Basis: 82 facts checked against 25 primary sources (SEC filings, company releases, grid and regulator documents); our valuation model, 7 Oct 2026.

Vistra owns 43,641 MW of power plants in Texas, the eastern United States and California, and sells electricity to about 5 million customers in 16 states and the District of Columbia (Vistra 10-K, 27 Feb 2026). A megawatt (MW) measures how much power a plant produces at one moment. A megawatt-hour (MWh) is one megawatt delivered for one hour, and it is the unit in which power is sold. Vistra’s plants expect to produce over 230 terawatt-hours a year, where one terawatt-hour is a million MWh.

Most of Vistra’s capacity burns gas (Exhibit 1).

Vistra is paid in three ways. First, it sells each megawatt-hour into a wholesale market at a price that changes every hour. In Texas that market is run by ERCOT, the Electric Reliability Council of Texas, which operates the state’s grid. Its eastern plants sell into PJM and three other grids; PJM operates the grid across the mid-Atlantic states and parts of the Midwest.

Second, PJM pays plants to be available when called, through a yearly auction for capacity. The auction for 2028/29 cleared at $325 per MW per day, the maximum the rules allow (PJM, 14 Jul 2026), and Vistra cleared about 10,924 MW in it (Vistra 8-K, 14 Jul 2026). Third, Vistra sells most of its output a year or more ahead at fixed prices, which is called hedging, so most of a year’s revenue is set before the year starts.

The retail business sells Vistra’s power to households and businesses under brands such as TXU Energy, which has about 2.6 million electricity customers in Texas (Vistra 10-K, 27 Feb 2026). Retail earned $1,622m of Adjusted EBITDA in 2025 (Vistra FY25 results, 26 Feb 2026). Adjusted EBITDA is the company’s measure of operating profit before interest, tax, depreciation and one-off items. Owning the customers gives Vistra a buyer for its own power at a price agreed in advance.

A data centre wants the plants that run day and night: the six nuclear units (6,448 MW) and the 28 combined-cycle gas plants (22,167 MW). Their licences matter to a 20-year buyer. Beaver Valley 1’s licence runs to 2036 and Davis-Besse’s to 2037 (Vistra 10-K, 27 Feb 2026), so a contract signed in 2026 outlasts them, and Vistra’s deal with Meta commits it to plan further licence extensions at all three of its PJM nuclear plants (Vistra, 9 Jan 2026).

AI data centres need power every hour of the day and new plants take years to build, so the fastest supply is a plant that already runs and is already connected to the grid.

A large AI data centre draws close to the same load day and night, and a power cut stops the computers. It needs firm power: supply that runs whenever it is called, which today means gas or nuclear plants rather than wind or solar alone.

New plants are slow and expensive. GE Vernova, which makes gas turbines, grew its backlog and reserved slots for gas power equipment from 100 GW to 116 GW in the second quarter of 2026 and expects at least 125 GW by the end of the year (GE Vernova, 22 Jul 2026). Combined-cycle gas plants due in 2026 and 2027 were reported at $1,116 to $1,427 per kW, and recent projects routinely report $2,000 per kW or more (GridLab, Sep 2025). Vistra agreed in January 2026 to buy ten running gas plants from Cogentrix for about $730 per kW (Vistra, 5 Jan 2026). Vistra’s own two new gas units in the Permian Basin, 860 MW in all, were under construction by February 2026 and are due online by the second quarter of 2028 (Vistra Q2 presentation, 7 Aug 2026).

The demand is large on paper. ERCOT’s preliminary forecast, filed on 15 April 2026, shows about 278,003 MW of demand by 2029 and 367,790 MW by 2032, against a record peak of 85,508 MW (ERCOT, 15 Apr 2026). ERCOT itself says the forecast is too high: it expects a summer 2026 peak of about 90,500 to 98,000 MW against the forecast’s 112,000 MW, and it has proposed cutting the forecast for data centres and crypto mining by 32% (ERCOT, May 2026). The queue of large users asking to connect holds more than 1,800 projects and more than 474 GW, many unlikely to be built (Texas Tribune, 14 Aug 2026).

Vistra has signed four long-term contracts with data-centre buyers, 4.2 GW in all (Exhibit 2). Each is a power purchase agreement: a contract to buy a plant’s output at an agreed price for a set number of years.

