VST

Hold at $143 a share against $143.17: value and price are level. Vistra's next five years are mostly written down already: 2027 is 94% hedged, PJM capacity is sold at $325-333 a megawatt-day through May 2029, and two 20-year nuclear contracts (Amazon at Comanche Peak, Meta in Ohio and Pennsylvania) start delivering in late 2026 and 2027. Those, plus the Cogentrix gas plants, take Adjusted EBITDA from $7.3B in 2026 to about $9.6B in 2031 in our central case. The price is what that path is worth: our value is 18 cents below it, so the rule (Buy at or below fair value) gives a Hold by a hair, and any of several $2-3 judgments below would tip it either way. It does not need more data-center contracts or higher power prices, and it does not pay for them either. What it cannot absorb is ERCOT power staying near $30 and the PJM capacity price falling toward the collar floor once the collar ends: that world is worth $64.

What the business is and who pays

Vistra owns about 44 GW of generation (end-2025: 27.0 GW gas, 8.7 GW coal, 6.4 GW nuclear, 1.3 GW solar and batteries; 10-K) and sells to about 5 million retail customers, mostly through TXU Energy in Texas (139 TWh of retail sales in 2025). The retail book is the natural buyer of the plants' output: when wholesale prices rise, the plants gain and retail margins shrink, and the other way round. Adjusted EBITDA in 2025 was $5.912B: Retail $1.622B, Texas $1.834B, East $2.282B, West $0.244B (release, 2026-02-26). Since then Vistra added the Lotus gas plants (2,600 MW, closed 2025-10-22) and has Cogentrix (5,496 MW) pending.

Three sources of revenue matter and they behave differently:

• Energy. Sold forward through hedges two to three years out. As of 2026-08-03 Vistra had hedged about 100% of 2026, 94% of 2027 and 72% of 2028 (release, 2026-08-07), up from 84% and 58% in February. Hedging delays price moves; it does not remove them.
• Capacity. PJM pays generators to be available. Vistra cleared 10,314 MW at $329.17 for 2026/27, 10,566 MW at $333.44 for 2027/28 and 10,924 MW at $325.00 for 2028/29 (8-Ks 2025-07-23, 2025-12-17, 2026-07-14). Each was at a price cap PJM agreed with the states; the 2028/29 auction would have cleared at $554.72 without it and fell 6.8 GW short of the reliability target (PJM auction report, 2026-07-14). The collar runs through the 2029/30 auction (PJM filing ER26-1556). What follows is not decided.
• Long contracts. Comanche Peak: 1,200 MW to Amazon Web Services for 20 years, delivery from Q4 2027, full by 2032; management expects 8-10% more free cash flow at full use (8-K 2025-09-29). Meta: 2,176 MW from Perry and Davis-Besse from late 2026, full by end 2027, plus 433 MW of uprates from 2031 to 2034, for 8-10% and 5-7% more free cash flow (8-K 2026-01-09). No price is disclosed; the accretion implies a premium of roughly $40-50 per MWh over the market at the time (our inference).

The nuclear production tax credit (45U) pays up to $15 per MWh (2024 dollars, indexed) when a plant's gross receipts are low, phasing out between $25 and $43.75. It paid Vistra $545M in 2024 and $220M in 2025 (10-Q 2026-06-30) and runs through 2032. It is a floor under the merchant nuclear fleet in a bad price year, and it ends one year after our forecast; the model removes it from the 2031 profit the fade carries forward.

Recent results and the bar set

(Releases 2025-02-27, 2026-02-26, 2026-08-07.) Management has beaten its own first guide in each of the last two years, in 2024 mostly through a tax credit it chose not to guide. That record is real but short; it says guidance is set to be beaten, not that the 2027 range will be. The CFO said on 2026-08-07 that 2027 is "trending towards the lower end" of the $7.4-7.8B midpoint opportunity because ERCOT forwards fell, and that Cogentrix and the Meta contracts add "roughly $700 million". Our 2027 is $8.0B: $7.35B core, $0.50B Cogentrix, $0.15B Meta.

Our forecast

The model builds Adjusted EBITDA from parts that move on their own (model.py, output.txt):

Where it differs from management: our 2027 core is $0.05B under the low end of their range; we haircut the Cogentrix EBITDA they quoted ($0.552B) by 10%; we take $0.15B a year off from 2028 for the 4.6 GW of coal plants retiring by mid-2028 (10-K Note 7); and we hold the PJM capacity price at $325 in nominal dollars while costs inflate 2.5% a year, so capacity shrinks in real terms. Merchant margins move -1, -1, 0 and +0.5 dollars per MWh in real terms in 2028-2031 on the 85 TWh still exposed: ERCOT forwards for 2026-2030 fell over the year and now sit at $53-55 on-peak, far below the $138 it costs to build a peaker (ERCOT board, 2026-06-01), while PJM forwards rose (CFO). We add one more gigawatt of data-center contract from 2030 at $15 per MWh over market; management says it is "in active negotiations" and has projects in ERCOT's Batch Zero, but nothing is signed.

