Constellation Energy (Nasdaq: CEG)

1. Snapshot

2. What the company does

In the restructured parts of the US, including the Mid-Atlantic, Illinois, New York, New England and Texas, independent generators keep the lights on. They sell into wholesale markets run by grid operators and to customers who choose their supplier. Constellation, spun out of Exelon on February 1, 2022, is the largest of them by nuclear capacity. It serves two needs from one fleet: grid operators' need for firm supply every hour, and large buyers' wish for carbon-free power at a known price.

Tracing one MWh.

• A Pennsylvania reactor burns uranium bought under long-term contracts. Fuel expense was $952M in 2025, about $5/MWh. It runs nearly every hour, with refuelling outages averaging 22 days every 18–24 months.
• The MWh is sold one of three ways: into PJM's spot market, through the retail book to a business on a contract of about two years, or under a 20-year contract such as those with Microsoft and Meta.
• The plant is also paid for being available. At $329/MW-day, capacity is worth roughly $13–14 per MWh for a unit that runs all the time (inferred).
• Grid operators settle weekly to monthly, and retail customers pay within about a month.
• Costs barely move with price. O&M was $6,159M in both 2024 and 2025. So price changes fall almost entirely to profit.

Retail is the hedge, not the profit centre. The company served ~147 TWh of retail load, mainly commercial and industrial (C&I), plus ~57 TWh of wholesale load in 2025. It claims over 32% of the direct C&I market, with 77% renewal on power contracts. The book lets it sell its generation forward at a margin over wholesale cost. Headline revenue ($25.5B in 2025) includes this pass-through, so the company measures segments on revenue net of purchased power and fuel (RNF).

From 2021 to 2025 owned capacity was nearly flat, at 32,400 MW and then 31,676 MW. The mix moved toward nuclear and away from gas. Calpine reverses that. It adds 21 GW of gas, the 730 MW Geysers geothermal complex, ~800 MW of batteries and ~62 TWh a year of retail load, mostly in Texas, California and the Northeast.

3. Industry, competitive position & moat

Restructured power has four layers: generators, non-profit grid operators, regulated wires utilities and retailers. Only generators carry real price exposure. Nuclear plants gain the most when prices rise, because their costs are fixed while prices are usually set by the marginal gas plant. PJM's annual price tracked Henry Hub gas with a correlation of about 0.88 over 2005–2025 (inferred from Market Monitor and EIA data).

The market is regional. PJM, ERCOT, NYISO, ISO-NE and CAISO run separate markets with different rules, and ERCOT has no capacity market. Constellation's weight is in PJM, whose capacity auction has cleared at its administrative cap three times running.

Constellation holds about two-thirds of the merchant nuclear fleet. EIA counts 96 reactors and ~98.4 GW of US nuclear, about 18% of US generation in 2025. Constellation's ~22 GW is about 22% of that. Most other reactors belong to regulated utilities and earn a regulated return. Of the roughly one-third that sells at market, Constellation owns about two-thirds (inferred). Merchant peers are far smaller in nuclear: Vistra has 6.4 GW, Talen 2.2 GW, PSEG ~3.7 GW, and NRG none.

Barriers that bind:

• New nuclear. Vogtle 3 and 4 cost about $34B and ran about seven years late, and no large reactor is under construction. The first commercial SMR construction permit came only in September 2026.
• New gas. Turbines are being reserved for 2031 delivery, and interconnection takes a median four to five years.

Barrier that doesn't: retail. The company itself calls retail entry barriers low.

What a rival could not reproduce. First, the fleet itself: licensed, depreciated, connected to the tightest market, and impossible to build or buy at scale. Second, operations: a 94.7% capacity factor, with outages of 22 days against an industry 33–38. One point of fleet capacity factor is ~1.9 TWh, about $95M a year at $50/MWh (inferred). Third, a BBB+/Baa1 balance sheet that can carry 20-year obligations and the collateral hedging requires.

Who wins here, and does CEG fit? The winner is the owner of the largest, lowest-cost, longest-lived firm fleet in the tightest market, with the credit to sign long contracts and the regulatory skill to shape market rules. Constellation fits better than any peer. Its own complaint started the FERC proceeding behind the December 2025 co-location order. The caveat is that size makes it the natural target of price caps and antitrust remedies: the DOJ required divestitures before Calpine could close. Peers have also signed hyperscaler nuclear deals (Talen–Amazon for 1,920 MW, Vistra–Meta for 2,176 MW), so Constellation is the largest seller, not the only one.

4. Growth engine

Volume hardly grows. Nuclear output was 175 TWh in 2021 and 183 TWh in 2025, mostly because of the South Texas stake. Growth comes from price, contract mix and acquisitions. Adjusted EPS went from $6.28 (2023) to $8.67 (2024) and $9.39 (2025), with diluted shares down only from 324M to 314M. In 2025, tax-credit revenue fell from $2,080M to $320M as prices rose above the phase-out, yet adjusted earnings still grew. Market revenue replaced the subsidy.

Management targets base EPS growth above 20% a year through 2029 including Calpine, then a "rolling 10%+". The target has been raised twice, from at least 10% to at least 13% and then to the current figure. No filing reconciles results against the earlier targets.

5. Margin, cash & capital allocation

Against ~$40/MWh of net revenue sits a fixed cash cost base of roughly $30 per owned MWh: O&M excluding decommissioning accretion, plus taxes other than income (inferred, including retail and gas-fleet costs). Depreciation is low at $985M, because the plants are old and largely written down. Decommissioning is pre-funded, with $19.4B of trust assets at end-2025 against a $12.9B obligation. About $5.3B of the trust is owed back to ratepayers.

