Big tech builds some power on its own sites, and demand built there never reaches a seller like Constellation.
Is it a good business? · 6 · What does its record show?
Sales jumped with Calpine; profit per $1 swings
Sales, $B
19.6
24.4
24.9
23.6
25.5
34.2
Dec 2021Dec 2022Dec 2023Dec 2024Dec 2025last 12 mo
Sales$25.5B → $34.2Brevenue, 2025 to the past year
Profit2.0% → 18%operating profit per $1 since 2022
Debt$23.4Btotal debt after the purchase
Profit per $1 moves with power prices, hedges and a nuclear tax credit, not with volume.
Is it a good business? · 7 · How much room is there to grow?
Servers take a rising share of power
22% to 33%of commercial power to servers by 2050
About 7%in 2025
15 to 20 yearsits tech contracts run
Government forecasts measure the demand; the contracts are what turn it into income.
Is it a good business? · 8 · Who is in charge?
A chief executive who beats his own forecast
Since 2022Joe Dominguez has run it
Under 1%of the company held by insiders
4 yearsrunning he beat the midpoint of his own profit forecast
Shane Smith took over the books with the Calpine purchase.
The story right now · 9 · Why do tech firms want nuclear?
AI data centres want clean power all day and night
About 7%of commercial power to servers, 2025
Microsoftrestarting a closed reactor
Amazon, Googlenamed as buyers
Data centres run constantly, and tech firms promised clean energy. Nuclear is the one clean source that never stops, so they sign for years.
The story right now · 10 · Why could a grid rule hurt?
Charging data centres more could cool demand
Data centres need powerGrid costs rise.
PJM weighs a ruleCharge data centres.
Buyers pay moreDeals get dearer.
Old plants left alone?Not yet decided.
PJM runs the grid across the mid-Atlantic, where much of the story sits. The rule would hit the very customers Constellation sells to.
The story right now · 11 · What changed lately?
The forecast rose and Amazon and Google were named
$12.50 to $12.80raised profit per share forecast
About 4,560 MWcontracted with three customers
$24.7B → $23.4Btotal debt in three months
34.4%below the year's high
Strong quarter
It also agreed to buy a 609 megawatt gas plant in New England for $715M.
The story right now · 12 · What is everyone arguing about?
Is there room left for a setback?
Many investors think
The clean, nuclear way to own AI power
We think
Earned by scale, with little room for setback
We are wrong if deals keep coming and the Three Mile Island restart holds its 2027 date.
The price and our view · 13 · What is a share worth?
The price sits near our sunny future
$191what we think a share is worth · the last close is of that
$191our value
Attractive below $152
Neutral
Unattractive above $229
$90.37gloomy: growth halves and less kept per $1
$191our estimate: experts' path, then slowing
$253sunny: growth beats forecasts by a quarter
Value per share in three futures. None is a prediction of the share price.
We estimate what a share is worth from the cash the business should earn over many years. Our rule: buy only when the price is at least a fifth below our estimate, as a cushion for being wrong.
The price and our view · 14 · How could we be wrong?
Debt, and a rule aimed at its customers
Debt nearly tripledBiggest
For the Calpine purchase.
A rule aimed at buyersShowing now
PJM is weighing it.
A restart could slipWatch
Three Mile Island in 2027.
How PJM shares costsUnknown
Not yet decided.
More shares may come, and some buyers in recent deals are unnamed.
The price and our view · 15 · How will we know?
February's outlook tests the restart
The restart on track for 2027Good sign
A new long tech contractGood sign
PJM charging data centres moreBad sign
FEB 202728
Fourth-quarter 2026 results and outlook, 28 February. Companies report every three months; each report is a test of the story.
Our view also changes on price alone: if the shares fall to within a fifth of our estimate, we move to Neutral.
The price and our view · 16 · So, what do we think?
The best signed nuclear pipeline, priced like it
$191our value
Attractive below $152
Neutral
Unattractive above $229
UnattractiveMedium conviction
Constellation runs America's largest nuclear fleet and sells round-the-clock clean power to tech firms on long contracts. But debt nearly tripled for a purchase, and a grid rule aimed at its customers is being weighed. Unattractive.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
A data centre needs electricity every hour of every day, and big tech buyers want it free of carbon. Nuclear supplies about 19% of America's electricity, the largest carbon-free source that runs around the clock, and Constellation owns more of it than anyone.
The January purchase of Calpine, for about $22B, added about 23 gigawatts, mostly gas plants, plus California's largest geothermal complex and a retail business. A buyer's alternative is paying the day's market price for power, or building its own plant.
