Vistra at $156: Merchant Power and the AI Demand Bet

Vistra doesn’t sell you a widget. It sells electrons into a market where the price resets every five minutes, and its profit in any given year depends on where power prices land, what its hedges were struck at, and how tight the reserve margin runs across Texas and the mid-Atlantic. That is what a merchant power producer is: not a regulated utility earning a guaranteed return on a rate base, but a generator taking the market price, with all the upside and all the whiplash that implies.

The whiplash is right there in the numbers. Vistra’s return on invested capital ran -5.7% → 11.5% → 16.4% → 7.1% from fiscal 2022 to 2025. Net income went from a $1.2 billion loss to a $2.7 billion profit and back down to $944 million. Free cash flow swung from -$0.8B to +$3.8B to +$1.3B. Try running a clean discounted-cash-flow model on that series and you’ll get nonsense, which is exactly why the naive reverse-DCF on this stock is a trap I’ll walk through below.

And yet, at about $155.73 (as of July 8, 2026), Vistra trades at roughly 14.4x forward earnings while 18 analysts carry a mean target of $222.89, about 43% above the price, with a strong-buy consensus. The bull story in one line: the AI datacenter buildout is a structural demand shock for electricity, it tightens power markets, and Vistra’s nuclear and dispatchable gas fleet is positioned to sell into it. The bear story in one line: merchant power is cyclical and this balance sheet carries real leverage, so a peak can look a lot like a plateau right before it isn’t. Let me value it the way a merchant generator should be valued, on earnings and EBITDA and the durability of the power-price uplift, not on a free-cash-flow number that bounces around like a heart monitor. Every figure below is computed from Vistra’s own statements.

#First, understand what you’re valuing

A regulated utility is a bond with a logo. It earns an allowed return on its assets, the regulator sets the rate, and the cash flows are boring and predictable. Vistra is the opposite animal. It’s a merchant generator: it builds and buys power plants, then sells the output into wholesale markets like ERCOT (Texas) and PJM (the mid-Atlantic), where prices float. It hedges some of that exposure forward, which smooths results but also means a given year’s profit reflects hedges struck a year or two earlier, not just today’s spot price.

Bolt on two more pieces. Through the Energy Harbor acquisition, Vistra added a fleet of nuclear plants, carbon-free baseload that runs around the clock and is suddenly very valuable to hyperscalers who want clean, firm power for datacenters. And through its TXU and Dynegy heritage it runs a large retail electricity book, selling directly to homes and businesses, which acts as a partial natural hedge to the generation side.

So when you value Vistra, you are not valuing a stable annuity. You are valuing a leveraged bet on power prices, capacity payments, and the reserve margin, with a nuclear option (literally) on the AI demand theme. That single fact reshapes every number that follows.

#The ROIC line tells the merchant story, not a moat story

For a wide-moat compounder like Visa or Coca-Cola, I want to see ROIC sitting far above the cost of capital, year after year, as the fingerprint of a durable advantage. Vistra fails that test, and it’s important to be honest about why: it doesn’t have that kind of moat. Merchant power is closer to a commodity business. The “advantage,” such as it is, comes from scale, fleet dispatchability, an integrated retail hedge, and irreplaceable nuclear and gas assets in the right markets, not from pricing power over a captive customer.

Read the line below correctly. The -5.7% in FY22 reflects a loss year (largely mark-to-market hedge losses as power prices spiked). The climb to 16.4% in FY24 was a genuinely strong power-price year. The drop to 7.1% in FY25 pulls ROIC back below the ~8.9% cost of capital. That dip is not evidence the business is deteriorating. It’s the combination of the merchant cycle, hedge accounting, and a denominator that grew as Energy Harbor’s nuclear assets landed on the balance sheet. The right way to judge a merchant generator is through the cycle: average ROIC across these four years is positive but modest, and it clears the hurdle only in the good years.

i

×

A saw-tooth, not a moat

This is return on invested capital versus the ~8.9% cost of capital (the dashed line). For a durable-moat business you want the line to sit calmly above the dashes every year. Vistra's line does the opposite: it dives below zero in FY22, vaults to 16.4% in FY24, then falls back to 7.1% in FY25. That is the signature of a merchant power producer, where profit is a function of volatile wholesale prices and mark-to-market hedges, not of a protected franchise.

