The AI Power Crunch Is Not a Hidden Story — And Neither Are Its Real Winners

I've been very surprised by the number of articles calling Vistra, Constellation Energy, and a third unnamed utility "hidden winners" of the AI power crunch. They aren't hidden. They are the most discussed names in all of utilities. They have been covered in nearly every market column for the past 18 months. And if you bought them at the peak of that coverage, you now have a painful drawdown and a dividend yield below 0.7%.

That being the case, the article title is a false narrative. The real question is not which utility is quietly profiting from the AI boom. The real question is whether the two companies that were loudly hyped for it have actually delivered free cash flow, income, and durable shareholder returns to justify their valuations. The answer, in my opinion, is no.

Let's start with the narrative. It sounds plausible: hyperscalers are building data centers that devour electricity, utilities with nuclear or flexible generation can sell that power at premium prices under long-term contracts, and therefore those utilities are the "hidden winners." The truth is that this story has been public for over a year, priced into the stock by a factor of two or more, and the market has already voted with its feet.

Vistra (VST) trades at $141 today, down 36% from its 52-week high of $220 and down 30.6% over the trailing year. The company has real contracts: 20-year power purchase agreements with Meta for over 2,600 megawatts across three nuclear plants, and a separate 20-year deal with Amazon Web Services. Those are legitimate assets.

But the numbers tell a different story about what shareholders have actually received. Vistra's free cash flow fell 50.6% year over year to $614 million over the trailing twelve months. The company generated $4.67 billion in operating cash flow and spent $4.06 billion on capital expenditures — leaving barely any surplus. The dividend yield is 0.65%. That is not an income stock. That is a speculation dressed up as a utility. The payout ratio of 15% sounds comfortable, but it is comfortable because the dividend itself is tiny relative to the share price — which means the stock price does all the heavy lifting on total return, and the stock price is currently declining.

The balance sheet amplifies the concern. Vistra carries $35.7 billion in total debt against only $5.6 billion in equity, for a debt-to-equity ratio of 341.6%. In a rising-rate or recessionary environment, that level of leverage is a structural risk, not a manageable one.

Constellation Energy (CEG) tells a similar story, though with better absolute growth. Revenue is up 26% year over year, driven by the high-profile $16 billion Microsoft power purchase agreement to restart the Three Mile Island nuclear facility — expected online in 2028. The company also operates the largest nuclear fleet in the United States, which gives it genuine structural positioning for 24/7 carbon-free baseload power.

But Constellation's Q2 2026 earnings, reported on August 6, revealed a crack in the narrative. The company beat on EPS ($2.55 versus $2.33 expected) but missed on revenue ($7.50 billion versus $7.70 billion expected). The QoQ revenue decline was 32.53%. Management raised full-year adjusted EPS guidance to $11.50–$12.50 from $11.00–$12.00, which is supportive, but a revenue miss on the flagship AI-play story warrants attention.

More concerning for income investors: free cash flow of $309 million. On a $94 billion market cap. The dividend yield is 0.59%. Constellation has only three consecutive years of dividend growth, and the stock has fallen 37% from its 52-week high of $413 to $261 today. The PEG ratio of 4.60 suggests that even the 27x trailing P/E is expensive relative to the growth rate.

This is where the false narrative becomes most damaging. Both Vistra and Constellation have been sold to income-conscious investors as plays on a secular energy trend. But they offer less than a 0.7% dividend yield. They are not income stocks. They are growth speculation stocks with nuclear plants. The distinction matters enormously.

Of the utilities most connected to AI-driven electricity demand, I favor NextEra Energy (NEE) for its dividend commitment, free cash flow, and valuation relative to growth. NextEra trades at $84.60, with a market cap of $176.5 billion. It generates $3.03 billion in free cash flow — five times Vistra's and ten times Constellation's. Its dividend yield is 2.80%, with 23 consecutive years of dividend growth. The PEG ratio of 0.34 suggests the stock is cheap relative to its earnings growth trajectory. EV/EBITDA of 18.0x is higher than Vistra's 11.6x, but it reflects a regulated business with far more predictable cash flows and far less balance-sheet risk — debt-to-equity of 161.7% versus Vistra's 341.6%.

NextEra is not a pure-play nuclear AI power seller. It is a diversified utility with massive renewable capacity, regulated distribution operations through Florida Power & Light, and data center power contracts in development. That diversification is a strength, not a weakness. It means the company doesn't rely on a single counterparty or a single technology to deliver returns. It also means the stock is less of a narrative bet and more of a structural holding.

The strongest counterargument to this view is straightforward: Vistra and Constellation have already signed multi-decade contracts worth billions. The revenue is locked in. The risk is minimal. That being the case, there are two problems. First, locked-in revenue does not equal free cash flow if capital expenditures continue to absorb the bulk of operating cash flow — and at current capex levels, both companies are reinvesting nearly everything they earn. Second, 20-year contracts signed today will not meaningfully change free cash flow for several years. Three Mile Island doesn't restart until 2028. The Meta uprates at Vistra's plants run through 2034. Investors who need income or near-term cash flow are being asked to wait.

For investors who want exposure to the structural shift in electricity demand driven by AI infrastructure, NextEra Energy offers the combination of real free cash flow, a meaningful dividend yield, and a proven track record of dividend growth. For investors who bought Vistra or Constellation at their highs hoping for a quick trade on a narrative, I rate both as "Sell" — not because the companies are broken, but because they are over-leveraged, under-yielding, and their biggest catalysts are still years away.

The AI power crunch is real. But its winners aren't hidden. They're the companies that produce cash, pay dividends, and don't require you to wait three years for a reactor to come online.

I rate NextEra Energy as a "Buy." I rate Vistra and Constellation Energy as "Sell" for income-oriented portfolios, with the caveat that they may have a case for traders comfortable with high volatility and no dividend yield.