, and $RACE sit at the final stage where status consumption occurs. Hermès and LVMH capture high-margin discretionary purchases; Ferrari does the same at the ultra-luxury end. The spend is not core to any of them, yet it is recurring, high-margin, and directly linked to liquidity events that have no natural ceiling in the near term.

These names function as the final confirmation layer rather than the primary driver. When newly liquid capital seeks visible signals of arrival, the marginal purchase tends to land at the scarcest brands. That behavior is culturally sticky and price-insensitive in the upper tail, which is exactly where concentrated IPO proceeds concentrate.

The parallel here is simple. Status goods do not require ongoing relationships or regulatory permissions; they require only recognition. That recognition already exists inside the brands that survived prior wealth waves, so incremental spending flows to them without new competitive entry.

Subscribe now

Safer anchors versus asymmetric bets

The more durable compounders sit where switching costs and regulatory barriers are highest. $SCHW, $ICE, and $NDAQ combine scale with structural mandates that new entrants cannot easily displace. $RMS adds pricing power through brand scarcity that survives macro shocks.

Higher-beta expressions include $BX, $MS, and $BAER, where inflows or execution can move results materially but where outcomes also depend on the timing and size of the actual IPO proceeds. $HOOD offers the purest activity beta yet the lowest durability if competition intensifies.

The distinction matters because the mechanical flows I described earlier—index inclusion, custody hand-offs, margin calculations—land first inside the platforms that already maintain regulatory licenses and data integrations. Those platforms are the anchors. The higher-beta names capture the variable portion of the same flow but must re-win the client at each liquidity event.

The same logic separates the two groups on pricing power. Anchors can maintain or even widen margins because clients face real replication costs; higher-beta names compete more directly on price or convenience and therefore absorb more of the documented fee compression pressure.

The bet

This is no longer a hypothetical. SpaceX is already public — it listed on June 12 at a 1.77 trillion dollar valuation and now ranks among the largest companies in the country, with its fifteen-trading-day path into the Nasdaq-100 already running. OpenAI and Anthropic have both filed confidential S-1s and are queued behind it. Index-inclusion mechanics have already been updated to accelerate mega-cap entries. The underwriting fee pools alone run into hundreds of millions per deal. Yet the market continues to price the downstream infrastructure as if that liquidity might never arrive or might arrive at permanently lower margins. The historical pattern and the concrete mechanics already in place suggest the opposite outcome.

The gap between the mechanical triggers already coded into index rules and custody systems and the valuation multiples that still reflect fear of fee compression alone stands out as most asymmetric. The compression is real, yet it applies to a larger base of assets that must travel through fewer hands. That combination has produced durable revenue expansion in every comparable cycle I have studied.

The bet therefore rests on sequence rather than sentiment. The shares must list before they can be advised, allocated, or spent. Each mandatory step already routes through institutions whose permissions and integrations predate the current wave. Those institutions do not need to win new clients; they need only process larger volumes inside relationships that already exist.

Not financial advice. This content is for informational and educational purposes only and reflects my own opinions, not a recommendation to buy or sell any security. I am not a licensed financial advisor. Do your own research and consult a qualified professional before making any investment decision. I may hold positions in the securities discussed.

The mechanics are straightforward once the jargon is stripped away. Agents cannot operate in a vacuum. They need permissions that survive employee turnover, audit trails that regulators can follow, and a single source of truth that does not reset every time a new model appears. The company that already owns those controls inside an enterprise gains a structural advantage. New point solutions must fight for access; the existing orchestration layer simply extends its reach.

Subscribe now

Historical precedent shows the pattern. When Adobe replaced perpetual licenses with Creative Cloud subscriptions, revenue climbed from $4.4 billion to $19.4 billion while recurring revenue crossed 90 percent. The model did not weaken. It became stickier because the customer’s work now lived inside a persistent environment. Microsoft’s move from on-premise servers to Azure produced the same outcome: a $75 billion run rate at 39 percent growth even as legacy server revenue declined. Incumbents that control the environment where work happens tend to expand total value rather than lose it.

Genesys offers another concrete parallel. Its PureCloud platform migrated contact-center customers from on-premise licenses to cloud delivery and recorded nearly 530 percent year-over-year revenue growth in the first three quarters after the shift, with customer wins rising 115 percent. The feared displacement never materialized because the new model concentrated governance and workflow execution inside one governed system. The same logic applies when agents replace seats: the layer that already enforces compliance and maintains context captures the expanded activity.

Consider what happens when the number of actors inside a system multiplies by an order of magnitude. An enterprise that once tracked a few hundred human users now tracks thousands of agents that each carry their own permissions, decision histories, and cross-references to other agents. The cost of fragmentation rises sharply. A single missed audit trail or conflicting memory state can invalidate an entire quarter of automated work. That requirement does not favor lightweight newcomers. It favors the systems already wired into the enterprise’s compliance fabric and data backbone. The shift therefore does not flatten value; it concentrates it where the rules and the records already live.

The feared collapse

The bear case starts from a simple observation: one agent can replace dozens of licensed users. Seat-based ARR therefore looks fragile. If pricing shifts to $2 per conversation or $0.99 per resolved ticket, the predictable cash flow that justified high multiples disappears. Gross margins that sat at 78 to 82 percent could compress into the low sixties once inference and observability costs run through every workflow. The market has already acted on that fear.

Between January and February 2026 roughly $2 trillion in software market value vanished. Atlassian shares fell 35 percent after the company reported enterprise seat counts declining for the first time in its history. Salesforce dropped 28 percent on similar concerns. The SaaS Capital Index median ARR multiple compressed from 7.0x to 3.8x even though many companies continued to post double-digit revenue growth. Valuations now embed the assumption that the underlying business model itself is broken.

ServiceNow illustrates the dislocation cleanly. Subscription revenue grew 21 to 22 percent year over year, yet the stock halved from its 2025 peak. Investors priced the risk that agentic automation would erode the very workflows the platform monetizes. The narrative treats consumption pricing as inherently more volatile and less valuable than the seat model it replaces.

The same compression hit broader indices. The iShares Expanded Tech-Software ETF fell more than 21 percent year to date and roughly 30 percent from its September 2025 peak by March 2026, even as names such as Adobe continued reporting positive revenue growth. The selloff priced in model obsolescence rather than adaptation.

Salesforce revenue growth slowed to a consensus 9.6 percent for fiscal 2026 while the stock traded down more than 26 percent year to date and over 40 percent from its highs. Public SaaS companies reported AI-driven gross margin resets from the 78-82 percent range into the 60-70 percent band as inference, evaluation, and observability costs scaled with every automated workflow. The sector de-rating exceeded historical shocks even though fundamentals showed resilience in top-line expansion. The prevailing view treats these margin pressures and seat declines as terminal rather than transitional.

Subscribe now

The orchestration layer

The same forces that threaten per-seat revenue strengthen the platforms that already manage identity, permissions, and memory. Agents do not roam freely inside large organizations. They are forced through controlled environments because the cost of mistakes is too high. Microsoft Entra Agent ID automatically assigns unique identities to agents built in Copilot Studio or Azure AI Foundry. ServiceNow’s Knowledge Graph and CSDM extensions give agents a shared, auditable view of enterprise processes. Once agents operate inside that layer, switching costs rise because the accumulated context and governance rules cannot be ported without recreating months or years of work.

Consumption pricing on top of these platforms expands revenue per workflow rather than shrinking it. Salesforce Agentforce reached $800 million in ARR, up 169 percent year over year, with 60 percent of bookings coming from existing customers expanding usage inside the CRM data layer. Intercom Fin hit nine-figure ARR at $0.99 per resolved ticket. Hybrid subscription-plus-usage models have shown 31 to 38 percent higher growth than pure subscription arrangements. The agent does more work, the platform records more activity, and the customer pays for outcomes rather than headcount.

The thought experiment clarifies the mechanism. Imagine an IT department that once managed 5,000 human seats and now manages 50,000 agents. Every agent must pass through the same permissions and memory layer to avoid compliance failures. That requirement concentrates value in the platform that already owns those controls. The alternative—scattered point solutions—creates audit gaps and version conflicts that enterprises will not tolerate. Data gravity therefore pulls agents inward, not outward.

Adobe and Microsoft supply the historical parallel. Both faced the fear that their old models would collapse. Both absorbed the shift and grew total revenue substantially. The orchestration layer that survives does not merely replace seats; it becomes the place where the new volume of agent activity is governed and monetized. ServiceNow’s May 2025 AI Platform launch added an AI Agent Fabric for cross-vendor coordination with Microsoft, NVIDIA, Google, and Oracle, deepening the same gravitational pull inside one governed system. ServiceNow’s April 2025 Cisco partnership further embedded secure AI adoption at enterprise scale. The pattern repeats: governance requirements and integration depth turn the feared volume increase into a moat rather than a threat.

Here is the map

A quick word on where these names come from, because they are not cherry-picked. Before writing a word of this piece I mapped the entire theme: 607 companies screened across 218 value-chain nodes and 985 supplier-customer edges, with all 244 public names scored on theme exposure and criticality. Fourteen made the cut. The chart below is the whole field — the grey cloud is everything that touches the theme, and the gold dots past the gate are what survived.

The agent value chain breaks into four clusters. Workflow orchestration platforms embed and run teams of agents inside existing enterprise processes. Vertical process platforms add agentic capabilities to industry-specific data and compliance requirements. Billing and observability layers capture usage volume and monitor agent behavior at scale. Data and decision platforms supply the persistent memory and long-running logic that agents need to operate across time. Value pools where governance, memory, and workflow execution already intersect because those are the constraints that become more binding as agent counts rise.

