Jan Zuazo Pfingsten

$NBIS = scarcity of compute. $RKLB = scarcity of space infra público. $BE = scarcity of power. $AMPX = scarcity of energy density. $ONDS = scarcity of autonomous systems. You don't need to hold everything in the market, if you buy this one's on market fear. You'll outperform.

Defense stocks are in a bear market. 5 things happening behind the curtain $ONDS $KTOS $AVAV $RCAT $LMT 1. Private money is at all-time highs. While public defense stocks fell 20%, Anduril just raised $5B from Thrive and a16z at a $61B valuation. Record highs. ~28x sales. Public comps like $ONDS trade at 5x with a comparable counter-drone stack. Same industry, two markets, opposite directions. One of them is wrong. 2. Backlogs grew 24% while stocks fell 20%. The five defense primes ended FY2025 with $1.36 TRILLION in combined backlog, up 23.7%. $KTOS backlog topped $2B, 72% funded. The market is selling companies whose order books just hit records. Sentiment down, demand up. 3. The Pentagon just proposed $75B for drones. FY2027 request: $75 billion for drones and counter-drone tech. Last year’s autonomy line was $13.4B. One office, DAWG, went from $225M to $54.6B. A 24,000% increase. The largest single boost of any defense program in history. It sits under SOCOM, a DZYNE customer, which $ONDS acquired two weeks ago. Announced in April. The sector fell anyway. 4. Peace doesn’t disarm. It rotates. After the Iran ceasefire, $21B in weapons sales to the Middle East were approved in Q1 alone. Kuwait bought $2B of counter-drone systems to protect civilian oil infrastructure. Wars end. The fear of drones doesn’t. Post-war money flows INTO counter-UAS, exactly where $ONDS lives. 5. We’ve seen this movie. The 90s. The post-Cold War “peace dividend” crushed defense stocks. What followed was the biggest consolidation wave in history, the one that created $LMT and Boeing as we know them. Those who bought that bear bought the giants at liquidation prices. The wave is repeating: Lockheed acquired Ultra Maritime last week. In a bear market, small caps with full tech stacks become targets. The market is selling the war that’s ending. The budget is buying the wars that come next, and they’re drone wars. I’m buying the divergence.

I see More about $BE

Data center capacity is projected to add ~106GW from 2026 through 2030 reaching 174GW and implying more than $5T of total spending. That buildout could drive 50–60% of new U.S. power capacity through 2030 spanning $GEV turbines, $BE on-site fuel cells, $CEG and $VST generation and $VRT cooling and electrical infrastructure as hyperscalers work around years-long grid queues. Once the power is secured, spending flows into $NVDA and $AMD accelerators, $AVGO and $MRVL custom silicon and networking, $TSM manufacturing and HBM from $MU, $SKHY and Samsung. I also think as AI shifts toward more distributed and latency-sensitive inference in 2027 then demand should broaden beyond mega-campuses into edge and colocation facilities while $OKLO and $SMR offer longer-term nuclear optionality.

Possible socialarb trade with $IMAX stock, everyone is looking for an Imax 70mm cinema. About 35 in the world. But then the laser IMAX is abundant. With people crazy on the odissey and later this year Dune. I think earnings can crush expectations. IMAX today 39$ a share

@ChrisCamillo possible socialarb trade with $IMAX stock, everyone is looking for an Imax 70mm cinema. About 35 in the world. But then the laser IMAX is abundant. With people crazy on the odissey and later this year Dune. I think earnings can crush expectations. IMAX today 39$ a share