Two things drive the swings specifically. First, hedge accounting: when power prices move, Vistra books mark-to-market gains and losses that can dominate reported returns in any single year. Second, the invested-capital denominator grew as the Energy Harbor nuclear fleet was added, which mechanically pulls ROIC down until those assets earn into the higher-demand environment they were bought for.

The investing takeaway is to resist two mistakes: don't read the FY24 peak as the new normal, and don't read the FY25 dip as decay. Value this business on a mid-cycle view of returns, and remember that the entire bull case rests on the AI demand surge lifting the whole cycle (pushing the average of this line up) rather than on any single year.

#Margins and cash both swing, and that’s the point

Here’s where merchant power gets genuinely tricky to value. Operating margin and free cash flow both swing hard, and they don’t swing in lockstep. FY23 shows the tell: an 18% operating margin but a $3.8 billion free cash flow gusher, the highest of the four years, because of how hedge collateral and working capital settled that year. FY25 is the mirror: a 10.7% margin and only $1.3 billion of free cash flow.

This is why the free-cash-flow number is nearly useless as a valuation anchor for Vistra. A generator posts and recovers billions in collateral as forward power and gas prices move; those flows hit the cash-flow statement but tell you little about the earning power of the plants. One year’s FCF can be inflated by unwinding hedges, the next depressed by posting margin. If you build a DCF off a single year’s FCF, or worse, off a growth rate extrapolated from these numbers, you are modeling noise.

i

×

The gap between the two lines is the whole warning

The left axis is operating margin; the right axis is free cash flow in dollars. On a normal business these would rise and fall together: more margin, more cash. On Vistra they diverge. FY24 has the fattest margin (23.7%) but less free cash flow than FY23, which had a far thinner margin (18%). That disconnect is the fingerprint of a merchant generator whose cash flows are dominated by hedge collateral and working-capital swings, not by the P&L.

Mechanically, when forward power and natural-gas prices move, Vistra must post or recover collateral against its hedge book, sometimes billions of dollars in a year. That cash movement is a financing artifact of running a hedged generation fleet, not a signal about how profitable the plants are. It reverses over time, but in any single year it can swamp the operating picture.

For valuation this means one thing: anchor on earnings and EBITDA across the cycle, and treat reported free cash flow as directional at best. It is the reason the EV/FCF multiple of ~54x and the reverse-DCF that falls out of it are both meaningless here: they divide a real enterprise value by a distorted, cherry-picked cash number.

#Revenue is steady, earnings are not

Put revenue and net income side by side and the merchant nature jumps out. The top line is almost boring: $13.73B → $14.78B → $17.22B → $17.74B, a tidy compounding of volumes and prices plus the acquisitions. The bottom line is anything but: a $1.23 billion loss in FY22, a $1.49 billion recovery in FY23, a $2.66 billion peak in FY24, then back down to $944 million in FY25.

Most of that bottom-line violence is unrealized hedge gains and losses flowing through the income statement. When Vistra hedges future output and prices spike, it can book large paper losses even as the underlying business is fine (that’s largely the FY22 story). When prices cooperate and hedges settle favorably, profit balloons. This is why the trailing P/E of 26.1x is close to meaningless right now: it’s dividing the price by a depressed 2025 earnings number. The forward P/E of 14.4x tells you the market expects earnings to roughly normalize higher next year, consistent with the consensus next-year EPS near $11.28 in the data.

i

×

The bottom line is a mark-to-market story

The tall bars (revenue) climb steadily; the short bars (net income) lurch from negative to a big peak and back. On a normal industrial company these would be roughly proportional. Here they aren't, because a large share of Vistra's reported profit in any year is unrealized gains or losses on its hedging positions, which swing with power and gas prices independent of how much electricity it actually sold.

The FY22 loss was driven substantially by mark-to-market hedge losses during a period of spiking prices, an accounting outcome, not a sign the plants stopped earning. The FY24 peak and FY25 pullback reflect the reverse and the normalization. This is why "earnings" for a merchant IPP have to be read across several years, and why analysts lean on adjusted EBITDA, which strips out much of the mark-to-market noise, to gauge the underlying trend.