Walmart’s experience with e-commerce offers a useful analogy for how the clusters interact. Traditional retailers feared pure online players would strip away margin. Instead, the companies that already controlled physical inventory, customer data, and fulfillment absorbed digital channels and grew omnichannel revenue. The same pattern appears here: the platforms that already sit at the intersection of workflow, compliance, and data capture the incremental agent activity rather than losing it to new entrants.

The four clusters are not interchangeable. Workflow orchestration carries the highest switching costs because agents inherit the process maps and exception rules already built inside the system. Vertical platforms add domain-specific barriers that general tools cannot replicate without recreating regulatory logic. Billing layers convert every additional agent action into a metered event. Data and decision layers supply the memory that prevents agents from hallucinating or losing context after reorganizations. The clusters reinforce one another: an agent that runs inside a governed workflow also generates billable events and requires durable memory. The company that owns two or more of these layers compounds the advantage.

Workflow Orchestration Platforms

$NOW ServiceNow runs the workflows that IT, HR, and customer service organizations already use. Its platform now hosts pre-built teams of agents that execute inside the same governed environment. The consensus view treats the stock as exposed to seat erosion; the shares halved from the 2025 peak despite 21 to 22 percent subscription growth. What is actually happening is that governance mandates and the Knowledge Graph requirement force agents into ServiceNow’s orchestration layer rather than allowing them to operate independently. Switching costs rise because the accumulated process context and audit trails sit inside one system. The business profile strengthens.

$CRM Salesforce positions Agentforce as the operating system for agentic work on top of its CRM data. The street narrative sees slowing core growth offset by AI monetization hopes, with the stock well below 2025 peaks. The concrete development is that Agentforce ARR reached $800 million with 60 percent of new bookings from existing customers expanding inside the trusted data platform. Salesforce processed 19 trillion tokens and delivered 2.4 billion Agentic Work Units across Agentforce and Slack. Agents increase the volume of activity that must be recorded and governed; the platform that already owns the customer records captures that expansion. Pricing power moves from seats to outcomes without losing the data gravity that keeps customers inside.

$PATH UiPath converts its RPA foundation into the execution fabric for autonomous agents. Consensus views it as legacy automation threatened by newer AI tools. In practice, agents multiply the number of workflows that require reliable orchestration and auditability. The automation layer becomes the default runtime for those workflows, raising switching costs and platform value rather than eroding them. Enterprise deployments that once required separate tools now converge on a single governed runtime because compliance and exception handling cannot be fragmented.

$PEGA Pegasystems supplies low-code workflow and decisioning platforms used for CRM and ERP processes. The market narrative often treats it as a mature vendor in a crowded field. What the agentic shift actually does is embed decision agents directly into existing case management flows. Switching costs increase because the accumulated business rules and outcome histories live inside the platform; moving agents elsewhere requires rebuilding those rules from scratch. Barriers to entry rise because new entrants lack the production decisioning logic already tuned to regulated processes. The profile strengthens as agent volume scales inside the same low-code environment.

$HUBS HubSpot provides CRM and marketing automation to mid-market companies. The consensus frames the business around seat-based marketing and sales tools. In reality, its workflow automation directly supports agentic actions such as lead qualification and campaign orchestration. The platform that already owns customer data and campaign history becomes the natural home for agents that act on that data. Switching costs rise because agent actions inherit the existing lead scoring and nurture sequences. The business strengthens as agents increase the frequency and precision of those sequences without requiring new integrations.

Vertical Process Platforms

$DSY Dassault Systèmes supplies 3D and PLM environments where aerospace and automotive teams already model products. Agentic simulation and generative design agents now run inside the same model. Domain-specific data gravity intensifies because regulatory and engineering compliance rules are embedded in the platform; agents cannot bypass them without creating certification risk. The business strengthens as the volume of simulated work increases inside a single governed system.

$SGE Sage Group delivers accounting and HR software to small and mid-sized businesses. Agentic bookkeeping agents handle ledgers natively. SMB customers lack the resources to stitch together multiple point solutions, so the platform that already manages compliance and financial records absorbs the new agent activity. Switching costs remain high because the ledger history and regulatory filings sit inside one environment.

$TNE Technology One provides ERP systems to public-sector and higher-education institutions. Agentic workflow agents manage compliance processes that carry statutory requirements. The vertical specialization creates barriers that general platforms cannot easily replicate; agents must respect those rules or the output becomes unusable for audits. The profile improves as agent volume grows within the existing compliance layer. Removing this platform would stall agentic ERP rollouts at more than 200 institutions that rely on its specialized statutory controls.

$TEMN Temenos supplies core banking SaaS to tier-two banks. Its Infinity platform now includes agentic advisory and operations modules. The market narrative positions it as a traditional banking vendor. What actually occurs is that regulatory reporting and customer advisory rules are already codified inside the system. Agents that advise on accounts or execute operations must operate within those rules or produce outputs that fail audits. Switching costs therefore rise sharply; a new platform would require re-encoding decades of compliance logic. The business profile strengthens because agent volume increases inside a vertical moat that general orchestration layers cannot easily replicate.

$APPN Appian supplies low-code BPM platforms for enterprise process automation. Consensus views it as one of several BPM vendors. The agentic shift embeds agent orchestration directly into existing case and workflow engines. The platform that already owns process definitions and exception paths becomes the execution layer for agents that follow those paths. Barriers to entry increase because accumulated process knowledge cannot be ported without recreating the underlying logic. The profile strengthens as agent counts scale inside the same governed BPM environment.

Billing & Observability Layers

$ZUO Zuora operates the billing infrastructure that converts usage into revenue. Consumption pricing for agents directly increases the number of billable events the platform records. The consensus treats the company as niche subscription software. The reality is that hybrid models combining subscription with usage have already demonstrated 31 to 38 percent higher growth when agents replace seats. Pricing power moves in Zuora’s favor because every additional agent conversation becomes a metered transaction inside its system.

$DT Dynatrace supplies observability across enterprise stacks and has integrated explicit agentic monitoring capabilities. As agent counts rise, the volume of actions that must be tracked and explained increases. The platform that already sits inside existing monitoring workflows captures that incremental activity without requiring new integrations. The business strengthens because observability becomes a non-discretionary requirement rather than an optional add-on. Enterprises cannot tolerate unexplained agent decisions in regulated environments; the layer that already provides that visibility absorbs the new demand.

The interaction between billing and observability creates a reinforcing loop. Every agent action that generates a billable event also generates telemetry that must be explained. The company that owns both layers captures the full margin on both the usage and the explanation. New entrants would need to replicate both the metering engine and the monitoring fabric simultaneously, raising the capital and integration cost of displacement.

Share

Data & Decision Platforms

$PLTR Palantir’s AIP platform runs persistent agents on enterprise data. The core value is long-running logic that maintains context across organizational changes. Agents require exactly this form of durable memory and decision auditability. The platform that already owns the data foundation becomes the place where those agents operate, raising barriers to substitution. The market narrative often frames the stock around government contracts, yet the same memory and decision layer applies directly to commercial agent deployments where context must survive reorganizations.

$ESTC Elasticsearch provides vector memory and Agent Builder tools that let customers construct custom agents on top of search and observability data. The market narrative positions the stock as a data foundation that benefits from AI traction yet trades well below highs. What is actually happening is that agentic workloads need reliable context retrieval and reranking; the layer that supplies those capabilities sees increased usage as agent volume grows. The profile strengthens because memory and context become central rather than peripheral. The Q4 FY2026 launch of Agent Builder and Elastic Workflows explicitly targets this requirement.

Subscribe now

Safer anchors versus asymmetric bets

$NOW and $CRM sit on the durable side of the risk curve. Both already operate at large scale inside regulated environments where governance requirements are non-negotiable. Agent volume expands inside platforms that customers cannot easily exit without recreating years of process history and compliance records. The valuation compression prices the opposite outcome. ServiceNow’s partnerships with infrastructure providers and Microsoft’s Entra identity layer further lock agents into these controlled environments.

$PATH and $ZUO represent higher-leverage expressions of the same shift. UiPath’s automation layer becomes the runtime for thousands of new agent workflows; Zuora’s billing engine scales directly with conversation volume. Both carry smaller current footprints than the largest orchestration platforms, so the upside depends on execution in a still-forming market. The asymmetry is larger, but so is the sensitivity to adoption timing and competitive response. The same logic applies to $PLTR and $ESTC, where persistent memory and context retrieval become non-discretionary as agent counts scale.

$PEGA, $APPN, and $TEMN occupy an intermediate position. Each owns vertical or process-specific logic that general platforms cannot replicate quickly. Their smaller scale relative to the largest names creates more operating leverage if agent adoption accelerates inside their domains, yet they still benefit from the same governance and memory requirements that protect the larger anchors. The risk curve therefore runs from the broad orchestration platforms at the low end to the specialized vertical and decisioning platforms at the higher end, with the common thread that all of them already sit inside the constraints agents cannot bypass.

The bet

The platforms that absorb agents into governed, persistent systems of record capture more value per workflow than the seat-based model ever did. Current valuations price the opposite outcome. The dislocation that erased $2 trillion in market value between January and February 2026 created the entry point. The companies that already own the controls agents cannot bypass will compound from here. The historical pattern from Adobe’s subscription transition, Microsoft’s Azure pivot, and Genesys’s cloud migration shows that the layer enforcing governance and memory expands rather than contracts when the underlying work model changes.

The asymmetry is straightforward once the mechanics are clear. The market has marked down the entire sector on the assumption that volume will migrate to new, lighter tools. The evidence from identity mandates, compliance requirements, and data gravity points the other way. Agents multiply the number of actions that must be governed and remembered. The platforms that already perform that governance and memory absorb the multiplied activity at higher switching costs and with consumption-based pricing that scales revenue rather than caps it. That gap between the priced-in fear and the mechanics of agent deployment is the opportunity.