$BE : Why everything under $230 is a buy - and 10 things about Bloom most people here still don’t know At today’s price ~215$, the earnings gap is fully erased. You’re buying the company as if the biggest beat in its history never happened, while backlog sits at $20B. That’s why sub-$230 isn’t a dip to me. It’s a mispricing. Hidden angle 1 — Scandium isn’t mined. It’s recovered from titanium processing waste. ~10M tons processed yearly, more than half outside China (Canada, Australia). The short’s “China dependence” thesis dies on this fact. NioCorp’s 2022 filing confirms Bloom has “contracts with numerous existing and emerging scandium suppliers.” Hidden angle 2 — The electrolyte got thinner. Recent BE patents point to a scandia-doped zirconia layer around ~50 microns, implying ~20 tons of scandium per GW - 1.5x to 3.75x LOWER intensity than the short modeled. Their “physically impossible” math used the wrong input. Hidden angle 3 — The service annuity nobody models. Every GW installed generates ~$100M/year of service revenue for 10-20 years. Service margin went from 4.8% to 18% in twelve months. By 2030 the installed base alone throws off $1.5-2B/year recurring. It builds itself with every sale. Hidden angle 4 — Bloom isn’t a bridge until nuclear arrives. It’s the complement when it does. The same solid-oxide platform runs in reverse as an electrolyzer, and Bloom has active pilots with Westinghouse and Idaho National Lab: reactors run flat-out 24/7, Bloom turns off-peak output into hydrogen, then feeds it back as dispatchable power at peak. Nuclear is baseload; Bloom is the flexibility nuclear physically can’t provide. SMRs won’t replace Bloom’s boxes — they’ll plug into them. Hidden angle 5 — ITC changed the price equation. The restored 30% credit takes Bloom’s effective cost from ~8¢ to 5-6¢/kWh — parity with gas turbines (JPM’s math). Some customers deferred 2025 installs into 2026 to capture it. Part of this year’s demand surge is that, not just AI. Hidden angle 6 — The competition quit. FuelCell Energy abandoned SOFC in June 2025 ($64.5M impairment). Sunfire exited fuel cells. Bloom is now the only commercial-scale solid oxide player in the West. A monopoly formed by desertion. Hidden angle 7 — Wall Street pays the capex now. Brookfield expanded its Bloom framework to $25B — the first time an infrastructure giant treats fuel cells as an asset class. Then this week: $1.7B from IDF and Oaktree, Morgan Stanley on tax equity, MUFG on senior debt, to deploy Bloom cells for Nebius. Bloom projects are now financeable infrastructure, like solar farms. Deployment scales without touching Bloom’s balance sheet. CEO says no dilution needed - this is why. Hidden angle 8 — Permits are the real moat. No combustion = “minor source” air permits. No water use. No smoke. While gas turbines face 2028-2029 delivery and full reviews, Bloom deploys in 90 days. When the permit IS the bottleneck, regulatory speed is revenue. Hidden angle 9 — The beat pattern. Last three quarters: +23% beat, guidance raised, +39% beat ($751M vs $540M), raised again. Management already said Q2 will be “at least as strong as Q1.” Analysts keep modeling the old curve. July 28 is the next iteration — with 10.5% short interest on top. Hidden angle 10 — The free option. Bloom’s electrolyzer was validated by Idaho National Lab as the most efficient commercial-scale electrolyzer in the world, 20-25% more hydrogen per MW than PEM/alkaline. Same platform, same factory, ~2GW/yr capacity ready. The market prices it at zero. Do the 10GW math: ~$3M per MW verified ASP, management on record that 2GW capacity supports 4x 2025 revenue. Run it forward: $30-40B annual revenue potential by decade’s end, plus the service annuity. Against a $59B market cap. The bear case is a short fund’s spreadsheet. The bull case is a $20B backlog, a patent trail, and Brookfield and Morgan wiring money. DYOR

Two weeks ago Meta announced it would sell its excess compute. $NBIS dropped 14%. The market screamed “glut.” Here’s what actually happened since: SpaceX leased its entire Memphis data center to Anthropic: $1.25B per MONTH through 2029. Then rented more capacity to Google at $920M/month. Anthropic signed a 20-year, $19B lease with TeraWulf. Now Meta is in talks to lease Anthropic up to $10B of compute. Read that again. xAI sells to Anthropic. Google sells to Anthropic. Meta wants to sell to Anthropic. Direct rivals in the model race, selling each other the weapon. When mortal rivals trade the input of the war, the input is scarcer than the rivalry. And prices? Zuckerberg himself said short-term compute deals carry “large premiums.” The feared glut was absorbed before it existed. There is no excess. There’s a queue. Eagle view: a wholesale compute market is being born, and it just exposed the hierarchy of scarcity: Power > Compute > Models. Frontier labs with 90% inference margins are wiring billions UP the stack to whoever owns the electrons. Anthropic didn’t just rent compute - it leased someone else’s power plant for 20 years. Meta bought its own 932MW gas plant. The value moved from the machines to the place you’re allowed to plug them in. I’m not betting on which model wins the war. I’m sitting where every check flows: $BE — the electrons. Only commercial-scale SOFC on earth. 90-day time-to-power. $20B backlog. $NBIS — the compute. Full-stack efficiency, $46B backlog, sold out. The labs fight the war. I own the tollbooth

Space X has bottomed here at 125$ $SPCX

NVIDIA CEO HUANG: WE ARE AT THE BEGINNING OF THE CYCLE

@elliotrades Agreed. Bloom is below the price where they posted best earnings in their history. Make of that what you want. I'm buying

This AI selloff is driving fears that MEMORY (cyclical) has topped. That could be the case. But you see a chart like this for $BE And you're trying to tell me that ENERGY TOPPED? Like they won't need to power data centers now? Fells like market is here to take out the levered degens I smell opportunity Playing cautious into midterms but then I'd expect interesting action on the AI darlings

Risk/reward seems attractive again. Lots of cheap stocks with durable competitive advantages that are going to crush numbers for the next 6-12 quarters. Time will tell!