For the reader, the practical lesson is that the trailing P/E of 26x is anchored to a depressed number and the forward P/E of 14.4x reflects an expected earnings recovery. The valuation question is not "is 26x too high"; it is whether the recovery the forward multiple prices in is durable, which is a question about power prices and AI demand, not about accounting.

#The balance sheet is where the risk concentrates

Merchant generation is capital-intensive and Vistra funds it with a lot of debt. Total debt sits near $20 billion against $658 million of cash, for net debt of roughly $19.3 billion, or 2.84x EBITDA. A regulated utility can carry high leverage because its cash flows are guaranteed by a regulator. A merchant generator carrying this much debt against volatile cash flows is a different risk profile entirely. Interest coverage of about 1.9x in FY25 is thin, meaning EBIT covered interest less than twice over in a soft year.

I want to be neutral and precise here, not alarmist. The leverage is a deliberate, financeable structure for an asset-heavy generator, and in a strong power-price environment it magnifies equity returns beautifully, which is a big part of why the stock has run. But it cuts both ways. If power prices soften or a bad hedge year hits, the same leverage compresses the equity fast, and thin coverage leaves less margin for error. This is the number I’d watch most closely, along with the maturity schedule and any collateral demands in a volatile price environment.

i

×

Utility-scale debt on merchant-scale volatility

The debt bar dwarfs the cash bar: roughly $19.9B of total debt against $658M of cash, for about $19.3B net debt and a net-debt-to-EBITDA ratio of 2.84x. For reference, I flag anything above 2.5x as a caution, and Vistra has been above that line in two of the last four years per the red-flag screen.

The reason this matters more for Vistra than for a regulated utility is the volatility of the cash flows servicing the debt. A regulator guarantees a utility's revenue; the wholesale power market guarantees Vistra nothing. Interest coverage of ~1.9x in FY25 shows how a soft year squeezes the cushion. The debt also isn't the whole story on cash calls: a hedged generator can face large collateral posting when prices move, which is a separate liquidity consideration.

None of this makes the balance sheet broken. It makes it a constraint and an amplifier. In the bull scenario, leverage turbocharges equity returns as power prices and capacity payments rise; in the bear scenario, it does the reverse and faster. Any thesis on Vistra has to price the leverage as a core part of the risk, not a footnote.

#Capital allocation: buybacks funded partly by debt

Over the last five years Vistra deployed cash roughly like this, as a share of cumulative free cash flow: about 115% to capex, 81% to buybacks, 28% to dividends, and 62% to acquisitions (chiefly Energy Harbor). Those percentages sum well above 100% because the spending was funded partly by debt and asset sales, not just by internally generated cash, which is consistent with the leverage picture above.

The buybacks stand out, and they deserve a precise reading rather than praise or scorn. Repurchases absorbed about 81% of five years of free cash flow, and because total spending across capex, buybacks, acquisitions and dividends ran to nearly 290% of FCF, a meaningful share of that buyback was funded not by internally generated cash but by debt and asset sales while net debt climbed from ~$13.3B (FY22) to ~$20.1B (FY25). That is a fact to price, stated neutrally: returning capital via repurchase while adding leverage magnifies per-share outcomes in both directions. Net stock issuance was -5.8% of sales in the latest year, so the share-count reduction is genuine, not dilution masked by stock comp (stock-based comp is a modest 0.6% of sales). Whether that’s value-accretive depends entirely on the price paid: buying back stock is smart below intrinsic value and a heavier bet on a levered balance sheet if done at a cyclical peak multiple. Given how much the stock has moved, that’s a judgment reasonable people can disagree on, and it’s worth watching whether repurchases continue at these higher prices or slow.

The red-flag screen returns three marks over four years: ROIC below 10% in two years, net debt/EBITDA above 2.5x in two years, and one year of negative free cash flow. For a merchant IPP, none of these are damning on their own, they’re characteristic of the model, but together they underline that this is a cyclical, leveraged business, not a steady compounder.