Emil Hartela Investing is a reader-supported publication. I map bottlenecks and themes like this one; supply chains, pure plays, and the names the market hasn’t connected yet.

Disclaimer: This is my own research and opinion for informational purposes only — not financial advice, and not a recommendation to buy or sell anything. Many names here are small- and micro-cap or non-US listed, which carries elevated risk. I hold positions in some of them, so assume I am biased. Figures may contain errors or be out of date. Do your own research. DYOR. NFA.

In the May edition of the Haypp Group web traffic breakdown we can see some abnormally large growth. For those new to the series, this is where I track Haypp's websites to front run earnings and generate alpha. The correlation between web traffic and revenue is extremely high. Let's dive right into it.

Core — Mature Nordic

Core markets are showing some pretty flat behavior for Q2. There isn’t really anything super special going on here. Last quarter was a surprisingly strong period for Haypp core markets and some moderate to flat growth is likely where we will land in Q2. It could even be slightly down depending on how June develops. But to sum it up, this quarter for core markets is stability following the rapid growth seen in Q1.

snus.com — Norway

Read more

Subscribe now

The Markets

Today’s markets are showing more and more signs of being in an unsustainable place. Anything connected to semis and AI has either run up several hundred percent in a few years, or it is trading at all time lows as the market decides entire categories, SaaS being the obvious one, are about to be made worthless. Rarely have I seen so much conviction priced into so many names in both directions at once.

My investing style has always been contrarian to some extent. Yes, valuation and cashflow are key; however, so is the market’s perception of reality. I don’t believe for one second that good investing comes down to being better at analysing how fundamentals are going to develop. Instead the value lies in the divergence between where fundamentals will actually go and what everyone else thinks. That gap is where the money is, and right now it is unusually wide.

The reason it is so wide comes down to one thing the market is still struggling to price; for the first time we are not automating physical work, we are automating cognition itself. The last great boom, the internet, connected everything but never really moved the productivity needle. This time is different. I have a hard time seeing how disconnecting thinking from humans doesn’t lead to enormous productivity gains long term, and an equally hard time seeing how it doesn’t reshape who, and what, actually holds value in the economy. That second part is the one almost nobody is underwriting properly.

Layered on top of that is a world coming apart at the seams. I don’t think many price the Iran disruption and global unrest realistically. Europe is struggling, key value chains that feed into Asia are severely weakened, and trust is deteriorating globally. So we have the single most powerful productivity force in a century colliding with the most fragile geopolitical backdrop in decades, and the market is trying to hold both in its head at the same time. It isn’t doing it well.

From this worldview I will walk through every position I hold, have held, and am considering holding; which names are insulated from all of this, which ones quietly benefit, and the one or two that frighten me more than I would like to admit.

Read more

This post is about those companies, and specifically the hidden ones found in Finland.

Unlike their counterparts elsewhere in the world, these names have traded flat for the past few years, some of them down. But they sit on the same bottlenecks. And if the buildout plays out the way the rest of the sector is already pricing in, they have the same potential to see large gains as the spending works its way down the chain.

A note before we start: this is a survey of a cluster I find interesting, not a set of price targets. Some of the international names these get compared to have run well ahead of their fundamentals, and I’ll flag where the Finnish ones have real warts too. The point isn’t that these are guaranteed to follow. It’s that the attention has rotated through the US, Sweden, and Germany and largely skipped Finland, despite the same structural setup.

Canatu: the pick-and-shovel inside the lithography machine

Start with the one I’ve done the most work on. Canatu is a Vantaa-based deep-tech company that’s been quietly building since 2004. That long, boring history is the whole reason nobody owns it: it spent most of its life in the lab, and it’s only now hitting the point where the science turns into revenue. The market hasn’t repriced it because, until recently, there was nothing to reprice.

The first-principles case is simple. Making advanced chips runs through one chokepoint: EUV lithography machines, built only by ASML. Inside each machine sits a consumable part that protects the most expensive component, and as the machines get more powerful, the old version of that part stops working. Canatu makes the technology for the new version. So without getting into the chemistry, the position is this: the AI buildout needs chips, the chips need these machines, the machines need this part, and the current materials are hitting a wall. Canatu sits exactly on that wall.

What makes it a business rather than just a clever product is the model. Canatu doesn’t manufacture and ship the part itself. It sells a reactor to a partner, sells the proprietary inputs that only it can supply, and then collects a royalty on every single unit the partner produces. One sale turns into three revenue streams, and the best one, the royalty, recurs forever with almost no cost or manufacturing risk attached. They carry none of the capital burden of building factories around the world; the partner does. Canatu just owns the IP and clips a fee off everything that comes out. That’s the kind of capital-light, high-margin, recurring structure that, if volumes scale, becomes very valuable very fast.

It’s not theoretical anymore. The model went live in October 2025 when a Korean partner, FST, took a commercial license. The interesting wrinkle: Samsung owns a stake in FST. So Canatu is effectively plugged into Samsung’s supply chain without ever appearing as a direct supplier, which is about as good as it gets for a company this size.

Now the honest part, because the bull case usually overstates it. Canatu is not the only player here. Mitsui, the established incumbent, is developing a competing version of the same part with its own backing. So the right way to think about Canatu is not “irreplaceable monopoly.” It’s the other door. The entire chip supply chain relying on one supplier for a critical consumable is exactly what makes the big fabs nervous, nobody wants a single point of failure on something that gates their most expensive equipment. Canatu is the credible independent alternative, and the only one offering partners a way to make this in-house under license rather than buying from a competitor. In a market this strategic, being the trusted second source isn’t a consolation prize. It’s a structurally excellent place to sit, and the licensing model means even winning a modest share compounds into real money over time.

A few things to keep honest. The payoff got pushed from 2027 to 2030, so this is a multi-year hold, not a quick flip. The company still loses money (2024 revenue was €22M, up 62%, but it ran a €5M net loss) because it’s spending to build capacity, which is normal at this stage but demands patience. And the share count is the detail most holders miss: thanks to how the company came public, more shares get issued as the price rises, so the realistic fully-diluted figure is closer to 44-46 million than the ~34 million your broker shows. If you’re doing the math, use the bigger number.

The most recent signal is the one I’d weight most. In May, Dr. Maximilian Slawinski took over as CEO, coming from Soitec, a company whose entire competence is taking advanced semiconductor materials from “we can make a little” to “we can make a lot.” That is exactly the problem Canatu has to solve next. Moving from one-off deliveries into real commercial volume is a different skill than inventing the technology, and they just hired someone who has done precisely that. For a company at this specific inflection, it’s the right hire at the right time.

Aspocomp: the tiny re-shoring bet

Aspocomp is a different kind of idea, smaller, scrappier, and riskier, but it sits in front of a real tailwind that the big banks have started naming out loud.

The first-principles case starts one layer away from the chips themselves. Everything in an AI server has to sit on something, and that something is a printed circuit board. As the boards get denser and more complex to handle AI workloads, they’ve quietly become a constraint of their own. Goldman’s spring 2026 supply-chain work flagged PCBs and the laminates underneath them as one of the genuine bottlenecks in the buildout, with prices climbing 30%-plus a year and shortages they don’t expect to clear before 2027. So this isn’t a Finnish curiosity. It’s a global chokepoint that a major bank has put in writing, and Aspocomp is one of the few European names that plays in it.

Here the interesting part isn’t really the company, it’s the geography. Almost all of this production sits in Asia. Europe has been trying, through the Chips Act and now its successor, to drag critical supply chains back onshore so the continent isn’t fully dependent on imports for the things that matter most, defense electronics, semiconductors, aerospace. A PCB maker with an actual European plant and defense and semiconductor qualifications is exactly the kind of supplier that policy is designed to favor. That’s the core of the bet: not that Aspocomp out-competes the Asian giants on cost, but that Europe increasingly needs a domestic option and is willing to support one.

And there’s early evidence the demand is real. In its most recent quarter the company posted a record order book, with the bulk of new orders coming from the semiconductor and defense segments and part of the backlog stretching into late 2027, unusually long visibility for a company this size.

Now the warts, and they’re real. This is a genuinely tiny company, a market cap around €30 million, with a single plant. Recent results have been bumpy: even as orders hit records, sales actually dipped and margins compressed hard, dragged down by low-margin contracts signed earlier and a literal equipment breakdown that hurt output. That’s the risk in one picture, the order book says one thing, the income statement says another, and the gap between them is execution. Scaling a small single-site manufacturer into a hundred-million-euro revenue company, which is their stated long-term ambition, is hard, capital-hungry, and exactly the kind of thing small caps stumble on. None of that is hidden; it’s why the stock is where it is.

So I’d frame Aspocomp honestly: not a chokehold story like Canatu, and not a clean compounder. It’s a “right place as Europe re-shores, in a category the banks are now flagging” bet, in a name small enough that almost nobody is looking. Worth watching, not pounding the table.

Subscribe now

Modulight: the photonics wildcard

The third name is more of a wildcard, and I’ll keep it shorter because I’m earlier in my own work here. Modulight is a Tampere laser company that’s been around since 2000. The core business today is medical lasers, mostly for cancer treatment and eye conditions, a real, revenue-generating business, not a story stock. But the part that’s interesting in this context is that it runs its own vertically integrated facility in Finland, doing semiconductor fabrication, laser assembly, and optical calibration all under one roof. That’s an unusual amount of capability for a company this small, and it gives it a genuine foothold in photonics, with early exposure to defense optics, quantum, and semiconductor applications.