THE FUTURE IS ON SALE RIGHT NOW Drawdown from highs: • $OKLO −76% • $NVTS −63% • $CRWV −60% • $ASTS −58% • $IONQ −57% • $PL −57% • $AAOI −56% • $RKLB −55% • $IREN −54% • $ONDS −53% • $TE −51% • $OUST −42% • $ARM −42% • $BE −39% • $NBIS −38%

$IRDM launched a new PNT chip today. This will be $RKLB after merger closes.

Some bozo once told me on X-Spaces: "Bad things happen under 200-day moving average." $RKLB closed at $76.20 today, below 200-day SMA at $77.21. Please sell all your stock. All of it. While I stack orders of naked leap put sales. Thank you! 🙏

Distracted by macroeconomic factors regarding Iran and Korean semiconductors, the market is valuing $NBIS merely as a capital-intensive GPU lessor. However, this week’s developments, specifically the Reflection partnership and the asset-light model, reveal that Nebius is deliberately building a structure to capture value across raw compute (hyperscalers), open-weight inference (labs like Reflection), and third-party infrastructure (asset-light), all atop a unified software layer. This is not just a "neocloud"; it is a platform that integrates into every layer of the AI value chain. Yet on year-end guidance, NBIS trades at ~6.7x EV/ARR. The market hasn’t repriced the shift. The whole thesis rests on one number: ARR scaling from $1.9B today to the guided $7-9B by year-end. July 29 earnings matter more than any headline or contract announcement, because price tracks the run-rate, not the news flow. Hit the trajectory → 6.7x forward is cheap, re-rate follows. Land mid ($5B) → ~9x, fairly valued, not a bargain. Miss badly → the 23x on today’s ARR gets exposed, multiple compresses. Everything else is noise until the ARR print.

The Reflection deal is a proxy for $NBIS inserting itself into the open-weight side of the AI market split; the side that drives demand for distributed compute.

lets talk market psychology for a second, using $ONDS but really this is about how you invest, period. conviction is what protects you from fear. thats the whole game. and its deeper than it sounds, because when you actually have it, something flips in your head. the stock drops, everyone panics, people are running for the door and screaming that its over. and youre just sitting there happy. because they are literally handing you the thing you want, cheaper. they think theyre selling you a falling knife. youre thinking thank you. but heres the part nobody says out loud. that feeling cannot be borrowed. you cant get it from a guru, a chart, or some guy on X with a nice follower count. conviction is built, brick by brick, by doing your own work. and if you build the tower on someone elses foundation, the first time real wind comes, youre terrified and the whole thing comes down. same tower, same wind, totally different outcome, just based on who laid the base. when your conviction comes from inside, from work you actually did, a red day doesnt read as danger. it reads as a discount. a cheaper price on something you already decided you wanted. thats it. thats the entire difference between the people who make it and the people who get shaken out at the exact wrong moment. and this is where i watch so many people on X fail. theyre glued to the screen all day, refreshing, reacting, feeling productive because theyre “working”. but its fake work. they cant see behind the curtain, so the second the wolf shows up they move like sheep. every single time. its the boy who cried wolf, except they run at every shout. if you know whats behind the curtain, you dont need to sit there all day hunting for reasons to sell. or hunting for reasons to feel safe either. you already know. the work is done. the noise is just noise. so please, silence it. theres a lot of people here with no experience who literally cant log off, theyre addicted, and they make bad decisions because they think they have to analyze every tick. it doesnt just steal your time. it makes you trade worse. staring harder does not make you see more. i dont care where ONDS goes in a week. a month. longer. genuinely dont care. i did the work, i know what i own, and short term price is just other peoples emotions on a screen. do your own work. build your own foundation. then let the wind blow.

NEW YORK TO ENACT THE FIRST STATEWIDE DATA CENTER MORATORIUM IN THE U.S., PER NYT. Gov. Kathy Hochul will pause approvals for new hyperscale data centers using 50MW+ of power for one year while the state studies energy, water and environmental impacts. The order takes effect immediately, but does not affect projects that already have required permits. Hospitals, universities and back-office financial services are not expected to be impacted.

$NBIS has secured a $1B+ AI compute deal with Reflection AI through 2029 providing access to $NVDA GB300 chips. Reflection is building open-source models to compete with OpenAI and Anthropic.

$BE