Vistra has not disclosed the price per MWh in the Meta or AWS contracts (Vistra Q2 results, 7 Aug 2026). Two newer items bring money but no signed contract. Helix Digital Infrastructure, a data-centre fund launched by the private equity firm KKR in June 2026 with over $10bn of committed capital, counts NVIDIA and the Kuwait Investment Authority, Kuwait’s sovereign wealth fund, among its founding investors and names Vistra its preferred power partner (KKR, 11 Jun 2026). Vistra has committed up to $1bn to it, with no MW or price agreed (Vistra 10-Q, 10 Aug 2026). On 5 October 2026 the US Department of Energy offered a conditional loan of up to $4.2bn for uprates at Vistra’s Pennsylvania and Ohio plants; the loan documents are not yet signed (US Department of Energy, 5 Oct 2026). The shares rose 10.8% the next day.

Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:

After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.

The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.

“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”

Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.

Learn More

Vistra made $5,912m of Adjusted EBITDA in 2025 and guides to $6,800m to $7,600m for 2026, and it has already sold about 100% of its 2026 output and about 94% of its 2027 output at fixed prices (Vistra Q2 results, 7 Aug 2026).

The company’s cash measure is free cash flow before growth: the cash left after running costs, interest, tax and maintenance, before any spending on new plants or acquisitions. It was $3,592m in 2025 (Vistra FY25 results, 26 Feb 2026), and the 2026 guidance is $3,925m to $4,725m. The 2027 range of $7.4bn to $7.8bn of Adjusted EBITDA uses power prices from 31 October 2025 and leaves out both the Cogentrix plants and the Meta contracts (Vistra Q2 results, 7 Aug 2026).

Because the output is sold ahead, a fall in power prices barely touches 2026 and moves 2027 only a little. 2028 is about 72% sold and Vistra publishes no price for it, so 2028 is the first year in which the power price can change the result by much.

Vistra spends the cash in four places (Vistra Q2 presentation, 7 Aug 2026):

With 2026 and 2027 mostly sold, what a buyer pays for today is the power Vistra produces from 2028 onward, and how much of it ends up on 20-year contracts instead of at market prices.

The market prices Vistra as an AI infrastructure company in the making; the contracts signed so far support less than that.

At $160.50 a buyer needs Vistra to sign about 1.3 GW more of 20-year contracts at the median $90.06/MWh by 2028, on top of the 4.2 GW already signed. In plain terms: a buyer at today’s price is assuming Vistra signs faster over the next two years than it has since January.

The dated facts behind our view:

We value Vistra at $145.11 a share, 9.6% below the $160.50 close on 6 October 2026.

We split Vistra’s cash into three kinds and discount each at the rate its own kind of investor would ask. Cash sold for 20 years to Meta and Amazon is discounted like their bonds, at 6.99%. Cash from power sold at market prices is discounted like a power producer’s shares, at 9.12%, a rate matched to how far Vistra’s shares have moved with the market over five years; retail cash sits between the two, at 7.76%. Each GW moved from market prices onto a 20-year contract at $90.06/MWh raises the value, because it earns more than the ERCOT 2029 forward of $53.42/MWh and is discounted at a lower rate.

We then value Vistra in four worlds, each driven by a different measure of AI demand. The weights start equal and are moved by five pieces of dated evidence; leaving out any one of them moves the weighted value within $133.01 to $152.17.

If ERCOT power falls to $38/MWh, the price needs 3.5 GW of new contracts (Exhibit 5). That is more than the 3.1 GW of nuclear Vistra has left, so the last 0.4 GW would have to come from gas plants.

The input that moves the value most is the contract price, and Vistra has not disclosed it. We use $90.06/MWh, the median of four reference prices: Jefferies’ estimate of about $70 for Constellation Energy’s nuclear contract with Meta and Morgan Stanley’s $98 for its contract with Microsoft (Utility Dive, 12 Jun 2025), the middle of Goldman Sachs’ $85 to $100 for new Vistra nuclear contracts, and $88 implied by Talen Energy’s contract with AWS on our own arithmetic. At $70 the price needs 2.5 GW of new contracts; at $100 it needs 1.0 GW.

Other power companies give a cross-check. Our weighted value is 8.9 times our estimate of $8.1bn of 2027 EBITDA (including Cogentrix and Meta), and the share price is 9.6 times. Constellation Energy trades at 16.4 times trailing EBITDA and NRG Energy at 14.1 times (Yahoo Finance key statistics, 7 Oct 2026). On the same trailing basis Vistra trades at 11.5 times.

At $160.50 the shares trade $15.50 above the top of our entry zone, so we rate them HOLD.

Disclosure. At the date of publication, we have no position in Vistra Corp. (NYSE: VST).

Not investment advice. This note is for information and education. It is not investment advice, a personal recommendation or an offer to buy or sell any security. Figures are from public filings and reports as dated in the text; estimates are ours and can be wrong. Do your own research.