The first-principles appeal is the same shape as the others: a small European company with real manufacturing capability in an area where the world is increasingly worried about depending on Asia. And there are two things I like about how it’s built. First, the balance sheet is genuinely strong, mostly equity, a net cash position, and enough runway to fund itself toward profitability without being forced to raise money and dilute holders, which is rare for a company at this stage. Second, it’s quietly shifting toward a recurring “pay-per-treatment” model, where it earns a fee every time one of its hospital systems runs a procedure, rather than just selling a machine once. That’s the kind of model that compounds if it scales.

Here’s the honest framing, though, and it’s why I’d hold this one at arm’s length for now. Modulight sits a step further from the current AI and semiconductor boom than Canatu or Aspocomp. Its revenue today is medical. The photonics, defense, quantum, and semiconductor angles, the parts that would tie it directly into the trade everyone’s chasing, are early-stage optionality, not the core of the business yet. So the bull case is more diffuse: you’re backing a well-run, profitable niche company that happens to hold several call options on hot themes, rather than a pure-play sitting squarely on a named bottleneck.

That cuts both ways. It makes Modulight less obviously a rotation beneficiary than the other two. But it also means that if attention swings toward European photonics and defense optics under the same re-shoring logic, a small, profitable, vertically integrated laser maker in a NATO country is exactly the kind of name that could catch outsized interest in a hurry. Worth watching, and worth a proper deep dive, which I’ll do before I say anything stronger.

Subscribe now

Nokia: proof the rotation is real

I’ll end with the one everybody already knows, because it’s the proof of concept for everything above. Nokia spent a decade as a byword for a Finnish tech champion that lost its way, the phone company that missed the smartphone, then a telecom-equipment also-ran trading sideways for years. The kind of name nobody on FinTwit wanted to touch.

And then it re-rated, hard. Not on the old telecom story, but because the market reframed it as an AI-infrastructure play. The acquisition of Infinera plugged Nokia directly into optical networking, the systems that move data between and inside the data centers AI runs on, and suddenly its order book and guidance started reading like a different company. The stock has roughly doubled off its lows as investors repriced it from a legacy equipment vendor into something exposed to the same buildout that’s lifting everything else in the supply chain.

The point isn’t to pitch Nokia here, it’s already been found, and a doubled stock is a different risk-reward than an undiscovered one. The point is what it demonstrates. A large, sleepy, unloved Finnish tech name sat flat for years while the AI trade raged everywhere else, until the market finally connected it to a bottleneck it was actually exposed to. Then it moved fast.

That’s the whole thesis of this post in one example. The attention rotates. It went through the US hyperscalers, then the picks-and-shovels names, then Sweden and Germany, and at each step it eventually reached the companies sitting on a real bottleneck, regardless of how boring or overlooked they’d been. Nokia is the Finnish name where that has already happened. Canatu, Aspocomp, and Modulight are names where, on the same logic, it largely hasn’t yet.

I’m not promising it will. Plenty of these setups never get their moment, and I’ve tried to be honest above about where each of these has real warts, execution, dilution, timing, size. But the structural pattern is hard to ignore: the same bottleneck themes that have repriced names across the US, Sweden, and Germany are sitting, mostly undiscussed, on a handful of Finnish balance sheets. Nokia shows what it looks like when the market finally notices. The rest of the cluster is still waiting.

That’s the bet. Not that any single one of these is a sure thing, but that the gap between “same bottleneck, same tailwind” and “a fraction of the attention” is exactly the kind of inefficiency worth being early to.

Disclaimer

This post is for informational and educational purposes only. It is not financial advice, an investment recommendation, or an offer or solicitation to buy or sell any security. I am not a licensed financial advisor, and nothing here is tailored to your individual circumstances, risk tolerance, or objectives.

I hold positions in some of the companies discussed, so I am not impartial. The names covered here are small-cap and micro-cap stocks listed on Nordic exchanges, which carry elevated risk: low liquidity, high volatility, wide spreads, limited analyst coverage, and real potential for permanent loss of capital. Several are not yet profitable and may need to raise capital, diluting existing shareholders.

Some figures and claims in this post draw on company filings, third-party analyses, and external market data that I have not independently verified in every case, and which may contain errors or become outdated. Forward-looking statements and any scenario or valuation figures are inherently uncertain and should not be relied upon as predictions.

Do your own research. Verify everything independently, and consider consulting a qualified financial professional before making any investment decision. You alone are responsible for your investment choices.

DYOR. NFA.

At the start of the war I sold out of most of my biotech book, rotated into cash, and took a handful of smaller tail positions: $LYB, $WEAT, $BAK and a few others, mostly expressed through options. Those worked out far better than I had any right to expect. Most of them ran up over 100%, and with the clock ticking on the options, I closed them out and started hunting for the next thing.

I think I have found two very compelling names. Here is the origin story.

The current market is dominated by AI capex. Photonics stocks have risen wildly. Part of me wants to partake in all of these exciting technologies, because I have managed to spot AI trends before every investor poured into them in the past. Companies like Google and Nebius Group are key examples. But unlike photonics today, those stocks were extremely cheap when I bought them. Photonics is not.

So I recently started searching for AI-related bottlenecks, trying to work out where all of this capex is actually going. I asked Grok to help me uncover emerging AI capex spend. It came back with several names, most of them clustered around the same theme. The first one I came across was Accelsius, which you get exposure to via $INV. Accelsius is essentially a bet on AI data center cooling demand.

As data centers become more power-hungry on the back of next-gen chips, the nature of cooling demand is shifting. Today, if you want to cool a data center, you are most likely using air cooling, essentially mimicking how a modern building is cooled, just at far larger scale. There are several problems with this.

First, cooling a data center with air means you need fresh water to cool that air, and we all know local communities are increasingly worried about their fresh water supply. Second, and more fundamental, air cooling can only get you so far before it hits a hard physical ceiling on how much heat can be pulled off a chip. That ceiling sits around 280 watts of thermal design power, and modern AI chips have blown well past 700.

The solution is liquid cooling, and it can be done two ways: either by heating the fluid up, or by letting it boil. Most of the market is currently focused on heating over boiling, because the technology is more mature. But looking at the physics and the trajectory of power demand per chip, two-phase cooling, where the fluid is boiled to extract the heat, becomes inevitable. And the thing being boiled is not water but engineered dielectric fluids and refrigerants that vaporize around 50C, which is exactly where the chemistry comes in.

Yesterday I wrote all my thoughts on $INV in a longer piece that I will link below. In short, the company is building the two-phase cooling solution, but getting comfortable with the management and the execution is another thing entirely. Luckily, there is a far safer way to bet on this market than $INV. So that is what I bought instead.

Subscribe now

Selling shovels in the cooling gold rush

Read more

I found the name in a slightly unusual way. I asked Grok for several interesting bottleneck theses outside of photonics, ones that have not yet been hyped to the same degree. The original thesis traces back to AAIG, and there is a fairly active X community forming around the name that has helped me orient myself as I dug in. I also subscribed to MVC Investing after someone in the comments mentioned they were looking into the name as well. So this is not a lone-genius writeup; it is me building on a thread others started and trying to take it deeper. Go check out both and Investing's work.

Let us jump straight into it, starting with the technology and the sector as a whole, because that is what explains why this opportunity exists in the first place.

Subscribe now

Why This Opportunity Exists: The Cooling Bottleneck

Billions upon billions are being poured into data centers right now, and the pace is only accelerating. The reason is straightforward once you accept the premise that AI demand is real and durable. It is not just chatbots. Inference at scale, agentic AI that runs long chains of reasoning and calls tools in the background, and the emerging wave of physical AI in robotics and autonomous systems all share one trait: they are enormously compute-hungry, and that hunger compounds. Training a model is a one-time burst, but inference and agentic workloads run continuously, around the clock, across millions of users and tasks. If you believe that future is arriving, then you believe the demand for compute, and therefore for the physical infrastructure that compute runs on, is going to keep climbing for years. That belief is what is driving the capex wave, and it is the foundation everything else in this writeup rests on.

But compute does not exist in a vacuum, and this is where it gets interesting, because the build-out is starting to run into hard physical constraints. The first is power. Electricity demand from data centers is rising fast enough that it is straining grids, pushing up local power prices, and turning into a genuine political issue in the regions where these facilities cluster. The second is water, and this one is sharper than most investors appreciate. There is a growing movement against data centers at the local level, precisely because of how much electricity and water they draw from the communities around them. We have already seen public hearings and local opposition over facilities consuming the water supply that towns rely on, and a meaningful number of new data center projects are effectively stalled because of exactly these power and water concerns. So the constraint is not abstract. It is showing up as stalled permits, community pushback, and rising input costs.

This is the backdrop for what has become a kind of frenzy on X and across the investing world: the hunt for the bottleneck. If compute demand is going to explode but the build-out keeps hitting physical walls, then the companies that sell the picks and shovels, the ones that relieve those bottlenecks, become extraordinarily valuable. The most popular version of this hunt has been photonics, where the names have already been bid up aggressively. It feels like half of the platform has gone into bottleneck-finding mode, with accounts like Serenity leading the charge on the silicon and photonics side. The logic is sound: find the chokepoint, find the company that widens it, and you have found a structurally advantaged business.

This thesis is not about photonics, though. It is about cooling, and specifically about the bottleneck inside the bottleneck. Goldman Sachs has named liquid cooling one of the primary constraints in the data center market, and once you understand the physics of why, it becomes hard to unsee.

Start with how a data center actually works at the physical level. You have rows of racks, each rack filled with chips, and those chips run on electricity. Electricity in means heat out; it is a law of physics, not a design choice. Every watt of power a chip consumes becomes a watt of heat that has to be removed, or the chip throttles, fails, or melts. For most of the history of computing, this heat has been handled the way you would cool a room: with air, supported by water-based systems. You chill the air, blow it across the chips, capture the heat, and reject that heat to the outside, very often by evaporating water in cooling towers. That evaporation is the crux of the water problem. It is not that the chips drink water, it is that the traditional method of getting rid of their heat consumes enormous quantities of it by evaporating it into the atmosphere.

