Ignition Portfolio Update #4
Everything from the week.
Saturday, July 25, 2026 · Zero trades this week.
Last Tuesday morning, Nvidia disclosed a 9.3% stake in Nebius. It then rose 19% in one session.
I didn’t do anything.
Okay, maybe I smiled a little bit.
But then on Friday, Nebius fell 15%. Nvidia still owned its 9.3%. The contracts were all still signed. What changed was the market’s mood about AI spending — the Magnificent Seven lost something like $800 billion of value on Thursday alone, and anything within shouting distance of a GPU got dragged down after it.
What’s hilarious is that Google had possibly one of the greatest earnings reports of any recent company, yet the stock dropped like a rock. Same goes for GE Vernova. And Nokia, too.
I couldn’t be happier about the majority of my portfolio and where it sits in respect to innovation and progress. Though I am entering a season of rebalancing, where I am redeploying capital away from crowded, cyclical, disconnected-from-reality companies into less known, undervalued, future compounders.
The portfolio finished the week down 1.5%. The S&P lost 0.6%. The lead survives once again.
Before we continue:
Get access to my personal portfolio, watchlist, insights, stock picks, deep dives and more with the annual plan. Subscribe or upgrade now!
1. The Free View
15 positions, 8 themes. Same roster as last week — nothing added, nothing trimmed, nothing sold. The tickers for Kraken and Adyen, again, cannot display real-time info because they are non-US companies.
Power & Generation GE Vernova ($GEV), Constellation Energy ($CEG)
AI Infrastructure Micron ($MU), Nebius ($NBIS)
Physical AI Nokia ($NOK), ($KRKNF) Kraken Robotics
Space & Defense ($KRMN) Karman S&D, ($RKLB) Rocket Lab
Nuclear & Materials ($CCJ) Cameco, ($UUUU) Energy Fuels, ($SMR) NuScale
Software & Ad-Tech ($APP) AppLovin
Fintech ($PGY) Pagaya, ($ADYEY) Adyen
Biotech ($RXRX) Recursion Bio
2. Developments
The Nebius whiplash. Monday brought two pieces of real news: the Nvidia stake, and a $775 million senior secured debt facility, which is the first in company history, and an answer to the funding question bears have leaned on all year. Tuesday it climbed 19% and Baird initiated coverage at Outperform. By Friday, capex fear had spread across the whole neocloud group and Nebius gave almost all of it back, closing the week up barely 1%. Nothing about the company was different on Friday. Only the mood was. A position whose value swings that much on mood is exactly the kind of position you have to hold on something sturdier than mood.
GE Vernova reported a monster quarter and got punished for it. Revenue up 22% to $11.1 billion. Organic orders up 88%, to $24.2 billion. Free cash flow guidance raised from $5–5.5 billion all the way to $6.5–7.5 billion, with turbine slots selling years out. The market looked at all of that, saw an EPS miss, a still-bleeding wind division, and a warning about a $100–200 million tariff hit, and knocked the stock down almost 9% on Wednesday. Thursday it clawed most of that back while ten or so banks raised their targets. I trimmed 15% at roughly $1,070 before any of this, and the reasoning holds up: a premium multiple walking into an earnings date in a nervous market is a coin flip, and I’d rather flip it holding 85% of the position. The trim gave me room to be a little wrong.
Nokia showed up! Yes, this Nokia. AI & Cloud revenue doubled — up 105% — on €2.8 billion of AI order intake, about half of which converts to revenue within a year. Gross margin hit 45%. Management raised the full-year operating profit outlook. And the stock fell 5% anyway, on legacy telecom softness plus worries about memory costs eating into next year. I’m sure Kevin Xu wasn’t happy. I’m down around 4% from my $10.10 entry. Fine. I didn’t buy Nokia for this week’s close; I bought the nervous system of physical AI, and the order book just confirmed construction is underway.
Power kept on winning. Constellation rose 8.7% and NuScale 9.3% — the top two tiles on this week’s heatmap — with a U.S.–Saudi nuclear agreement in the headlines behind both. Ten days after my $247 entry it’s up 11%. The thesis hasn’t needed restating since I wrote it: the data centers already under construction will need electricity, and nobody on either side of the capex debate disputes that part. And more than just data centers, every aspect of the consumer economy needs more power than currently available. Homes, data centers, factories — every single aspect.
And chips stayed broken. The semiconductor index spent the whole week inside its bear market. Intel beat on Friday and dropped 8% for its trouble. Micron fell 7% on Friday — and still ended the week up 8.5%, because the first four days were that strong. A market that punishes good results is a market pricing the crowding, not the earnings, and that process takes as long as it takes. Its triggers are memory-pricing conditions, not chart levels, and none fired. Paid subscribers will see my detailed, rational thesis on the semiconductor industry and why I believe it’s the second weakest part of the buildout (behind the model layer).
One more thing. The Fed decides on rates Wednesday (!!!).
Enough said, let’s dive into the portfolio.
