Opinion · June 2026
Power Before Chips: What the Nuclear PPA Wave Says About TLN and VST
Most of the AI-infrastructure conversation on this site is about chips — SNDK, AMD, MU, INTC, NBIS, CRWV. Two smaller positions, Talen Energy (TLN) and Vistra (VST), sit on a different leg of the same buildout: the power those data centers actually run on. Both are up meaningfully since I bought them, and both have been genuinely volatile this month even though the underlying contracts behind them run for decades. That combination is worth examining honestly rather than just citing as a thesis win.
The real deals behind the thesis
Talen's Susquehanna nuclear plant supplies Amazon Web Services under an 18 billion dollar, 1.9 gigawatt power-purchase agreement running through 2042 — 1,920 megawatts of carbon-free nuclear power locked in for close to two decades. Vistra signed its own nuclear PPA with Meta in January 2026: 2.6 gigawatts across PJM-grid plants, starting late 2026, with an option for Meta to draw power from a new 300-megawatt small modular reactor. Both deals share a structure: hyperscalers paying utilities for firm, round-the-clock, carbon-free power over multi-decade terms, because a data center that needs to run continuously can't tolerate the intermittency of a spot power market or the emissions profile of gas peaker plants.
The honest complication: contracted cash flows don't mean a calm stock
The instinct is to treat "20-year contracted revenue" as a lower-volatility, bond-like characteristic. The actual price action doesn't support that. TLN traded from a $336 low on June 10 to a $438 high by June 22 — a roughly 30% range in twelve trading days — and VST moved from $138 to $167 over a similar stretch. These are not calm stocks, despite sitting on some of the most durable contracted cash flows in the book. What the PPAs actually buy is confidence in the underlying business model over a decade-plus horizon, not day-to-day price stability. Conflating "durable cash flow" with "low volatility" is a mistake worth naming explicitly, because it's the kind of thing that sounds right in a thesis writeup and turns out wrong the first time the stock has a rough week.
Why I still think this is a different bet than the chip cluster
The volatility is real, but the source of it is different. Semiconductor names move on sentiment about AI capex broadly — a Samsung earnings report, a Broadcom guidance miss, a memory-pricing rumor can move the whole group regardless of company-specific news. TLN and VST's swings this month have tracked more company- and deal-specific catalysts: PPA announcements, regulatory news, capacity updates. That's a real distinction even if it doesn't show up as lower volatility in the price chart. Combined, TLN and VST are 4.15% of the book — a modest sleeve inside the Emerging Growth bucket, sized well under any cap, which is the right way to hold a thesis I believe in structurally while being honest that "structurally sound" and "low-risk" aren't the same claim.