Capital: maintenance and nuclear fuel $1.54B in 2026 (guide) plus $0.10B for Cogentrix, rising with inflation; growth capex $0.7-0.9B a year, mostly the PJM uprates and life extensions (a program we put at $3.5B over 2026-2034; DOE offered a conditional loan of up to $4.2B on 2026-10-06, World Nuclear News, press). Cogentrix closes late 2026 for $2.3B cash, $1.5B of assumed debt and 5 million shares (10-Q Note 2). Net debt rises from $20.3B at end-2025 to $22.0B at end-2027 and is held at 2.75 times EBITDA after, under management's "below 3x" target (8-K 2026-01-05); the rest is paid out, valued at fair value with the share count held. Cash taxes stay low through 2028 on $3.8B of federal NOLs (10-K Note 6).

Value

The method values each case's 2031 business twice and takes the average:

• Long fade: returns on new capital fade from 8.0% to a mature 8%, against an 8.53% cost of capital. New investment roughly earns its cost, so growth after 2031 adds little. The fade is worth 9.0 times 2031 EBIT, $102 a share at exit.
• Outside view: 14.2 times 2031 EBIT, $209 a share. This is a judged change from the method's default of 20.4 times, the mean of Vistra's and its peers' ten-year median EV/EBIT. For a hedged generator those years are distorted twice: GAAP EBIT swings with unrealised hedge marks (CEG at 80.9 times in 2020, Vistra at 54.4 times in 2017, loss years dropped), and 2023-25 holds the data-center rerating. We price the 17% of 2031 EBITDA under signed 20-year investment-grade contracts (the premium and the underlying margin on the Meta and Comanche Peak megawatt-hours, and Meta's capacity) at the regulated utilities' median (PEG and EXC, 20.2 times) and the rest at the merchant generators' pooled median for 2016-2022 (the ten valid NRG and Vistra years, 12.9 times). This follows the business mix, not the company's own multiple from a period with different growth.

The two halves are 105% apart. The gap is the mature return: the fade assumes Vistra's new capital earns about its cost (Cogentrix on management's own numbers yields 6.6% after tax), while the market's history paid 13 times for merchant generators whose returns were lower still. We take the average as the method asks; the judgment range below shows what each choice is worth.

Judgment range (central fair value; view against $143.17, where Buy is at or below fair value and Avoid below 85% of the price, $122):

The call turns on how much of Vistra the market should value like a utility. At 14.2 times value and price are level. A reader who believes the contracted share will reach a third of EBITDA, or that the 2023-25 rerating is permanent, gets a Buy from the same operating forecast.

Operating sensitivities (central $143): merchant margin ±$5 per MWh from 2029: -$10.6 / +$10.9; PJM capacity from June 2030 at $175 (collar floor) -$12.6, at $555 +$19.3; 2027 core EBITDA $7.0B -$11.7, $7.8B +$15.1; mature return 6.5% -$7.5, 10% +$8.3; no new contract -$3.1, three gigawatts +$6.2.

Cases. Pessimistic ($64): 2027 core $7.0B; ERCOT load slips after the queue audit and batteries keep prices low, -$6 per MWh by 2029; PJM capacity at $250 for 2029/30 and $200 after; no new contracts; uprates cost 30% more; mature return 6.5%; leverage cut to 2.5 times; the Helix stake worth nothing. Optimistic ($246): 2027 core $7.6B; load growth of 4-6% a year in ERCOT (management's estimate) lifts merchant margins $6 per MWh by 2031; PJM capacity $450 from 2030; 2 GW of new contracts at $20 over market from 2029; mature return 10%. Each case takes its own exit multiple through the method's case spread (9.8 and 20.7 times EBIT). The pessimistic case stacks every adverse driver and the low multiple, so it sits below 1% of our own simulation (the optimistic case sits near the 90th percentile); with symmetric tails the case-weighted value would be higher than $149 (the advocate's $152 on the draft).

The market against our view

At $143.17 the price is our value: our central case earns 9.9% a year to the exit from here, against a 9.94% cost of equity. The reverse calculation says the price needs nothing beyond our central: 2027 core EBITDA of $7.36B, or PJM capacity at $327 after the collar. Half our simulated values sit above the price.

