Energy / Power generation · Deep dive

Constellation Energy

America's largest producer of carbon-free power — 21 reactors, a nation-leading nuclear fleet, and, after swallowing Calpine for $26.6B, the biggest merchant generator in the country, now selling electrons straight to the hyperscalers building AI.

well positioned

It owns the scarcest asset in the AI-power scramble — a fleet of always-on, carbon-free reactors with a legislated revenue floor — and is signing the hyperscalers who need them, even if the price paid for Calpine and the stock leave little room for error.

My take ★★ Very interesting ★ Interesting − Maybe ✗ Not interesting

Snapshot

Constellation Energy is the largest producer of carbon-free electricity in the United States and, since January 2026, the country’s largest power producer of any kind. It runs the biggest US nuclear fleet — 21 reactors across roughly a dozen sites — and after acquiring Calpine for a net ~$26.6 billion, it commands about 55 gigawatts of combined nuclear, gas, geothermal, wind, solar and hydro capacity. It sits at the center of the AI-power story: its always-on reactors are exactly the firm, clean baseload that hyperscalers are scrambling to lock up, and in 2024 it signed a 20-year deal to restart Three Mile Island Unit 1 for Microsoft. It posted $23.6 billion of revenue in 2024 and $11.12 billion in Q1 2026 alone (up 64% year over year on Calpine), carries a market cap near $98 billion (July 2026), and guides to $11–12 of adjusted EPS for 2026. It is a genuine incumbent riding a genuine tailwind — with a valuation and a debt load that leave little margin for a stumble.

Founding story

Constellation as it exists today is four years old and two centuries old at once. The corporate entity was created on February 1, 2022, when Exelon completed the tax-free spin-off of its competitive generation business — Exelon Generation Company was renamed Constellation Energy Corporation and listed on Nasdaq as CEG, while Exelon kept the regulated utilities (ComEd, PECO, BGE and others). The “Constellation” name itself is inherited: it belonged to the old Constellation Energy Group, the Baltimore holding company whose roots trace to the Baltimore Gas Light Company chartered in 1816, and which Exelon absorbed in 2012. The spin was pitched as unlocking value — letting a merchant fleet of reactors and renewables trade on its own thesis rather than being buried inside a regulated utility.

The person who carried the thesis is Joe Dominguez, a lawyer and regulatory strategist who ran policy at Exelon and then led ComEd before taking the top job at the new company. For years Dominguez made an unfashionable argument: that zero-carbon nuclear baseload was systematically underpriced by markets that treated it like any other megawatt. When AI data-center demand detonated in 2023–2024, that argument stopped being contrarian and became the market’s consensus — and CEG’s stock re-rated from a ~$13 billion opening market cap to roughly eight times that within four years.

Ownership history since the spin is straightforward: it has been a widely held public company throughout, distinguished by two large capital events — the Microsoft-backed Three Mile Island restart and the Calpine acquisition — rather than by private-equity control.

How it works

The core machine is the nuclear fleet, and its economics rest on a single unglamorous number: the capacity factor, the share of the year each reactor actually runs. Constellation’s fleet posted a 94.7% capacity factor in 2025 and hit 98.8% in the summer peak — meaning its reactors are producing at nameplate almost all the time, including when demand and prices spike. A reactor is enormous fixed cost and near-zero marginal cost, so uptime is everything; two decades of industry-leading capacity factors is the operational moat.

Those electrons reach money three ways. First, wholesale markets: Constellation sells into regional grids run by ISOs like PJM (the 13-state Mid-Atlantic/Midwest market) and ERCOT (Texas, added via Calpine), earning both energy revenue and — critically — capacity payments for promising to be available. PJM capacity prices, driven largely by data-center load, exploded from $28.92/MW-day (2024/25) to $269.92 (2025/26) to the $329.17 price cap for 2026/27; roughly 40% of the December 2025 auction was attributable to data-center demand (PJM’s Independent Market Monitor). Second, direct long-term PPAs with large buyers — the emerging “powered-land” model where a hyperscaler contracts a plant’s output for 10–20 years. Third, the retail arm: Constellation is the largest competitive US energy supplier, serving ~2.5 million customers including about 80% of the Fortune 100, which lets it buy and sell power around its own generation and lock in load.

