Vistra Corp. VST

Key data

The business

Vistra is the largest competitive, non-regulated power producer in the United States, about 41,000 megawatts of gas, nuclear, coal, solar and batteries. It makes money two ways: a retail arm (TXU Energy and other brands) selling electricity to roughly five million homes and businesses, and a wholesale arm selling the plants' output into open markets, heavily in Texas. The two halves fit together, retail demand hedges some of what the plants produce. In September 2024 it bought Energy Harbor, adding about 4,000 megawatts of nuclear and making Vistra the country's second-largest competitive nuclear owner. That nuclear baseload is the crown jewel: it runs all the time, which is exactly what a data center wants. The thing to hold in your head is that this is not a rate-regulated utility with a guaranteed return. It earns market prices. When Texas power spikes in a heatwave it prints money, and when gas is cheap and the weather is mild, spreads compress. The moat is scale plus that nuclear fleet plus an integrated retail book, not a regulator's blessing.

The numbers

The reported profit line is close to useless quarter by quarter, and that is the first thing to understand.

Q1 2026 came in at $2.87 against an expected $1.32, then Q2 2026 landed at $0.76 against $1.61, with the winter quarter between them at $0.55 against $2.31. Those are not operations swinging 80% in a quarter, they are unrealized marks on the hedge book moving with forward power and gas curves. The watch-item from a day earlier, that you value this on cash rather than book earnings because GAAP dances with hedge marks, held cleanly: the Q1-to-Q2 whipsaw is that exact effect, and the cash line is the honest gauge.

Reported EPS ran $7.00 in 2024 then collapsed to $2.18 in 2025 on the same hedge noise, while the business barely changed. Cash tells a steadier story: operating cash flow was $5.5B, $4.6B and $4.1B across 2023 to 2025. Against that, growth capital climbed from $1.7B to $2.8B for solar, storage and nuclear uprates, and buybacks ran near $6.0B over five years, a real dent in a $50B company that steadily shrinks the share count. The cost was leverage: long-term debt rose about $3.3B since the end of 2024 to $19.6B, roughly 3.0x net debt to EBITDA, most of it to fund Energy Harbor and the build-out. What this memo believes that the tape does not is that the 32% haircut off the high overweights the optical earnings volatility and underweights the cash the fleet throws off, and the print that settles it is full-year cash generation holding near its historical run-rate.

That backlog matters for what it is not. $7.1B of contracted revenue against $17.7B a year means most of Vistra's output is sold merchant, at whatever the market pays. The visibility is thin, and one wholesale counterparty is 36% of net exposure, $236M. This is a price-taker with a hedge overlay, not a contracted annuity.

Management

CEO James Burke runs it, and the insider ledger leans one way: about $66M sold across 23 transactions in a year against $1.2M bought. Burke himself sold $4.4M in October 2025, and two other officers sold $10.2M and $9.5M in November. Plan status is not disclosed in the filings pulled, so whether these were pre-scheduled 10b5-1 sales or discretionary cannot be told apart, which is itself the reason to note it rather than lean on it. Capital allocation is the clearer signal: roughly $6.0B of stock retired over five years, much of it bought back when the shares sat far below today, funded alongside debt-financed growth and the Energy Harbor deal. The record is aggressive return plus aggressive build, carried on a rising balance sheet.

How it fails or surprises you

AI power demand lands (right tail). If a hyperscaler signs a long-term supply or colocation deal for the round-the-clock nuclear output at a premium, the cash re-rates and the peer discount closes. The stock is 32% off its high, so almost none of this is paid for today. First tell: an announced data-center power or colocation contract.

Merchant prices roll over. With only $7.1B contracted against $17.7B of revenue, a mild Texas summer or a gas-price slide compresses spark spreads across the unhedged bulk of the fleet. The plants keep running, the margin does not. First tell: ERCOT realized prices and wholesale segment cash earnings falling sequentially.

The leverage meets real hedge losses. Net debt near 3.0x and $19.6B outstanding is fine while cash covers buybacks, dividend and growth capex. If today's hedge marks harden into cash losses, or refinancing bites, the "cash is steadier than the print" read is the thing that breaks. First tell: free cash flow before growth slipping under the capital-return bill, forcing debt higher again.

Closing thoughts

Full-year cash tells you if the fleet is still compounding at its historical run, and any announced data-center deal tells you if the nuclear premium is real. Those either happen or they don't. Merchant power prices you cannot underwrite in advance; you watch the realized spreads quarter by quarter and know the read breaks if they fall. The tails are close to balanced and both are wide. If prices roll and leverage tightens at once, this is a levered price-taker and the drawdown is real. If AI demand contracts the nuclear at a premium while the fleet keeps compounding cash into buybacks, a name at 10.7x against peers near 20x has a long way to travel. What is at risk on the downside is a merchant earnings air-pocket landing on a stretched balance sheet, what is on offer on the upside is a re-rating the market has stopped paying for.

The bet is still that America keeps needing more electricity, and that Vistra's gas and nuclear plants get paid more as Texas grows and AI data centers keep plugging in. It breaks if merchant power prices fall and the thin $7.1B contracted book cannot cushion the drop, and the pair that tells you first is realized power prices against the 3.0x leverage on that $19.6B of debt. If a full year of cash cannot cover buybacks, dividend and growth spending without borrowing more, the read that the cash is steadier than the reported number is simply wrong.