So here is the first reason this trade is interesting. Water is a finite resource, the political and regulatory pressure against data centers consuming it is mounting, and the traditional cooling approach is directly in the crosshairs. Something has to change in how these facilities reject heat, and that pressure is creating a forced migration toward new approaches.

The first step in that migration is liquid cooling, and the version most widely deployed today is what is called single-phase direct-to-chip. Instead of cooling the air in the room, you bring liquid directly to the chip. You run cold plates over the hot components, circulate a coolant through them in a closed loop, the coolant absorbs the heat directly at the source, carries it to a cooling unit called a CDU, dumps the heat, and cycles back. The key word is closed loop. Because the coolant recirculates rather than evaporating away, single-phase liquid cooling dramatically reduces a facility’s reliance on outside water. That solves the environmental and siting problem to a large degree. And it solves a second problem at the same time: liquid carries heat away far more effectively than air, so you can pack chips much more densely and run hotter, higher-power processors than air could ever keep up with. For the current generation of AI hardware, single-phase direct-to-chip is the workhorse, and it is genuinely good at its job.

But there is a deeper layer, and this is where the real opportunity lives. Single-phase cooling has a ceiling, and the chip roadmap is heading straight for it. The problem is in the physics of how single-phase works. It removes heat by warming up the liquid, which is called sensible heat, and a given volume of liquid can only absorb so much energy before it gets too hot to keep cooling effectively. To remove more heat, you have only one real lever: move more liquid, faster. And that lever breaks down in three compounding ways as chips get hotter.

First, the penalties of pumping faster do not scale linearly. The pressure you need rises with roughly the square of the flow speed, and the pumping power rises with roughly the cube. So doubling the flow to cool a hotter chip can cost you around four times the pressure and closer to eight times the power. Each generation of hotter chip makes the next increment of cooling disproportionately more expensive. Second, that faster flow creates real mechanical wear: erosion of the channels, stress on pumps, tubing, and connectors, and higher maintenance and failure risk. Third, and most fundamentally, there is a hard wall. Past a certain heat density at the chip surface, no amount of faster flow helps, because the limiting factor stops being how fast you move the liquid and becomes how fast heat can cross from the chip surface into the fluid at all. Beyond that point single-phase simply cannot keep a chip safe, regardless of how much water you throw at it. The consensus figure is that single-phase direct-to-chip becomes unviable somewhere around 2,500 watts per chip, and the chip roadmap is moving toward and past that line.

This is where two-phase cooling enters, and the distinction is the whole thesis. Two-phase works much like single-phase, with one transformative difference: instead of merely warming the coolant, it boils it. You circulate a dielectric fluid, a liquid engineered to be electrically safe so it can touch sensitive components, over the chip, and at the hot surface that fluid boils and turns to vapor. The vapor travels to a condenser, releases its heat, condenses back to liquid, and returns to do it again, all in a sealed closed loop. The reason this matters is a piece of physics worth internalizing: boiling a liquid absorbs vastly more energy than simply heating it. The energy a fluid soaks up when it changes phase from liquid to vapor, called latent heat, is on the order of several times greater per unit of fluid than what it absorbs by warming up alone. So two-phase can pull far more heat off a chip with far less fluid flow, which sidesteps the exact wall that single-phase runs into. It is not a marginal improvement. It is a different regime of heat removal, and it is the reason essentially every other high-heat industry, from power generation to refrigeration, relies on phase change rather than simple convection.

Put simply: as chips keep getting hotter, the industry runs out of room with single-phase, and two-phase is the architecture that picks up where it leaves off. That transition, from air, to single-phase liquid, to two-phase, is the bottleneck this writeup is about, and the company at the center of it is where we turn next.

Subscribe now

Meet the Company: Innventure and Accelsius

The asset that matters here is Accelsius. Innventure is the listed shell that lets you get exposure to it. That framing is worth holding onto from the start, because almost everything interesting about this opportunity lives inside the operating company, while the thing you actually buy on the market is the parent. We will get into what that structure means later. For now, the point is simple: when people talk about the $INV thesis, what they are really talking about is Accelsius.

So what is Accelsius? It is an Austin-based company commercializing two-phase, direct-to-chip liquid cooling for data centers, the exact architecture described in the previous section. It was founded in 2022 by Innventure, which operates as a create-and-operate platform that takes promising technology, builds a company around it, and scales it. Accelsius is the cooling business that emerged from that model, and it has become the centerpiece of the entire Innventure story.

The technology has serious provenance, and this is one of the more important and underappreciated facts about the company. Accelsius’s core two-phase cooling technology originated from research done at Nokia’s Bell Labs, one of the most credentialed industrial research institutions in the world. Crucially, Accelsius did not merely license this technology. It bought the underlying patents outright from Nokia, through a patent purchase agreement, while other software and know-how is licensed alongside it. Owning the foundational patents rather than renting them is a materially stronger intellectual property position, and the Bell Labs lineage gives the technology a pedigree that is rare for a company this young. The science had been developed at Bell Labs over several years but never productized, and Accelsius was built to take it to market.

The product line is branded NeuCool, and it is worth understanding what they actually sell, because it spans several layers of the cooling stack. At the chip level, the system uses what are effectively boiling cold plates, mounted directly onto the hot components, where the dielectric refrigerant nucleates into vapor before traveling to a coolant distribution unit to condense and cycle back. On the capability front, Accelsius has demonstrated some of the most impressive thermal numbers in the industry: in testing, its cold plate handled 4,500 watts on a GPU-socket test vehicle, reportedly the highest documented load for direct-to-chip liquid cooling, and notably the cold plate did not reach its own limit, the test equipment simply ran out of power. That figure matters because it shows headroom well beyond where single-phase gives out and well beyond current chip generations.

On the product side, the line includes the IR150, an integrated rack that combines the cooling and the IT space into a single plug-and-play enclosure rated to 150 kilowatts, designed for straightforward deployment without redesigning a facility’s plumbing, and the MR250, a row-based coolant distribution unit rated to 250 kilowatts and beyond, which Accelsius has described as the first in a planned series of increasingly powerful multi-rack systems with higher-capacity units on the roadmap. The system runs on a low-global-warming-potential, minimal-PFAS dielectric refrigerant, which matters given the environmental and regulatory backdrop. In short, Accelsius is one of a very small number of companies delivering true two-phase direct-to-chip cooling at rack scale today, with a credible product family already in the market and bigger systems in development.

Now the ownership and capital picture, because this is where the story gets its strategic weight. The most important recent event was Accelsius’s funding round, which closed with Johnson Controls leading and Legrand participating. The round valued Accelsius at roughly 665 million dollars post-money, and Innventure’s economic stake in Accelsius sits in the region of 43 percent, which is the figure to anchor on when you think about how much of Accelsius’s value flows through to the listed parent. Innventure consolidates Accelsius in its financials, so when you look at Innventure’s reported revenue, you are largely looking at Accelsius’s revenue.

The identity of those backers is the part worth dwelling on, because strategic investors are a different signal from financial ones. Johnson Controls is a roughly fifty-billion-dollar building systems and HVAC giant with deep relationships in the facilities and mechanical infrastructure of large buildings, exactly the layer where data center cooling lives. Legrand owns rack power distribution brands like Raritan and Server Technology and has strong penetration in data center rack infrastructure, particularly in Europe. So the two investors are not passive check-writers hoping for a markup. They are operators who bring distribution, manufacturing credibility, and customer access, and Legrand in particular has framed the relationship as a pillar of its data center strategy, with the two companies collaborating on integrating two-phase cooling into rack infrastructure. When giants like these put capital and their own reputations behind a small company’s technology, it is a stronger validation than a venture fund writing a check, because they are staking their customer relationships on the technology working.

Innventure itself is the publicly traded entity, and Accelsius is its flagship, though not its only, operating company. The parent also holds other ventures, including AeroFlexx in packaging and Refinity in advanced materials, but the overwhelming share of investor attention, and of the thesis, is on Accelsius and the cooling opportunity. That concentration is deliberate on the part of most people following the name, because cooling is where the large, near-term, AI-driven opportunity sits.

So to summarize the asset before we get into how to think about valuing it: Accelsius is a two-phase direct-to-chip cooling company with technology bought outright from Nokia Bell Labs, a product family already shipping across rack and row formats, industry-leading demonstrated thermal performance, a roughly 665 million dollar post-money valuation from its latest round, two major strategic backers in Johnson Controls and Legrand, and Innventure as the roughly 43 percent owner and the listed vehicle through which public investors get exposure. That is the thing at the center of all of this. The next question is how you actually put a value on something like it, and that is where most people, in my view, reach for the wrong tool.

Subscribe now

How to Value Something Like This: The Make-or-Buy Lens

The first thing I did when I started modeling Accelsius was the obvious thing: I tried to map out their revenue. Management is guiding for a 100 million dollar run rate by the end of the year, which means roughly 8 to 9 million dollars of revenue in December, exiting at that pace. Full-year 2026 revenue is likely to land somewhere around 30 to 40 million. On top of that, management has said they see a pipeline supporting something like eight times their 2026 revenue plan, which points to a pipeline north of 200 million dollars. The trouble is that it is genuinely hard to pin down how much of that pipeline actually converts to sales, and over what timeframe. So the revenue model gives you a rough shape, but not a foundation you would want to build a valuation on.