How it makes money
It sells the power its plants generate, into regional markets and, increasingly, under long contracts at a fixed price. Those contracts are the story: 15 to 20 years of income from investment-grade buyers. It also sells electricity and gas directly to about 2.5 million customer accounts.
Growth and profit
What growth costs
Constellation grows by buying power plants and spending on its own, not by selling harder, so the usual cost lines tell little. Its spending on plants has taken a larger share of every sales dollar each year, while profit margins have swung with power prices and trading gains. Whether growing makes it more or less profitable cannot be read from the costs yet.
Its record
The numbers
Is it growing?Fast
Sales grew 38% over the twelve months to Sep 2026, faster than the 8.3% in the year to Dec 2025.
Is it profitable?Healthy
Operating margin is 18%, up from 6.5% in the year to Dec 2023.
How steady is it?Uneven
Yearly sales growth has ranged from −5.4% to +24% over the last 5 years, with sales falling in 1 of them.
Does it earn good returns?Decent
Each dollar put into the business earns 11% a year after tax, up from 5.7% in the year to Dec 2023.
Is it making cash?Thin
It turned 3.9% of sales into free cash, $1.34B over the last twelve months.
Is the balance sheet safe?Heavy debt
It owes $22.8B more than the cash it holds, about 3.9 years of the cash its operations bring in.
Are shares being diluted?Creeping
The share count rose 7.6% in 3.8 years.
Why it wins
The evidence
The scarce thing is licensed, working nuclear capacity, and nobody can build comparable new capacity this decade. That scarcity is why Microsoft signed for 20 years to restart a closed reactor at Three Mile Island, and Meta for 20 years of a whole plant's output from June 2027.
Who could beat it
The rivals in full
The biggest buyers may skip the grid. Nearly every hyperscaler is now building power plants on its own sites, and OpenAI and Anthropic alone have leased more than 10 gigawatts of it: demand that never reaches a seller like Constellation. Rival generators chase the same deals, and management will not name the customers or prices in its current talks.
Each rival
Hyperscalers and data-center operators building their own behind-the-meter gas generation (xAI, and others tracked industry-wide) rather than buying grid-connected nuclear power through a long-term PPASPCX
The grid-connected, utility-scale PPA model Constellation's Microsoft, Meta, CyrusOne and Walmart agreements all depend on.
Other-basin and other-fuel generation growth (new gas-fired capacity in other RTOs, or other nuclear operators extending their own licenses) competing for the same data-center demand poolnot listed
Constellation's specific share of new signed data-center demand, if rivals convert comparable volume first.
How big it could get
The market and who has signed
The market
The market is data-centre electricity signed on long contracts. Government forecasts have data-centre servers rising from about 7% of American commercial buildings' electricity in 2025 to between 22% and 33% by 2050. Against that, Constellation has named about 2,135 megawatts of deals since September 2024. Nobody publishes what share of the whole any one seller wins.
Who has signed
Microsoft signed for about 835 megawatts for 20 years, and Meta for a whole plant, about 1,092 megawatts, for 20 years from June 2027. CyrusOne signed for 380 megawatts in Texas. In August another 920 megawatts arrived, of which only Walmart's 176 is named.
Who runs it
Leaders, pay and the cash
Who runs it
Joe Dominguez has run Constellation since it split from Exelon in 2022, and has beaten his own profit forecast every year since. Shane Smith took over the books when Calpine was bought. Insiders own under 1% of the company, so they gain little directly from the share price. Most of the bonus rides on yearly profit, and most of the share awards on cash generated before growth spending and on how the stock does against rivals; nothing rewards return on the money invested.
What they did with the money
It paid for Calpine with 50 million new shares and about $4.5B in cash, and debt rose to about $24.7B from about $9.0B in six months. When Calpine's former owners sold 11 million shares in June, Constellation bought back about $558M of them itself.
The last report
The latest report
Constellation raised its full-year profit forecast to $12.50 to $12.80 a share
What it did to the thesis Strong quarter
It answered all three questions we set: the forecast rose again, buyers beyond Walmart appeared and debt came down. The call then confirmed the Three Mile Island restart is on track for 2027, so the business case was met in full.
The report
The third quarter of 2026 earned $5.19 a share on the reported measure and $5.06 on management's adjusted measure, which strips out hedge gains and losses, against $3.04 a year before. Management raised its full-year adjusted range to $12.50 to $12.80, from $11.50 to $12.50.
In the quarter it contracted about 4,560 megawatts with three customers, naming Amazon and Google. Total debt fell to $23.4B from $24.7B three months earlier, and it agreed to buy a 609 megawatt gas plant in New England for $715M.