A word on the dividend, because Vistra pays one and it’s easy to overlook. The forward rate is about $0.92 a share, a yield of just 0.58% on a 15.2% payout ratio. That is a token dividend by design. Vistra’s five-year average yield was 1.82%, so today’s yield sits well below its own history, mostly because the stock has tripled, not because the payout shrank. The dividend per share has actually grown at roughly an 11% annual clip since 2019 ($0.50 to about $0.90). The signal here is simple: this is a buyback-first capital-return story, not an income stock. Management returns cash overwhelmingly through repurchases and keeps the dividend small and steadily growing. If you’re buying VST for yield, you’re in the wrong name; the return case is capital appreciation plus buyback-driven per-share growth, both of which live or die on the power-price cycle.

i

×

A growing dividend, a collapsing yield, and why that's fine

The rising line is dividend per share; it has climbed steadily from $0.50 in 2019 to about $0.90 in 2025, a mid-teens-to-double-digit growth rate. The falling line is the dividend yield, which dropped from over 3% in 2022 to under 0.6% today. Both lines are true at once, and the reason is the share price: the payout grew, but the stock grew far faster, so the yield compressed.

This is the shape of a company that treats the dividend as a modest, reliable, growing token rather than the main event. The payout ratio is only about 15% of earnings, leaving the vast majority of returnable cash for buybacks and reinvestment. That is a deliberate capital-allocation choice for a business whose earnings are too volatile to support a large, safe dividend across the cycle.

The takeaway for anyone screening on yield: VST is not an income stock, and the 0.58% yield tells you so. The total-return case rests on capital appreciation and share-count reduction, both of which are levered to power prices, not on the dividend check. I have deliberately excluded the 2016 figure (a one-time distribution around Vistra's emergence from bankruptcy) and the partial 2026 stub, which would otherwise distort the trend.

i

×

Spending above 100% of FCF, funded by the balance sheet

These slices are each expressed as a percentage of the five-year free-cash-flow total, and they add to well over 100%. That is not an error: it means Vistra spent more than it generated in free cash flow, funding the gap with debt and asset sales. The single biggest use was capex (keeping and growing the fleet), followed by an aggressive buyback program and the Energy Harbor nuclear acquisition.

The buyback is the most interesting choice. Reducing the share count by buying stock is powerful when the shares are cheap and questionable when they are expensive, especially on a levered balance sheet. Vistra genuinely shrank its share count (net issuance was negative), so the buybacks are real, but their wisdom depends on the price paid across the cycle.

For the thesis, capital allocation reinforces the core point: this is a company comfortable using leverage to buy assets and shares. In a rising power-price world that magnifies returns to equity. It also means the balance sheet has less slack if the cycle turns, which is why leverage keeps coming back as the central risk.

#The valuation trap, and how to value it instead

Now the part that matters. The local model spits out a few numbers that look precise and are actively misleading for this business, so let me disarm them before building something honest.

The reverse-DCF says the price implies ~13.6% FCF growth for a decade. Ignore it. That figure divides the enterprise value by 2025’s $1.3 billion free cash flow, a number distorted by hedge collateral. The implied “growth rate” is an artifact of a bad denominator, not a real hurdle. The single-scenario FCF-DCF spits out ~$104 fair value. Ignore that too, for the same reason: it grows a noisy $1.3B FCF base at 8.4% and discounts it, which is modeling noise with false precision. And the EV/FCF multiple of ~54x is not “expensive”, it’s meaningless, because the F in the denominator is a cherry-picked trough.

For a merchant power producer you value the earnings power and EBITDA across the cycle, and you cross-check against the demand catalyst. Here’s a cleaner frame built on the forward earnings and EBITDA the market actually trades on:

Method	Assumption	Implied value/share
Forward earnings, low multiple	~$11.28 EPS × 12x	~$135
Forward earnings, mid multiple	~$11.28 EPS × 15x	~$169
Forward earnings, high multiple	~$11.28 EPS × 18x	~$203
EV/EBITDA on FY25 actual EBITDA	10.6x × ~$5.23B, less net debt	~$107
EV/EBITDA on forward EBITDA	10.6x × ~$6.8B (+30%), less net debt	~$178
Analyst mean (18 analysts, strong buy)		$222.89
Current price		$155.73