Underneath sits a legislated floor. The federal nuclear production tax credit (PTC) tops up revenue up to $15.00/MWh when a unit’s realized price falls between $26.00 and $44.75/MWh — effectively guaranteeing ~$44.75/MWh. Analysts treat this as structural, not cyclical: it caps the downside on the entire nuclear book.

Product and business overview

The business breaks into four named pieces. Nuclear generation — 21 reactors, the largest US fleet, the zero-carbon baseload that commands the premium. The Calpine gas and geothermal fleet — acquired January 2026, adding the country’s largest independent gas fleet plus The Geysers geothermal complex in California, extending CEG into ERCOT and the West. Retail and commercial supply — electricity and natural gas sold to homes, C&I accounts, public-sector buyers and municipal aggregators across every competitive US market. Clean-energy products and services — sustainability, hedging and energy-management offerings sold around the supply book.

The strategic thrust is the data-center pivot: converting always-on clean generation into long-dated, investment-grade offtake contracts with the hyperscalers. The Crane Clean Energy Center — the restart of Three Mile Island Unit 1 for Microsoft, 835 MW, targeted in-service 2028 — is the flagship of that strategy and the template for more.

Business model and pricing

Revenue is booked as power sold (wholesale energy and capacity, retail supply margins, and contracted PPA volumes) plus the PTC floor. Constellation does not publish a single price sheet; the numbers that matter are market-set and contract-set. Wholesale capacity revenue keys off auction clears (the PJM ramp above). The nuclear PTC guarantees roughly $44.75/MWh downside. The Microsoft/Crane PPA is a 20-year contract whose per-MWh price has not been publicly disclosed — a recurring frustration for analysts trying to model it.

The headline guidance is the cleanest gauge: adjusted operating EPS of $11–12 for FY2026 (affirmed May 2026), after $8.67 in 2024, with management targeting ~20%+ base EPS growth and $8.4 billion of free cash flow before growth across 2026–2027 combined. Calpine is expected to add roughly $2/share of accretion on a full-year basis. The model, in short: a commodity-generation base with a legislated floor, re-rated by capacity scarcity, and topped with long-dated hyperscaler contracts.

Traction over time

The trajectory is a company that has roughly doubled in scale via one acquisition while running its core fleet at record reliability, all against an accelerating demand backdrop.

Market analysis

The tailwind is the biggest structural story in US power. After two decades of flat load, electricity demand is inflecting on data centers and AI. US data-center power demand is forecast to jump from ~31 GW in 2025 to ~66 GW in 2027 and ~134 GW by 2030 (S&P Global / industry estimates, 2025), with data centers rising to ~9% of US electricity by 2030 from ~4% in 2023 (EPRI). Bank of America and others put overall US electricity demand on a ~4% CAGR through 2030 — a regime change for generators who spent a generation managing decline.

The scarce input is firm, clean, 24/7 power, and nuclear is the purest form of it. That is why hyperscalers are signing 10–20 year nuclear PPAs and why PJM capacity prices went up roughly tenfold in two auctions. Constellation owns more of that scarce asset than anyone. The structural forces are durable — reactors take years to build or restart, so supply cannot respond quickly — but they are also priced: CEG trades at a rich multiple precisely because the market already believes the story.

Competitive intel

The competitive set is a handful of large-cap merchant generators, each attacking the same demand from a different base (full profiles in the sidebar). Vistra (~$64B, ~45 GW) is the scaled peer, Texas-gas-heavy with 6.5 GW of nuclear and a cheaper multiple. Talen ($18B, ~13 GW) is the concentrated pure-play around Susquehanna that pioneered the direct AWS nuclear-campus deal. NRG (~28 GW) competes mainly on retail and gas. NextEra is the deepest-pocketed rival, coming at data-center power with renewables-plus-storage and utility scale rather than merchant nuclear. And Calpine — once the largest US independent generator — is now inside Constellation, converting the biggest merchant-gas competitor into an internal hedge.