Once you get past the next year or so, management has not told us much, so you are forced to reach for industry TAM figures and sector growth rates. The number that gets quoted repeatedly is a roughly 40 billion dollar cooling TAM by 2033. But here is the thing: estimating how Accelsius fits into that TAM picture does not actually help very much, in my opinion. Saying they land at, say, 20 percent market share is just guessing. This is a startup. Picking a share number a decade out and multiplying it by a TAM is the kind of math that feels rigorous and is really just optimism with a calculator. So I do not think the conventional revenue-and-share model is where the real signal is.

Instead, there are only two things I think you can actually rely on when looking at Accelsius, and both come from outside the company’s own projections.

The first is industry validation from the players who set the standards, namely the chipmakers, NVIDIA and AMD. Two-phase is still in its early innings, and it is genuinely unclear exactly when the threshold gets crossed where single-phase no longer cuts it. It could be 2028, it could be later, it could be earlier. What matters is that we get confirmation from the reference designs that NVIDIA and AMD publish, because their roadmaps dictate what the entire ecosystem builds around. If Jensen comes out and says two-phase cooling is inevitable for the 2028 generation, we have a dramatically clearer picture than we have today. Until the chipmakers bless it, two-phase adoption is a question of when, and the chipmakers are the ones who answer that question.

The second thing to watch is what the incumbents are doing, and this is the heart of how I think you should value this asset. I will get into the competitive picture in detail later, but for now the exercise is to put yourself in an incumbent’s seat. The big cooling players are mostly doing air and single-phase today. But they know perfectly well that as chips get hotter, they will have to move to better cooling solutions. So they face a simple question: make or buy. Should they build two-phase technology in house, or acquire a company that already has it?

Building it in house takes real time and effort. These are large companies with plenty of capital, so money is not the constraint. Time is. Two-phase is a different engineering discipline, and developing it from scratch, designing, testing, and iterating through the reliability qualification that mission-critical cooling demands, takes several years. My guess is that they do not have those years to spare, and here is why. Once chips cross the threshold where single-phase can no longer cool them efficiently or economically, there will be a scramble, a window of a few years where everyone needs two-phase at once. We are already seeing the beginning of this. As I mentioned, two of Accelsius’s own investors are already strategic players who have chosen to buy in rather than build. And when you look at Accelsius’s main competitor, ZutaCore, you see the identical pattern: the incumbents circling that company also chose to buy exposure rather than develop their own. So far, essentially every major player that has looked seriously at this field has decided to bet on either Accelsius or ZutaCore, rather than join the game as a builder.

When you combine the make-or-buy logic with the future TAM of the sector, you start to get a much better picture of what an asset like Accelsius could be worth. The right question is not “what revenue will Accelsius book in 2030.” It is “how much would an incumbent be willing to pay to be in the two-phase cooling game at all.” And bear in mind the structural setup: there are several incumbents who will need two-phase, and not many companies actually making it yet. More buyers than assets is exactly the condition that drives prices up.

If you look at other sectors that have seen rapid valuation increases over the past few years, a common pattern emerges: incumbents go on acquisition sprees in the early innings of commercial sales, right at the proof inflection where the technology is de-risked but adoption has not yet ramped. Accelsius is not quite at that point yet, but we are starting to get closer to peak incumbent demand for the technology. The incumbents will look at a 40 billion dollar market and think, we need to be part of this, what are we willing to pay to participate. They obviously need upside on the purchase, but a 40 billion dollar market that is still growing can plausibly support well more than 40 billion in total value creation across its players. So it is not hard to see how an incumbent could be willing to pay several billion dollars for an asset that lets them play in that market, or even into the tens of billions if you want to be really optimistic about the scramble.

This is why the hype around Innventure is growing. Innventure owns roughly 43 percent of a potentially multi-billion-dollar company, and that ownership is being acquired through a listed vehicle that currently trades at a fraction of what that stake could be worth if the make-or-buy scenario plays out. It is genuinely easy to construct a scenario where Innventure is worth several hundred percent more than it is valued at today. The make-or-buy logic gives the story a floor that no revenue forecast could, because it does not require Accelsius to win the open market on its own merits, only to remain one of the few buyable ways into a market everyone will need to be in.

That is the bull case at its strongest. But there are some important things to keep in mind that complicate it, and those are what I want to focus on next.

Subscribe now

The Good, the Bad, and the Ugly

So far we have concluded that there is a large opportunity ahead of Accelsius, potentially valuing the company at multiple billions in the future. Innventure’s stake in that is roughly 43 percent. So it is not too big a stretch to see Innventure valued at double or triple where it trades today, somewhere in the 10 to 20 dollar range, if the scenario plays out. That is the part that gets people excited, and I think it is genuinely plausible.

But there are multiple risks and several things that make the story considerably less sexy than the headline. It is worth going through them honestly, because the difference between a good outcome and a bad one here runs straight through these issues. I will take them as the good, the bad, and the ugly.

The good, or at least the manageable. The first thing to internalize is that you only own 43 percent of Accelsius, and the rest of Innventure is other companies, namely AeroFlexx in packaging and Refinity in advanced materials. These other positions in the Innventure portfolio are interesting ventures, but they are ventures, which means they carry the same trait every venture does: the potential to require more capital and bring more dilution to the whole. When I value Innventure, I would not assign a large amount of value to the businesses beyond Accelsius. Yes, they might be worth 100 million or so collectively, but they also carry the potential to dilute a similar amount as they get funded. So I treat them as roughly a wash, neither a hidden treasure nor a disaster, just not the reason to own this. The reason to own this is Accelsius, and everything else is secondary.

The bad. The first real problem with Innventure is, obviously, dilution. We have already seen a great deal of it over the past year, and it is hard to see how, over the long term, there would not be more. This happens at two levels: at the holding company level, as Innventure raises capital to fund operations, and inside the holdings themselves, as Accelsius and the others raise their own rounds. Cumulatively, this can have a large impact on returns for anyone who wants to own this for a few years or more. A thesis can be completely correct on the technology and the eventual buyout and still deliver disappointing per-share returns if the share count keeps climbing in the meantime. This is the single most important thing for a long-term holder to keep front of mind, and it is the reason the capital structure matters as much as the technology.

The ugly. In 2021 there was a Hindenburg short report that painted a very negative picture of Innventure, focused largely on PureCycle, one of its ventures. This report came out around the time PureCycle went public, and before Innventure itself was listed. Like most short reports, it was aggressively negative, and short sellers are not neutral parties. But setting the tone aside, it identified some patterns that have, to some extent, repeated, and those patterns are worth understanding rather than dismissing.

The first pattern to be aware of is management’s tendency to make bold revenue claims and strong projections. With PureCycle, Innventure repeatedly signaled that the company was close to an inflection point and that revenues would soon be far higher. The useful thing about a five-year-old claim is that you can now fact-check it. Looking at PureCycle today, roughly five years later, it generates barely any revenue. Back in 2021, many investors listening to the inflection talk were modeling PureCycle making billions in revenue around now. In reality it is doing a couple of million, orders of magnitude below the guidance-era expectations. The lesson is straightforward: the management commentary should not be taken at face value. If Innventure’s management says they are at an inflection point, it is far better to wait for the proof than to take their word for it.

The second focus of the report was the team’s poor track record of taking portfolio companies to market. Most of their past ventures have gone bankrupt, and the report argued that they have consistently taken companies public too early. The pattern is related to the first point: the management comes across as highly opportunistic and quite promotional, spending a lot of time and effort making each venture look like the next big thing.

Now I want to address these claims directly, because the honest answer is more nuanced than either “ignore it” or “avoid the stock.”

First, on governance, something important has recently changed. An activist investor, Ascent Capital Partners, took a stake of around 7 percent and sent a pointed public letter to the board. The letter was sharply critical, essentially saying its patience had run out, and it made specific demands: materially cut corporate overhead, stop funding ventures beyond Accelsius until things stabilize, redirect capital into Accelsius, and reconstitute the board with more independent, experienced operators. Over the following weeks, the company responded substantively. It added new directors, published a formal capital allocation framework, and made meaningful cost cuts, with general and administrative expenses coming down significantly. The activist subsequently followed up with a public letter of support. This matters, and it is a genuine positive, because the activist pressure pushed the company toward exactly the disciplines that address the biggest concern here, namely overhead and dilution, and it tightened the focus onto Accelsius as the priority. Engaged shareholders forcing better capital discipline is precisely what you want to see in a holdco that has historically been criticized for the opposite. One honest caveat: an activist who has done well on the position has every incentive to declare victory, so the support letter is encouraging rather than a clean bill of health. But the direction of travel on governance is clearly better than it was.

On the Hindenburg report itself, my honest view is that it is largely correct in arguing that Innventure’s management should not be trusted completely. Where I think the report overreaches is in using failed ventures as evidence of poor performance, because on average ventures tend to fail; that is the nature of the model, and a string of failures is not by itself an indictment. You underwrite a venture platform on whether the winners are big enough to cover the losers, not on the body count. But, and this is the part I cannot wave away, it does worry me when a team consistently jumps on hype cycles and markets its story so aggressively that it repeatedly misses its own targets. I would much rather back a team that gives projections you can trust and then beats them quietly. So the management question is a real mark against the name, and I am not going to pretend otherwise.

But here is the most important thing, and it is the reason none of the above is disqualifying. Accelsius has its own management. Innventure has a parent-level management team whose record is, let us say, mixed. But Accelsius, the operating company that actually matters, is run by a separate team, and that team looks considerably stronger. So the promotional-parent concern and the quality of the operating business are two different questions, and conflating them is a mistake in either direction. The next section is where I want to dig into the Accelsius team and the company’s recent activity, because that is what gives the clearest picture of whether the operating business is the real thing the bull case needs it to be.

Share

The Accelsius Team and the Evidence on the Ground

If the parent-level management is the part of this story that gives reason for caution, the Accelsius operating team is the part that gives reason for confidence, and the contrast is sharp enough that it deserves its own section. This is a separate group of people, with a different profile, and they are the ones actually building and selling the product.