The Street is well above us. Its average 12-month target is $207.50 ($189 for the six set in the last month); discounted one year at our 9.94%, $189, so our value is 0.76 times it. The gap is the multiple, not the operating path. The Street's EPS ($10.19 for 2027, $15.58 for 2030 from three analysts) looks 13-33% above ours ($9.01 and $11.68), but ours holds the share count while the model pays out $4-5B a year. Spent on buybacks, the same operating path gives 2030 EPS of $14.32 to $14.95 (Codex) or $15.94 (the advocate, at $143-150 a share), within 8% of the Street through 2029. Run through our model, the Street's EPS path is worth $186-206 at our 14.2x and more at the default 20.4x (Codex, script C): the Street prices Vistra closer to the 2023-25 multiples our outside view sets aside. Its targets have fallen over the year ($218 to $189 in the last month) as ERCOT forwards fell.

Alpha, the quantitative model, ranks Vistra 688th of 696 (tier Weak; expectations E, growth E, momentum D, quality D, valuation C; read 2026-10-26 from a partial cache awaiting the daily rerating). We differ on growth: Alpha reads trailing GAAP results, which hedge marks and the 2025 Moss Landing and Martin Lake losses depressed, while our path rests on contracts signed for 2027 and after. Neither view gives a margin of safety at $143.

Review and what changed after the blind commit

Two reviewers read the draft (0a0318b) with the market and flags: an Opus advocate (red-team skill) and a Codex model auditor (gpt-5 class, codex exec, read-only). The central value moved from $140.03 to $142.98.

The correction that moved the value toward the price is the one both reviewers found independently, and it corrects the blind draft's own stated rule (price contracted EBITDA like a utility), so we take it. We did not take the advocate's two other changes, which rest on our judgment of management's numbers. Net, the call is a Hold that is level with the price; the PTC point alone is enough to put the value back under it.

Bear case and why the other side could be right

The case from below: Vistra's merchant fleet is priced off ERCOT, where around-the-clock power has settled near $30 for two years (CEO, Q2 call) and forwards keep falling as batteries and solar arrive. Governor Abbott's audit of the data-center queue may delay load that the 2027-2028 curve assumes. PJM capacity at $325 is a political price, set by a collar that ends with the 2029/30 auction; the states that asked for it want lower bills. If both happen, the pessimistic $64 is not extreme: EBITDA is still $7.9B in 2031 because of the contracts, but a buyer of a merchant generator with falling prices paid 6-13 times EBIT before 2023.

The case from above: the 2028/29 auction cleared at the cap with a 6.8 GW shortfall, and the uncollared price was $555. Data-center buyers are paying existing nuclear plants a large premium over market for 20 years, and Vistra has Comanche Peak's second unit and Beaver Valley (1.9 GW) not yet under contract. Two more such deals and a capacity reset to $450 give the optimistic $246.

Next results preview

Q3 2026 results on Friday 2026-11-06 at 10 a.m. ET (company release, 2026-09-29). The CFO said the 2026 and 2027 guidance would be updated on this call, with 2027 waiting for Cogentrix if it has not closed.

Each $0.1B of 2027 core EBITDA above ours is worth about $3.3 a share. The print that would move the value most is the 2027 guide: $7.7B or less including Cogentrix and Meta puts us near the pessimistic core and takes about $10 a share off; $8.3B or more adds a similar amount. A new data-center contract is worth about $3 a GW at our premium, more if priced like Meta's.

Catalysts and falsifiers

• 2026-11-06: Q3 results and the 2026-2027 guidance update.
• Late 2026: Cogentrix closing (FERC approved August 2026; either side may walk after 2026-12-31, extendable twice by 90 days; 10-K).
• December 2026: PJM's 2029/30 auction, the last under the collar.
• 2027: PJM's design for 2030/31 and after; ERCOT Batch Zero decisions after the audit; first Meta and Comanche Peak deliveries.
• Falsifiers: 2027 guide below $7.7B including Cogentrix and Meta; ERCOT 2028 forwards falling another $5 per MWh; a PJM rule for 2030/31 that caps capacity below $250; Comanche Peak energisation slipping past 2028.

Sources

NEI, Nuclear Costs in Context (2025 cost data, August 2026), for the review's nuclear cost. Primary: 10-K 2025 (filed 2026-02-27); 10-Qs Q1 and Q2 2026; 8-Ks 2025-05-21 to 2026-10-05 (Lotus, Comanche Peak, PJM auctions, Cogentrix, Meta, debt); earnings releases Q4 2024 to Q2 2026; company release 2026-09-29 (Q3 date); PJM 2028/29 auction report and ER26-1556 filing; ERCOT board item 14.2 (2026-06-01). Secondhand: Q2 2026 call transcript (Quartr via market data). Press: World Nuclear News 2026-10-06 (DOE loan); RTO Insider and Utility Dive on the 2028/29 auction and the collar. Market data statements and fiscal-year-end multiples (fetched 2026-10-26).