Look hard at the two EV/EBITDA rows, because this is where a lazy multiple analysis smuggles in the bull case. Hold the multiple constant at today’s ~10.6x and the answer depends entirely on which EBITDA you multiply. On FY25’s actual $5.23 billion of EBITDA, the same multiple that the market pays right now spits out only about $107 a share, below the current price. To get up to ~$178, you have to grow EBITDA to roughly $6.8 billion, about 30% above FY25, which is precisely the AI-driven power-price ramp the whole thesis is arguing about. So an “EV/EBITDA at a conservative current multiple” line that lands at $178 isn’t conservative at all: it’s the bull case wearing a conservative multiple as a disguise. Say the EBITDA base out loud or the method lies to you.

Two more honest observations fall out. First, the current price of $155.73 sits right in the middle of the forward-earnings frame, roughly 14x expected earnings, which is neither obviously cheap nor obviously expensive for a levered cyclical. Second, the Street’s $222.89 mean sits above almost every line in the table. Here’s the arithmetic on that gap. My mid frame uses consensus next-year EPS of about $11.28 at 15x, giving ~$169. To reach the $222.89 mean on that same 15x multiple, you need EPS of roughly $14.87, about 32% above the base consensus. So the Street is either underwriting 2027-2028 earnings well above the near-term consensus, or applying a materially higher multiple on the belief that the AI-power demand shock re-rates merchant generation durably. Either way, the mean target embeds an earnings or re-rating hurdle roughly a third above the current consensus, and that hurdle, not a vibe, is the thing you’d have to believe.

Intrinsic value · the verdict

Blended fair value

$175

Current price

$156

Margin of safety

12%

i

×

What you're looking at & why it matters

These are valuation multiples — the price paid per dollar of earnings, cash flow, or EBITDA, shown trailing (LTM) versus forward (NTM). A multiple means nothing alone; it only makes sense against the company's own history, its peers, and the growth and returns that justify it.

Each dot is a fair-value-per-share estimate from a different method; the dashed vertical line is today's market price. Dots to the right imply upside on that method; dots to the left mean the price is already ahead of it.

The methods disagree on purpose — a multiple built on a depressed-cash-flow year lands low, a growth DCF lands high. The honest answer is the range between them, not any single point.

Why it matters: seeing every method at once, against the price, keeps the valuation honest — it exposes how much of the outcome rides on the growth assumption instead of hiding behind one false-precise target.

Fair values from multiple methods (see above). Source: Yahoo Finance + computed on forward earnings/EBITDA (FCF-DCF rejected) · as of July 2026. Illustrative, assumption-driven, not a price target.

My read: on a mid-cycle earnings frame, a fair-value range of roughly $150–200 is defensible, with the price near the low end of that band and the analyst mean above the high end. Be clear-eyed about the tails, though. If you multiply today’s EBITDA by today’s multiple with no growth, you get ~$107, so the downside if the AI ramp disappoints is a real de-rating, not a soft landing. And the ~$203 top of my earnings frame only appears at an 18x multiple on already-recovered earnings. That’s not a call to do anything. It’s a statement that the stock is priced for a solid but not heroic outcome, while the sell-side is priced for the AI-power theme to keep lifting the cycle. The gap between those two is the trade the market is arguing about. Note two margins of safety: against my own $150-200 range, the price offers a small cushion at the low end; against the $222.89 consensus, the discount is a wide ~30%, but that consensus itself embeds the ~32%-above-consensus earnings (or a fatter multiple) you’d have to underwrite yourself.

#The bet: is the AI power uplift durable?

The bull case is genuinely strong, and it’s structural, not hype. AI datacenters are enormous, always-on electricity consumers, and their buildout is landing on grids, especially in PJM and ERCOT, that were not planning for a step-change in demand. When demand rises against a slow-to-build supply, wholesale power prices and capacity payments (what generators get paid just to be available) rise. Vistra’s fleet is exactly what a tightening grid needs: dispatchable gas that can ramp on demand and carbon-free nuclear baseload that hyperscalers increasingly want to contract directly. Capacity auction results in PJM have already stepped up sharply, which flows straight to a merchant generator’s economics. If a chunk of Vistra’s nuclear or gas output gets locked into long-term datacenter power purchase agreements at premium prices, that converts some of the cyclical upside into contracted, more durable cash flow, which is precisely what would justify the Street’s higher multiple.