Start with the CEO, Josh Claman. He is not a promoter cut from the create-and-operate mold; he is a career data center and enterprise technology operator with around thirty years in the industry. His background includes senior leadership at Dell, where he ran large multi-billion-dollar units including serving as general manager of Dell UK and heading enterprise business for the Americas, along with time at companies like NCR and AT&T. That is precisely the kind of profile you want selling infrastructure into the same enterprise and hyperscale buyers Accelsius is targeting; he has sat on the other side of those tables for decades.

The technical bench is the part that should reassure you most, because it directly answers the question of whether the technology is real. The CTO, Dr. Richard Bonner, is a genuine two-phase thermal specialist: 18 years in thermal product development, over fifty published papers, five US patents, a former AIChE Transport and Energy Processes Division Director, and a PhD in chemical engineering from Lehigh. This matters enormously in context, because one of the sharpest criticisms in the Hindenburg report about a different Innventure venture was that the team lacked deep domain expertise in the relevant science. Here, the opposite is true: the person leading the technology is an actual boiling and two-phase heat transfer expert. The supply chain side carries serious hardware-scaling pedigree as well; the supply chain lead spent 12 years at Dell running server supply chain planning, with prior engineering experience on NASA’s Orion program at Lockheed Martin, which is exactly the discipline a company needs when its bottleneck is manufacturing scale rather than invention. The broader finance and operations leadership brings backgrounds from companies like Dell and National Instruments that are known for scaling hardware revenue from millions into billions.

So the picture of the operating team is a credible, operator-and-engineer group with real data center, two-phase, and hardware-scaling expertise. That is a meaningful rebuttal to any lazy dismissal of Accelsius as just another over-marketed Innventure venture. The parent carries the promotional history; the operating company is run by people who look like they know exactly what they are doing.

Now to the evidence on the ground, because credentials matter less than what a company is actually doing, and over the past year Accelsius has put out a steady stream of milestones and signals that, taken together, are more substantive than I expected when I started digging. Let me walk through them.

On products and performance, the company brought its NeuCool line to general availability, including the IR150 integrated rack launched at Data Center World and the MR250 row-based CDU launched at the OCP Global Summit, where management explicitly described it as the first in a planned series of increasingly powerful systems with higher-capacity units on the 2026 roadmap. On raw capability, the demonstrated 4,500 watts per socket result, where the cold plate did not reach its own limit and the test rig ran out of power first, remains one of the most impressive thermal numbers publicly shown in direct-to-chip cooling.

On third-party validation, this is where the signals get genuinely interesting, and a lot of it comes from sources most investors never bother to read. Independent engineering analysis, attributed to Jacobs, has put numbers on the efficiency case: in a modeled 10 megawatt data center, two-phase direct-to-chip showed roughly 35 percent lower annual operating expense and around 12 percent lower five-year total cost of ownership versus single-phase, with no change in capital expenditure. That matters because it is an outside firm rather than the company validating the economics.

On deployments and channel, Accelsius has a named deployment at a large data center campus through DarkNX, a 300 megawatt project that stands as the single biggest named deployment in the space. It has placed equipment at Equinix’s co-innovation facility in Ashburn, the data center capital of the world, which means the technology is being shown to essentially every major operator that passes through. And the partnership with Legrand is visibly operational rather than just a press release: at Dell Technologies World, Accelsius exhibited its IR150 inside Legrand’s booth, showing the two-phase rack alongside Legrand’s power and networking gear, which is the strategic investment actually manifesting on a trade-show floor.

Perhaps the most striking single signal came from a separate event entirely. Accelsius was featured as a technology partner at the ribbon-cutting for a major server vendor’s customer solutions center, a facility of exactly the kind where that vendor walks its own enterprise customers through validated technology. The company’s own marketing lead posted about it, and the imagery showed the cooling technology on display in the vendor’s innovation lab. I want to be careful here, because this is validation and exposure rather than a disclosed supply contract, and it should not be inflated into one. But being placed in front of that vendor’s customer base, in that vendor’s own facility, is a meaningful step for a company this size, and it is not the kind of thing that happens to a business that is all marketing and no substance.

Then there is the hiring data, which is the kind of primary research that almost nobody does and that I find genuinely revealing. You can learn a lot about where a company thinks it is going by who it is paying to hire. Accelsius has been staffing like a company preparing for a production ramp, not like a lab still searching for product-market fit. It brought on a global commodity manager, a role whose entire purpose is to lock down annual component supply contracts, which you create when you have forward order visibility, not when you are running pilots. It has a manufacturing operations function at the VP level, unusual at this revenue stage, along with production and shift roles and systems engineers with hardware and aerospace backgrounds. None of that is an organization you build on speculation. You build it against demand you can see coming.

And there is a constant drumbeat of independent commentary reinforcing the direction. Data center industry voices have been writing about “building for a two-phase future” as the prospect of one-megawatt racks comes into view, and the IR150 launch drew a wave of posts from infrastructure professionals calling it a significant step for AI-ready cooling. Even Accelsius’s own deployment partner, DarkNX, has publicly framed two-phase as core to building AI environments that can handle both current high-density chips and future Rubin-class infrastructure without major redesigns later. When a customer is talking publicly about future chip generations, that is the kind of forward signal you want to see.

None of this is a contract, and none of it is revenue at scale. It is validation, channel activity, credentialed leadership, and hiring that all point the same direction. But that is exactly the point: in a pre-inflection business, the progression is the signal, and the progression here is consistent and improving. The operating company is doing the things a real company on the cusp of an inflection does. Whether that converts into the outcome the bull case needs is the open question, and it is what the final section is about.

Subscribe now

The Competition and the Industry Picture

I have referred to the competitive landscape a few times, so let me lay it out properly, because understanding who Accelsius is actually up against, and where the whole industry sits on the adoption curve, is essential to judging the thesis.

At the leading edge of two-phase direct-to-chip, this is effectively a two-horse race between Accelsius and ZutaCore. ZutaCore is the other credible pure-play two-phase company, and like Accelsius it is backed by serious strategic investors, in its case Carrier, alongside a hyperscaler server ODM and others, which gives it a strong direct line into hyperscaler design-in. Both companies have credible technology and early deployments, and crucially neither has yet landed the named, high-volume hyperscaler contract that would settle the race. So the outcome is genuinely open, and the two are not even competing on quite the same axis, which is the most useful thing to understand about the matchup.

There are two different ways to measure cooling performance, and Accelsius and ZutaCore have each pulled ahead on a different one. The first is per socket, meaning how much heat you can pull off a single chip. Here Accelsius leads: it has demonstrated 4,500 watts on a socket-level test vehicle, the highest documented figure for direct-to-chip, and the cold plate did not reach its own limit. ZutaCore’s HyperCool has been certified to cool up to roughly 2,800 watts per processor, which is formidable but lower. This is the chip-level headroom that matters most as the next chip generations push socket power up, because it is the axis on which single-phase physically gives out. The second measure is per unit, meaning how many racks one cooling unit can serve. Here ZutaCore currently leads: in late 2025 it announced end-of-row coolant distribution units rated at 1.2 and 2 megawatts, which serve many racks from a single centralized box. Accelsius’s row-based unit, the MR250, is rated at 250 kilowatts, with higher-capacity systems promised for 2026 but not yet shipped.

So the honest current state is that Accelsius leads on chip-level thermal performance while ZutaCore leads on system-level capacity per unit, and they are partly optimizing for different buyers. Accelsius’s integrated, plug-and-play products are well suited to enterprise, colocation, and neocloud players who want to adopt incrementally without redesigning a facility. ZutaCore’s high-capacity end-of-row units are aimed more squarely at hyperscale-density greenfield deployments. Which approach wins the early volume depends on whether the near-term buying comes from hyperscalers building from scratch or from a broader set of operators retrofitting and adopting rack by rack. It is worth being clear-eyed here: on the specific metric of capacity per cooling unit, which several knowledgeable observers argue is what ultimately decides the winner at the hyperscale frontier, Accelsius is currently behind, and the higher-capacity unit it has promised for 2026 is one of the more important things to watch.

It is also worth saying that this will not be a two-company market forever. The physics is known, and as the field matures more players will be able to enter, whether established thermal names or new entrants. There are already adjacent and emerging players in two-phase and advanced direct-to-chip beyond these two. But at the leading edge today, with shipping general-availability products, named deployments, and strategic backing, Accelsius and ZutaCore are clearly the front pair, and notably almost every incumbent that has looked seriously at the space has chosen to back one of them rather than build, which is the make-or-buy dynamic in action.

Now the bigger competitive reality, which is that the real near-term competition is not the other two-phase company at all; it is single-phase. The single-phase incumbents are the ones with the installed base, the relationships, and the mature products today, and they are formidable. The market leaders here, names like CoolIT, Vertiv, and Schneider, already ship single-phase coolant distribution units at very high capacity. CoolIT’s largest row CDU is rated at 2 megawatts and has been deployed at scale, enough to cool many of the densest current-generation AI racks. So when you compare Accelsius’s 250-kilowatt row unit to a 2-megawatt single-phase unit, the gap looks enormous, but a good chunk of that is a product-class difference, an integrated rack and an early row CDU versus a mature high-capacity row product, rather than a pure capability gap. Still, the takeaway is real: single-phase is not a weak incumbent waiting to be displaced. It is a capable, entrenched technology that is being actively stretched to handle more, and it is the thing two-phase has to displace, not merely outperform on paper.

