Vistra Is Down 30% From Its High. Here's What I'd Do Now.

Vistra shares have fallen nearly 30% from their $217.02 record high in September 2025 to about $154, as the company lost its AI-related valuation premium amid Moss Landing battery-fire disruptions and proposed PJM capacity-price caps. Despite only 3% revenue growth and a 69% EPS decline in 2025 from one-time hedge and Energy Harbor integration effects, analysts expect 2026 revenue to rise 29% and EPS to more than quadruple on data-center electricity demand. At roughly 15x next-year earnings versus 23x at its peak, the article argues the pullback offers a long-term buying opportunity, supported by Meta and Amazon data-center contracts.

Analysis

VST's valuation reset creates a more asymmetric setup only if its contracted and merchant power exposure can convert data-center load growth into realized capacity and energy pricing. The key distinction is geography: AI demand is valuable where transmission is constrained and reserve margins are tight, not simply at the national level. Moss Landing's removal also shifts the mix away from a high-value California storage optionality toward VST's Texas/PJM-facing generation portfolio, increasing sensitivity to ERCOT weather, gas spreads, and PJM market design.

Over the next 1-3 months, the principal catalyst is evidence that forward power and capacity curves remain sufficiently firm to support consensus 2026 earnings rather than additional revisions. A PJM capacity-price intervention would have disproportionate multiple consequences because investors have been underwriting scarcity rents; this is a regulatory-duration risk, not merely a one-quarter earnings issue. Conversely, a hot summer/winter, incremental hyperscaler power-contract disclosures, or improved clarity on replacement economics for Moss Landing could re-rate VST before reported earnings catch up.

Consensus may be too focused on VST as a direct AI proxy. AMZN and META benefit from securing power, but are economically on the opposite side of the equation: sustained wholesale-price inflation raises data-center operating costs and makes their power-procurement commitments more valuable strategically, while potentially pressuring AI infrastructure returns. The cleaner second-order beneficiary is nuclear exposure—CEG and NRG offer alternative scarcity/power-demand expressions—but VST's retail load book can partially offset merchant upside when procurement costs rise, making it less pure than headline framing implies.

Structural upside over 6-18 months requires new load to arrive faster than generation, transmission, and demand-response supply. The thesis fails if ERCOT/PJM forward power curves decline materially, PJM finalizes a restrictive capacity construct, or VST's 2026 EBITDA/FCF guidance does not demonstrate that data-center demand is improving realized margins rather than just volume.

Key Decisions for Investors

• Accumulate VST in tranches rather than chase a single-entry rebound; target a 6-12 month holding period and size as a regulated-policy-sensitive power allocation. Require confirmation from the next earnings release that 2026 EBITDA/FCF guidance is maintained or raised; exit/reduce on a material guidance cut or a sustained decline in ERCOT/PJM forward curves.
• Use a relative-value basket: long VST and CEG versus short XLU for 6-12 months. This isolates tightening power-market economics from rate-sensitive regulated-utility beta; principal risk is broad demand disappointment or a regulatory cap that compresses merchant capacity returns across both longs.
• For a defined-risk expression, consider VST 6-9 month call spreads only after PJM rule clarity or a positive forward-power update. Avoid outright near-dated calls: regulatory headlines and outage/remediation uncertainty can dominate the equity before fundamental earnings conversion.
• Monitor AMZN and META capex commentary for power procurement, interconnection delays, and data-center commissioning dates. Delayed energization would weaken the near-term load thesis for VST and other merchant generators even if long-run AI demand remains intact.
• Do not treat the headline multiple as sufficient evidence of value. Maintain a watch item on Moss Landing impairment/remediation costs and replacement capacity economics; a further balance-sheet or cash-flow drag would reduce the apparent discount and invalidate a simple mean-reversion trade.