What's happening now
The background
The shares are about a third below their high of the past year, and the 9 November results and call answered the tests: the forecast rose again, new buyers were named and management said the Three Mile Island restart is on track for 2027.
What is open now is PJM, which runs the grid across the mid-Atlantic and is weighing rules that would charge new capacity costs to data centres, the very customers this story depends on.
Where we disagree
Both sides in full
The market
The clean, nuclear way to own the AI-power boom, with Calpine mostly a story about getting bigger.
We believe
The higher price is mostly earned by scale, long nuclear licences and the best signed pipeline here, but it leaves little room for a setback.
The argument
Coverage treats the Calpine debt and the grid-rule debate as footnotes. We think the grid rule is still the clearest risk: it would charge the very customers the contracts are sold to. The debt has started to fall, which eases the other.
We are wrong if deals keep being named, Three Mile Island restarts on time and the rule leaves existing plants alone.
What it's worth
How we value it
Valued on the analysts' revenue path for three years, each later year closing 30% of the gap to 4%, on official operating profit with stock pay a cost and tax paid in full; the thesis's own evidence is the check on that path.
What the price assumes
Sales growth in the first year, then slowing as ours does, at the same profit per sale.
The price assumed
15.8%
Our estimate
1%
Measured at the close of
6 Nov 2026
The three cases
Bear case$90.37
Growth runs at half the analysts' path, 0%, 3% and 4%, the margin settles a fifth lower at 14.4%, and new capital earns only the 8% rate: the downturn test.
Our value$191
Revenue grows 1%, 5% and 8% over three years, the analysts' path (C36), then fades toward 4%, closing 30% of the gap each year; the margin moves to 18.0%, the analysts' year-three figure after stock pay and on the filed basis (3.6% less).
Bull case$253
Growth runs a quarter above the analysts' path, 1%, 6% and 10%, the margin reaches 19.8%, and new capital earns 15%.
What could go wrong
Every risk
The risks
Debt nearly tripled
Debt rose from about $9.0B to $24.7B in six months, then fell to $23.4B in the third quarter, while spending on plants kept free cash flow near zero.
A rule aimed at its customers
The National Energy Dominance Council has urged the grid operator to charge new capacity costs to data centres, the buyers the whole contract story depends on.
More shares coming
About 25 million Calpine shares unlock in June 2027. The last early sale knocked 7.7% off the price.
A restart that could slip
Microsoft's contract rests on restarting Three Mile Island in 2027. Approvals are clearing, but the reactor is not running yet.
What we still do not know
Who the unnamed buyers in August's 920 megawatts are, and whether their deals become disclosed contracts with sizes and terms.
How PJM finally shares out new capacity costs, and whether any of it lands on Constellation's existing plants.
What changes our view
Every route to a new view
Every route to a new view
→ Neutral
The share price falls to about $229, back within 20% of what we think the business is worth.
→ Attractive
The share price falls to about $152, 80% of what we think the business is worth.
The next test
It supports the case if
Management restates the second-half 2027 Crane date, total debt stays at or below $23.4B at the next quarter end, and the refreshed 2027 and 2028 guidance includes the new long contracts.
It pressures the case if
The Crane restart slips beyond 2027, debt rises again, or PJM's reform advances against incumbent generators.
We would also be wrong if
The Three Mile Island restart slips well past 2027.
The verdict
Scores and how we weighed it
The scores
Growth
3 of 4
Business quality
3 of 4
Financial position
3 of 4
Execution record
4 of 4
Per share capture
3 of 4
Potential
3 of 4
Risk profile
3 of 4
Valuation
2 of 4
Evidence quality
3 of 4
How we weighed it
ValueNegative
The price sits above our value and at the bull case.
BusinessPositive
The largest nuclear fleet, with long contracts to tech companies and strong execution.
GrowthPositive
The forecast rose again and new buyers appeared.
RisksNegative
Debt nearly tripled for a purchase, a grid rule may hit its customers and more shares unlock soon.
RivalsMixed
Near the middle of the power producers it competes with.
The groupMixed
Power producers move together on AI demand news.
Latest newsPositive
The last report answered all three questions we set.
In plain words
Our estimate of value
We forecast how much cash the business can earn over the next ten years and beyond, then ask what that stream is worth today to an investor who wants an 8% yearly return. We do it three times: a gloomy future, an expected one and a sunny one. We call a share Attractive only when its price is at least 20% below the expected estimate, and Unattractive when it is 20% above.