The bear case is the mirror. Merchant power has always been cyclical, and every prior boom in this sector eventually met new supply, demand disappointment, or a milder-weather year that broke the price. The AI demand estimates could prove too aggressive, or the buildout could slow, or new gas and renewable capacity could arrive to cap prices. And whatever happens to prices happens to Vistra’s equity with 2.84x leverage attached. A merchant generator at a cyclical peak, trading at what looks like a cheap forward multiple, is a classic value trap: the “cheap” multiple is cheap because the earnings are near a peak the market doesn’t expect to last.

There’s also a subtler risk hiding inside the bull case itself: counterparty concentration. The clean-cash-flow prize everyone wants is a long-term nuclear PPA with a hyperscaler, but Vistra’s nuclear fleet is only about 6,400 MW of its roughly 41,000 MW of capacity, and there are only a handful of buyers large enough to contract it (essentially Microsoft, Amazon, Google, Meta). A deal that locks in premium-priced, contracted cash flow also ties a chunk of the “durable” part of the thesis to one or two credit counterparties and to the pace of a single customer’s datacenter buildout. Concentration cuts both ways: it’s what makes the contracted upside valuable, and it’s what makes that upside fragile if a hyperscaler slows, renegotiates, or a deal stalls in interconnection queues and regulatory review (the kind of hurdle a co-location structure at a nuclear site has already run into across the sector). The bull case wants the nuclear PPAs; it should also size how much of the thesis rides on a small set of buyers.

What would change my view? On the bull side, signed long-term datacenter PPAs (especially for the nuclear fleet) at premium prices, and continued firm capacity-auction results, would convert cyclical upside into durable cash flow and validate the higher valuation. On the bear side, softening forward power curves, a stalling datacenter buildout, or rising net debt/EBITDA would tell me the peak is rolling over while the leverage is still on. The next earnings report (August 7, 2026) and the trajectory of capacity prices are where I’d look for the answer.

Because this is where the honest limits of this piece matter: I was not able to verify Vistra’s latest earnings-call transcript word for word while writing, so I’ve kept the forward section to well-established structural drivers and the analyst consensus in the data, and I’ve quoted no executive and cited no specific quarterly guidance figure I couldn’t confirm. The demand mechanism is real and documented across the sector; the exact magnitude of Vistra’s contracted upside is the thing you’d want to confirm directly from the company’s filings and calls before underwriting the bull case.

The tension is clean. You’re being offered a solid, well-run generation fleet at about 14x expected earnings, with a credible structural tailwind and a leveraged balance sheet that will make the ride loud in either direction. The market and the Street are arguing about how much of the AI-power premium is permanent. Where you land on that question, not on any single valuation number, is the whole thesis.

What’s your read on the merchant power names, is the AI-datacenter demand a durable re-rating or a setup for a classic cyclical peak? I’d like to hear the bear case. Come argue it with me on X or LinkedIn.

Keep reading: for the same quality-versus-price lens applied to a very different, capital-light business, see my breakdown of Visa’s payment-network moat at $352, and the rest of the Finance category works through more companies the same way.

Data sourced from Yahoo Finance via my local analysis tooling, as of July 8, 2026, computed from Vistra’s reported financial statements; free cash flow is computed from the cash-flow statement (operating cash flow minus capex) and, for a merchant power producer, is deliberately not used as a valuation anchor because it is distorted by hedge-collateral and working-capital swings. Valuation inputs (discount rate ~8.9% via CAPM) are derived from the financials and analyst estimates, not hand-picked; because reported earnings and free cash flow are volatile, I value the business on forward earnings and EV/EBITDA rather than a free-cash-flow DCF. I was unable to verify Vistra’s latest earnings-call transcript word-for-word at the time of writing, so the forward section is limited to well-established structural drivers and the analyst consensus in the data rather than unconfirmed quarterly figures or executive quotes. Figures are illustrative and may lag or contain gaps; always cross-check against Vistra’s 10-K/10-Q filings on SEC EDGAR before acting.