$APRTEC is a Swedish thermal management company validated in space by NASA and ESA, with a global tech customer integrating its EHD technology into chip-level product development. Every watt of AI compute becomes heat. In 3D-packaged chips that heat is trapped inside the package, and rack-level cooling cannot reach it. The solution requires a pump small enough to operate at microchannel scale with no moving parts. Mechanical pumps have a physical lower bound on miniaturization. Electrohydrodynamic conduction pumps do not. APR has spent 15 years developing EHD micropumps for highly demanding thermal environments, including spacecraft qualified by ESA and NASA. Space qualification imposes reliability requirements that are among the most demanding in any industry and are generally more stringent than those found in commercial data center environments. An unnamed global tech company has now integrated APR’s EHD pump into its product development and placed a pilot production order. That moves the relationship beyond pure evaluation. The confirmatory event is a volume order from the tech customer. It would move APR from a component in a competitive cooling layer to a potential node in near-chip thermal management. Risks: •The tech customer remains unnamed and the volume decision is not yet taken. The relationship could stall. •APR’s space and defense channel runs through ZIHET, which controls production, distribution and customer relationships. The economics accruing to APR remain uncertain. •Core EHD conduction pumping is prior art from 2003. Patent protection covers specific implementations, not the principle. A well-funded competitor could develop an alternative path. Position: Long, small and a volume order from the tech customer would materially change the size of this position.

$SPCX. The question is not whether it is a great company. It is when you get paid. SpaceX has built what is arguably the strongest competitive position in aerospace, with a near-monopoly in reusable heavy launch and a dominant LEO constellation in Starlink. The open question is when a public shareholder actually gets paid, and the filing answers it more honestly than the price. At around $2.1T, $SPCX trades near 112x revenue. The only profitable segment is Starlink, roughly $4.4B in operating income. Launch appears to generate lower economic returns than Starlink and is increasingly used to support internal demand. The AI segment is consuming capital at a pace that likely exceeds $30B annually. You pay a monopoly multiple for one profitable network alongside a lower-return launch business and a capital-intensive AI segment. History sets the base rate. Seven of the ten largest US IPOs of the past decade were underwater a year later. Didi fell 84%, Rivian 64%. Facebook spent over a year below its offer price before it became a great investment. The timing sits in the lockup. There is no single 180 day wall. After the Q2 report in late July, 20% of insider shares unlock, then 7% tranches roll for months, then 28% more after Q3. Up to 2.4B shares, roughly 4.3x the float, become sellable within 90 days. That supply lands while index funds are forced to buy $SPCX into the Nasdaq 100. Musk himself is locked until June 2027. My framework suggests a buy zone around $35 to $55. There you pay 30 to 50x Starlink forward EBITDA and the AI segment is free, which is where I value it until compute leasing becomes a revenue line instead of a cost line. No position. This is a wait. What gets me long is a price in that zone combined with one fundamental confirmation: either stabilising Starlink ARPU or compute leasing proving it can become a meaningful revenue stream.

$VRT has added $75B in market value over the past 12 months. The market has decided cooling is infrastructure. What it hasn't decided yet is which part of the cooling problem is actually solved. This article maps the thermal stack and where value may migrate next. It's also why I'm long $SHT and $APRTEC.

If you only followed sentiment, you’d think June had been a disaster. Portfolio says otherwise. June: 9.12% Biggest contributors: $CRDO $MRVL $MU $PENG $STM $AMPG $ARM $SIVE $AXTI

While everybody else is waiting for $SPCX to launch, I used the moment to add some $RKLB and $ASTS. Sometimes the better move is where the crowd isn’t looking yet. As Warren Buffett put it: “Be fearful when others are greedy, and greedy when others are fearful.”

My $SPCX position is now officially established. One share. Retirement is imminent. 🚀😄

I’ve had a good run with $LPK, but last week I decided to watch from the sidelines and wait for a new trigger. That trigger came sooner than I expected. The thesis on glass replacing organic substrates is as relevant as ever. Happy to be back.

I’ve always struggled with taking profits. Why sell a winner? But step by step, I’m learning that taking some profits from winners may be smarter than selling temporary underperformers, only to watch them rip the next day. So today I practiced. Sold some $AAOI, $AXT, $CRDO and $PENG. Added $MRVL and $AMPG. Then bought back $AXT and $CRDO at lower prices. Not perfect. But better than staring at the portfolio and calling conviction a strategy.

Stepped off the $AXTI ride at $69 back in March. Found it again at $78 yesterday. Not too late to rejoin the party. Here we go again.

So ... another red day. I suppose.

Today, the European Commission published the final Code of Practice on marking and labelling of AI-generated content. I’m all for transparency. But there’s a point where disclaimers stop clarifying the message and start replacing it. If the same logic applied fully to X, every post would become: One sentence of opinion. Nine sentences explaining why the opinion may not be an opinion, advice, recommendation, endorsement, promise, forecast or reliable indication of future performance. At some point, you don’t read the message anymore. You just scroll through the legal fog. This message is not financial advice. It is not legal advice. It is not political advice. It is not medical advice. It is not a recommendation to buy, sell, hold, panic, breathe deeply, log off or reconsider your life choices. The author may own stocks, change his mind, misunderstand regulation, overestimate irony or underestimate the willingness of institutions to make things unreadable. This post was written on a MacBook Air, with human intention, machine-assisted spell check, mild caffeine exposure and no guarantee that all jokes will perform as expected. Past humor is not indicative of future humor. ec.europa.eu/commission/pres…

Seems like we Europeans may finally get a chance to buy SK Hynix during normal waking hours. With a U.S. listing reportedly coming as soon as August, I’m looking forward to owning SK Hynix and keeping a normal sleep cycle. reuters.com/world/asia-pacif…

New institutional positions in $SIVE. BlackRock and Fidelity just showed up on the shareholder list, following JP Morgan last week. Looking forward to reading the analysis explaining why U.S. institutions keep buying a “meme stock”. 😄

The world grows more uncertain. AI infrastructure holds. Iran launches missiles at U.S. bases. VIX is around 19. Markets are nervous, but far from panicked. What hasn’t changed is the AI buildout. Advanced packaging remains constrained. HBM capacity is largely committed. CoWoS remains a key bottleneck. The core investment theses remain intact. Today’s focus is U.S. CPI at 14:30 CET. Higher oil prices can feed inflation expectations, which in turn affect multiples and market sentiment. The underlying company stories, however, remain largely unchanged. On Friday, SpaceX is expected to begin trading. With a $75B raise at a $1.75T valuation, it would become the largest IPO in history. It’s been a volatile week. The volatility is likely to continue, but the theses remain intact.

This is exactly why I bought Credo. One of the few stocks green yesterday.

I believe stock market will be red the rest of the day.

One good thing about frequent red days is that I get to practice patience. So far, I haven’t bought anything… …yet.