This is what makes the S-curve and the timing question the crux of the entire thesis. Two-phase is sitting at the transition from early commercial deployments into qualification: products just generally available, first named deployments landing, hyperscaler engagement at the proof-of-concept stage, and the current flagship chips still specified for single-phase. The reason single-phase is still the default is that it can be stretched further by brute force, pumping more fluid faster, even though the penalties of doing so rise steeply. So the key question is when the crossover arrives, the point where single-phase genuinely cannot keep up economically and two-phase becomes necessary rather than optional. The honest answer is that nobody knows precisely. It could be 2027, 2028, or later, and estimates keep moving. The chip roadmaps are the accelerant, and the single most important confirmation would come from the chipmakers themselves: if NVIDIA or AMD bless two-phase in their reference designs, the architectures the whole ecosystem builds around, that is the signal that the crossover is locked in. A reference-design win, or a public statement from the chipmakers that two-phase is the path for a given generation, would change the picture far more than any company press release. Watching the reference designs is, in my view, the clearest leading indicator there is.

The photonics comparison is instructive here, and it cuts in two directions. Photonics is the other great AI-bottleneck story, and its valuations have already popped hard, with names reaching multi-billion marks well before the revenue arrived. That tells you frontier-hardware valuations can run years ahead of adoption. But it also tells you that two-phase cooling has not yet had its photonics moment. The story has not gone fully hot, which means, if you believe the direction of travel, this sits earlier in its narrative cycle than photonics did, with arguably a clearer physical forcing function behind it. The flip side, and the genuine risk, is timing: if single-phase keeps getting stretched and the crossover slips out to 2029 or beyond, you are holding a cash-burning, diluting company while you wait for an inflection that keeps receding. That is the central risk of being early, and it is why the dilution discussed earlier matters so much. Being right on the technology and wrong on the timing can still cost you.

One last thread worth pulling, because it points to other ways to play this theme that may be safer than Innventure. Regardless of which cooling company wins, all two-phase systems need an engineered dielectric fluid to boil, and that fluid is a high-value, recurring consumable. The market for these fluorinated coolants is essentially a duopoly between Solstice Advanced Materials, ticker SOLS, the business spun out of Honeywell’s advanced materials arm, and Chemours, ticker CC. These are the coolant suppliers, the picks-and-shovels one layer below the cooling vendors, and they benefit whichever two-phase company wins, and indeed benefit from the broader liquid cooling shift generally. The catch is that both are large, diversified chemical companies where data center cooling fluid is only a slice of the business, so they are not pure plays, and the upside per dollar invested is more muted. But that is precisely the trade-off: they are slower, safer, more diversified bets that capture the theme without the binary technology risk and the holdco baggage that come with Innventure. For an investor who believes in the cooling bottleneck but does not want to underwrite a single early-stage company wrapped in a holding structure, the coolant names are a genuinely more conservative way to express the same view. I find them less exciting and lower-variance, which is exactly why they are worth mentioning: not every way to play a theme has to be the highest-beta one.

With the competitive and industry picture laid out, the final question is what actually moves this from here, which catalysts to watch and how I am thinking about positioning. That is where I will close.

Subscribe now

What I Am Watching and How I Think About Positioning

So where does all of this leave me. I want to close with the catalysts I am actually watching and how I think about expressing the trade, because the thesis is only useful if it connects to decisions.

Let me start with the question I have spent the most time on, which is how to express this at all, and specifically whether options make sense. The honest answer right now is that they mostly do not. The options chain for Innventure is poor. The contracts are expensive, the implied volatility is high, and the name does not have proper long-dated LEAPS. What that means in practice is that if you buy the options currently available, you are effectively betting that the stock rises by something like 100 percent within a few months, because anything less and the cost and time decay eat you alive. That is a brutal bar. You can be completely right on the direction and the eventual outcome and still lose on the option simply because the move did not happen fast enough.

To use the options, in other words, you essentially have to bet on market discovery and rerating rather than on a specific news catalyst landing in your window, and that is a genuinely hard thing to time. I will say that the signs of that kind of rerating are not absent. There has been a lot of activity around the name on X, and my own writing on it has drawn a good amount of engagement, which suggests the story can spread and the stock could rerate on hype and attention before any single hard catalyst arrives. But even granting that, the options being this expensive means they are not the ideal way to express the view. You are overpaying for the privilege of being right on a tight clock.

That mostly leaves pure stock, and when it comes to sizing I think you have to remember what this actually is: a venture. The sensible way to invest in a venture is to start small and add as the picture clarifies and the winner reveals itself, rather than going in large on conviction before the evidence is in. So my current thinking, if one wants to play the name at all, is a smaller position first, and then reassess around October, which is when I expect the next meaningful product catalyst, and decide from there whether the picture has improved enough to add. I would also note that, as covered earlier, SOLS and CC are genuinely attractive alternatives for expressing the broader cooling theme with far less single-company and holding-company risk, and for some investors those are the better vehicle entirely.

Now the catalysts themselves, in rough order of how much they would move the story.

The first is the product roadmap, and specifically the higher-capacity unit. I expect Accelsius to announce something around October, possibly earlier, that addresses the hyperscaler segment better than the current line does, following the launch cadence of the past year. If that lands, it closes the per-unit gap to ZutaCore and the single-phase incumbents and materially strengthens the competitive position. If October passes without it, that is a mild negative signal that the roadmap is slipping.

The second, and potentially the most powerful, is a chipmaker reference design. I think there are decent odds that within the next six months we see something like an AMD reference-design announcement, or comparable industry commentary, that validates two-phase for a specific high-power deployment. A reference-design win from NVIDIA or AMD is the single clearest confirmation that the crossover is locked in, because it is the chipmakers, not the cooling vendors, who dictate what the ecosystem builds around. This is the catalyst I would weight most heavily as a leading indicator.

The third is future funding rounds, which could act as a very strong rerating moment in their own right. If Accelsius announces it has completed new financing at, say, a 1.5 billion dollar valuation, the listed parent should see its implied value rise more or less instantly, because that is a fresh, market-set mark on the asset that Innventure owns a large slice of. A priced-up round is the cleanest way the private value gets crystallized into something the public market can point to.

The fourth is straightforward commercial progress: more named contracts on the scale of the 300 megawatt deployment, and crucially, whether booking growth breaks out of its current level rather than staying flat. Bookings accelerating is the proof that the pipeline is converting, and a named, high-volume hyperscaler contract would be the milestone that genuinely settles the bull case.

And the fifth, which is more of a slow-burn macro catalyst, is the environmental and water backlash. If states or regulators begin to say, in effect, that you cannot build new air-cooled or water-hungry data centers, the demand for liquid cooling would rise rapidly, and more investors would catch wind of the story. We are already seeing stalled projects and local opposition over power and water; if that hardens into actual policy, it becomes a forcing function that pulls the whole adoption curve forward, and a two-phase, water-free solution sits right in the path of that shift.

So that is the watch list: the October product cadence, a chipmaker reference design, a priced-up funding round, named contracts and booking acceleration, and the regulatory water backlash. Any one of them moving in the right direction strengthens the case; the reference design and the funding round are the two I think would rerate it hardest.

As for how I think about it overall, the thing complicating this for me is not the technology, which I find genuinely compelling, but the structure. Buying Accelsius directly as a pure play would be far easier to stomach. The added Innventure holding-company layer, with its dilution history and the promotional track record covered earlier, is what gives me pause, and it is the reason I would treat any position here as venture-style: small to start, sized to survive the dilution, and added to only as the catalysts confirm the picture. The honest conclusion is that this is a genuinely interesting, genuinely risky situation where the technology and the make-or-buy logic are the draw and the capital structure and timing are the risk. Whether those balance out in your favor depends a great deal on the catalysts above, which is exactly why I am watching them rather than reaching for the spreadsheet.

Share

Disclaimer: Nothing in this post is financial advice. I am not a financial advisor, and this is not a recommendation to buy or sell any security. It is my own research and opinion, shared for informational and entertainment purposes only. I may hold positions in the names mentioned, and those positions can change at any time without notice. Everything here could be wrong, including the facts, the assumptions, and the conclusions. Do your own research, consider your own circumstances, and consult a licensed professional before making any investment decision. Small, early-stage, and thinly traded names like the ones discussed here carry a real risk of permanent and total loss.

Haypp Group can be modelled using web traffic to a very high degree. I have consistently managed to make accurate predictions on Haypp Group’s revenue based on the web traffic of their several sites; the correlation between Similarweb panel visits and reported revenue sits at a Pearson r of roughly 0,94 on a YoY basis across the last nine quarters. In other words, when the platform’s visit footprint moves, group revenue moves with it almost one-for-one.

Previously I have mainly done quarterly web traffic updates, but I figured that doing monthly updates might be interesting to follow; it enables an earlier read into the quarter and also highlights interesting dynamics that would otherwise be missed.

And this month, the dynamics are unusual. April is showing the largest moves I have ever seen in Haypp’s web traffic data; one site has nearly doubled YoY and the Growth segment as a whole is tracking at a pace that, if it holds, would imply a step-change in quarterly revenue. At the same time, the Core segment is pulling in the opposite direction, which raises a real question about how the segments balance out at the group level.

So what is actually driving this? Is the explosive Growth reading the start of a structural shift, or noise off a soft Q1 base? And how should the Core pullback be interpreted against a Q1 that ran hot? Let’s dive into how the first month of Q2 looked.

Subscribe now

Core — Mature Nordic

In April the Core markets did not see much growth; so far we are mainly just seeing decline. At first glance this might look worrisome to someone who hasn’t looked at the stats before, but monthly swings are always large. Further, as haypp.com is both Growth and Core to some extent, this Core breakdown doesn’t factor in how haypp.com is doing; that will be covered in the Growth segment. Lastly, Q1 was a very strong quarter for the core market, and thus it makes the Q2 data look weak in comparison.

nettotobak.com — Sweden

Read more