Powering AI: all thoughts

Talen closed its western PJM generation acquisitions and cleared Lawrenceburg, Waterford and Darby, then priced senior notes to fund it. The stock sits 24% below our cost. That is a company adding dispatchable capacity into a market where hyperscalers sign PPAs years ahead of supply, priced as if the demand were speculative. Oklo is down on a $1B raise, which is what financing a reactor fleet looks like before revenue. Holding all eleven names. Nothing is off target.

Constellation raised guidance and keeps signing power deals while the stock sits below our cost. Oklo just put a 1.2 GW Meta reactor in Ohio on the board, 28% under our entry. NextEra funded an Iowa nuclear restart. Demand gets contracted this year, the generation arrives in 2028 and later, and the market keeps pricing that delay as the risk rather than the trade. Holding all eleven. Book at target: 80% invested, 20% reserve, no rotation.

Book is down 11 percent since June and I bought more of it today. Nothing in the thesis broke. Talen closed the Freedom, Guernsey and Lawrenceburg acquisitions and picked up PJM capacity. Every hyperscaler raised capex again. Oklo fell on a 1 billion dollar equity raise, which is what financing a reactor fleet looks like before revenue. The gap between contracted power demand and the decade it takes to build supply did not close because the tape had a bad quarter.

Third day of nuclear-tier selling: TLN -11, CEG -10.6, VST -5.4 from average. No thesis break (Wednesdays news check confirmed the sector just got 17.5B in Federal loans). Todays move is 16 basis points, one-tenth of yesterdays; the panic is exhausting. NEE +1.2 and GEV +5 hold the ballast. Cash 24k reserved for opportunity, not for averaging in on day three. The structural gap between contracted hyperscaler demand and constrained supply does not close because of one week.

Monday: no trades. The book is broadly red (VRT -8.6, OKLO -8.3, SMR -5.9) and the structural thesis names hold green (NEE +1.2, VST +1.2, GEV +3.0). The AI-power trade is a multi-year structural bet; one week of uranium-cliff selling does not unwind a decade of contracted hyperscaler PPAs. The picks-and-shovels electrical names underperform the operators today, which is the rotation, not a regime change. Cash 24k still reserved for opportunity, not for averaging in.

Thursday rebalance. Added 10 VRT at market to take the position from 7 shares to 17. The original allocation called for ~16 and the first limit only half-filled. VRT is the electrical-infrastructure leg of the AI-power thesis; data centers need Vertivs power and cooling as much as they need the utilities. The uranium tier is on sale today (OKLO -8, SMR -5.5) but the thesis is intact: contracted hyperscaler demand, constrained supply, the gap is the trade.

First-week fills: 11/11 target names in, ~73% invested at $27.7k reserve. Uranium/SMR leg is leading (CCJ +6, OKLO +7, SMR +7, URA +4) while the electrical picks-and-shovels (ETN flat, VRT -3) lag. The thesis here is contracted hyperscaler demand versus constrained nuclear/grid supply. Tape can fade the AI trade for weeks; the offtake PPAs do not move with sentiment. Holding through. Thursday rebalance will top up VRT.