Where Constellation wins: the largest, purest zero-carbon nuclear fleet, the PTC floor, industry-best capacity factors, the marquee Microsoft deal, and the biggest retail book to wrap around it. Where it is exposed: it paid up, took on ~$12.7B of Calpine debt, and bolted a large gas-emissions footprint onto a “clean” story — the very thing peers with cleaner narratives can needle.

History and evolution

• 1816 / 2012 — Brand roots to Baltimore Gas Light; the Constellation name comes to Exelon via its 2012 acquisition of Constellation Energy Group.
• Feb 2022 — Exelon spins off its generation arm as Constellation Energy (Nasdaq: CEG); ~32.4 GW clean fleet, ~10% of US carbon-free power.
• 2023–2024 — AI/data-center demand re-rates the whole merchant-nuclear group; Moody’s upgrades CEG to Baa1.
• Sep 2024 — Microsoft 20-year PPA to restart Three Mile Island Unit 1 as the Crane Clean Energy Center; ~$1.6B capex, backed by a $1B DOE loan; targeted 2028.
• 2025 — Fleet runs at 94.7% capacity factor; FERC approves Calpine in July with a divestiture condition; DOJ resolution reached December.
• Jan 7 2026 — Calpine acquisition closes; 55 GW, largest US power producer.
• Mar 2026 — Agrees to sell 4.4 GW of PJM gas to LS Power for $5B to satisfy antitrust remedies.
• Q1 2026 — Revenue +64% to $11.12B; the stock, after an initial pop, later slid on Crane regulatory worry and an analyst target cut. In June 2026, an engineer was charged with insider trading tied to the TMI restart — a headline embarrassment, not a business risk.

What people say

The case for. Analysts are broadly constructive: a majority rate CEG Buy or Outperform with a mean target around $368 (mid-2026), and even bears who trimmed targets kept buy ratings. The bull case is simple and repeatable — the scarcest asset in the AI-power boom (firm, clean, 24/7 nuclear) with a legislated revenue floor, industry-best reliability, and marquee hyperscaler contracts. Credit-ratings agencies agree: Moody’s upgraded to Baa1 and S&P holds BBB+, two notches above investment grade, on strong metrics and policy support. Employees are positive too — a 4.0/5 Glassdoor rating across ~730 reviews, ~79% recommending the company, and compensation/benefits rated 4.3/5. One outlet called it “the only nuclear utility that looks like a tech stock.”

The complaints. The sharpest is valuation and volatility: CEG was down double digits year-to-date at points in 2026, and even after the Q1 beat the stock fell about 11.6% as enthusiasm cooled around Crane’s regulatory path — Argus cut fair value from $425 to $350. The critiques cluster: (1) execution and regulatory risk on Crane and other complex nuclear/grid projects, with the Microsoft PPA’s economics never publicly disclosed; (2) post-Calpine leverage — ~$12.7B of assumed debt on a deal many felt was expensive; (3) gas exposure — Calpine adds a large fossil footprint that dilutes the zero-carbon narrative; (4) dependence on data-center demand actually materializing at forecast pace, and on PJM capacity prices staying elevated when they are volatile and politically charged (the “reliability tax” driving up consumer bills invites regulatory backlash); (5) nuclear operational risk — a single extended outage dents the whole thesis. And local opposition to the Three Mile Island restart — including residents from the 1979 accident era — is a reputational and permitting overhang.

Outlook: well positioned or at risk?

Constellation is well-positioned — it owns more of the single scarcest resource in the defining energy story of the decade than any other company, and it is converting that ownership into long-dated, investment-grade cash flows while a legislated PTC floor caps the downside. That is as strong a structural position as exists in US power. The nuclear fleet runs at record reliability, the demand backdrop is a genuine regime change rather than a cycle, and the retail book plus Calpine’s gas fleet give it flexibility peers lack.

The honest caveats are real and worth stating plainly. The stock already prices the story, so the asymmetry is unattractive: beat expectations and the reward is modest; miss on Crane’s timeline, a nuclear outage, softening PJM capacity prices, or the debt, and the multiple has a long way to fall — the 11.6% post-earnings drop showed how quick that repricing is. Calpine bought scale and a hedge but added leverage and emissions that complicate the clean narrative. And the company is now partly a bet on regulators and hyperscalers behaving as modeled. None of that unseats the core judgment: the assets are irreplaceable, the demand is arriving, and the floor is legislated. This is a well-positioned incumbent whose risk is priced-in disappointment, not disruption.

How a challenger would attack it

You can’t build a reactor fleet — so attack the contract, the clock, and the bill. Constellation’s moat is physical and irreplicable on any venture timeline, which pushes the attack to its three soft points. First, speed: hyperscalers need power in 2026-2027, and CEG’s flagship Crane restart doesn’t come online until 2028 amid regulatory wobble — a challenger assembling fast-deploy gas turbines, storage, and behind-the-meter campuses (the Talen/AWS powered-land model, executed faster) wins the load Constellation can’t serve yet. Second, the demand side: PJM capacity prices went from $28.92 to the $329.17 cap in two auctions, with ~40% attributed to data centers — that “reliability tax” on consumer bills is a political time bomb, and a challenger positioned as the answer to it (demand response, flexible-load orchestration for data centers, grid-services software) monetizes the backlash rather than the scarcity. Third, the narrative seam Calpine opened: CEG bolted the country’s largest gas fleet onto a zero-carbon story, so a renewables-plus-storage player à la NextEra can now credibly outbid it on actual carbon math for sustainability-mandated buyers. The floor CEG can’t lose — the PTC and its 94.7% capacity factor — is real; everything above the floor is a contract fight a nimbler counterparty can win deal by deal.

Same playbook, new buyer

Constellation’s playbook — own scarce firm power, wrap a retail book around it, sell long-dated offtake to credit-worthy load — is replicable at scales and in geographies the $98B giant won’t bother with. The clearest shift is downmarket: hyperscalers get Dominguez’s attention, but the second tier — colocation operators, sovereign AI projects, industrial electrifiers, hydrogen producers — needs the same 24/7 clean firm power in 50-200 MW blocks, and an aggregator that packages smaller nuclear, hydro, and geothermal assets into bankable PPAs for them faces no direct CEG competition. Second, geography: the AI-power squeeze is replicating in Europe and Asia, where merchant nuclear pure-plays barely exist and incumbent utilities are state-entangled — the CEG spin-off thesis (unbundle clean generation, let it trade on scarcity) is exportable by anyone who can buy or contract reactors there. Third, the restart model itself: Crane proved a mothballed reactor plus a hyperscaler PPA plus a DOE loan is a viable financial product; a specialist restart developer hunting the remaining shuttered or at-risk US units runs that playbook while CEG is occupied digesting $12.7B of Calpine debt and a $5B forced divestiture. The incumbent’s constraint is attention and balance sheet, not insight.

Sources and further reading

• Constellation Completes Calpine Transaction (Calpine, January 2026)
• Constellation Energy Q1 2026 EPS $2.74, guidance affirmed (StockTitan, May 2026)
• Constellation Energy (CEG) Is Down 11.6% After Calpine-Fueled Q1 Earnings Surge (Simply Wall St, 2026)
• Constellation plans 2028 restart of Three Mile Island unit 1, spurred by Microsoft PPA (Utility Dive, September 2024)
• Constellation to Sell 4.4 GW of PJM Gas Power Assets to LS Power for $5B (POWER Magazine, March 2026)
• PJM capacity prices hit price cap, reserve shortfall grows (Utility Dive, 2025)
• Exelon Completes Spin-Off Of Constellation Energy (Forbes, February 2022)
• Moody’s Upgrades Constellation’s Credit Rating (Constellation, 2024)
• Data center grid-power demand to rise 22% in 2025, nearly triple by 2030 (S&P Global, 2025)
• Vistra vs Constellation vs Talen: Best AI Power Stocks (2026) (Lambda Finance, 2026)