Cash flow needs care. A receivables-sale facility made GAAP operating cash flow negative from 2021 to 2024, so 2025's $4.2B is the first clean year. The company's preferred measure, free cash flow before growth, is disclosed only as forecasts: $8.4B for 2026–27 and $11.5–13B for 2028–29.

Capex roughly doubles to $5.7B in 2026 and $4.7B in 2027. That includes $3.9B of growth spending (Crane, uprates, co-location, licences), and ~29% is nuclear fuel, partly inventory built against Russian supply risk.

Where the cash went, 2022–2025: capex $9.6B, buybacks $2.4B (about 17M shares), acquisitions $1.7B (the South Texas stake), dividends $1.5B. Net borrowing was only ~$0.9B. The dividend nearly tripled, from $0.564 to $1.55 per share, with a stated ~10% annual growth policy.

Then Calpine. The price was 50M new shares plus $4.5B cash: $21.8B of consideration plus ~$12.6B of assumed debt, against a stated $26.6B net price (7.9x 2026 EV/EBITDA). The deal created $11.1B of goodwill. Required sales of PJM and Texas plants to LS Power for $5.0B and $0.86B are agreed but had not closed at the latest filing. Debt nearly tripled to ~$24.7B, and management targets its credit metrics by end-2027. Buybacks continued anyway, including 2M shares bought from former Calpine owners in their June 2026 secondary sale.

Incentives reward the same things. The annual bonus is 70% adjusted operating earnings. Long-term awards are 67% free cash flow before growth and 33% relative TSR, with a credit-rating penalty.

6. Cyclicality, constraints & what to monitor

Electricity demand barely falls in recessions, so the exposure is to price. Including Calpine, first-half 2026 segment margin came 28% from Calpine, 24% each from Midwest and Mid-Atlantic, 10% from New York, 7% from Other and 6% from ERCOT.

Hedging changed after 2023. Before then the company sold ~90%, 60% and 30% of the next three years forward. Once the tax credit began, it stopped hedging generation that way and stopped disclosing hedge levels. The only current disclosure says a $10/MWh fall in power prices plus a $5/MWh fall in spark spreads would have a "not material" effect on 2026–27 earnings.

The floor, precisely. §45U pays up to $15/MWh and loses $0.80 per $1 of receipts above ~$26. It reaches zero at ~$44.75 (2025 values) and ends after 2032. Between $26 and $45, receipts plus credit hold at roughly $41–45. 2024 proved it: PJM prices averaged $33.74, the company booked $2,080M of credits, and adjusted EPS still rose 38%.

Where the cycle stands. On every price measure the business is near the top of its own range. PJM energy averaged $50.73 in 2025, 50% above 2024 and well above the 2012–2020 average of ~$34, though below the 2008 and 2022 spikes. Capacity is at the cap. The tax credit pays almost nothing ($25M in 1H 2026), and capacity factor sits inside its historical 94–96% band.

Downside mechanism. If gas fell back to $2–2.50, the credit would rebuild nuclear receipts toward ~$41–45, as it did in 2024. The bigger exposure is capacity. A drift from the $325 cap to the ~$175 floor on ~18,900 cleared MW is about $1B a year (inferred). After 2032 the floor is gone.

7. Risks & open questions

• Political rationing of scarcity rents. Uncapped PJM capacity would clear at ~$530–555/MW-day. Governors have threatened to leave PJM, and states are debating utility-owned generation. As the largest seller, Constellation is the visible beneficiary of each spike, which invites the next cap.
• Demand routed around existing plants. If new large loads must bring new capacity, through the backstop procurement or interim adequacy rules, hyperscaler money funds new gas and batteries rather than premiums for existing nuclear. The early-2026 contracting "pause" showed how sensitive deals are to these rules.
• Leverage meeting a price reset. Debt is ~$24.7B and goodwill $11.1B. A price fall before end-2027 would hit earnings and credit metrics together. A downgrade below investment grade would require ~$3.4B of collateral.
• Fleet-wide regulatory exposure. All 25 units answer to the NRC, which cut Peach Bottom's 80-year licence in 2022 before restoring it in 2025.
• Fuel. Russian enriched uranium is fully banned from 2028, and ~35% of 2026–2030 uranium needs come from three suppliers.
• Calpine's gas. 21 GW more exposure to spark spreads, ERCOT softness and California adequacy rules, plus antitrust friction on further PJM growth.

What the filings could not answer:

• 2026–2028 hedge levels.
• Contract prices beyond the $20–50/MWh "value" range, and how PPA revenue counts toward §45U receipts.
• Post-Calpine capacity by fuel and region.
• Historical free cash flow before growth, and the 2025 capex split.
• Q4 2024 and Q4 2025 earnings materials.
• The economics of the Illinois plants after the May 2027 expiry of state support.

To resolve before forming a thesis:

• contracted versus merchant output and pricing for 2028–2032;
• whether PJM's caps outlive 2029/30;
• Calpine's realised cash yield against its $26.6B net price.

8. Investor takeaways

1. What it is: ~22 GW of irreplaceable, largely depreciated nuclear plants, now joined to ~23 GW of Calpine gas and a large retail book, selling mostly into PJM.
2. Core engine: low fixed-cost MWh sold at gas-set energy prices plus capacity payments, with a credit floor through 2032. Each $1/MWh ≈ $0.43/share before hedges.
3. Main growth lever: converting merchant output into 15–20-year contracts, plus PJM scarcity and Calpine in 2026. Volume grows only through Crane and uprates.
4. What could break it: regulators capping PJM prices and steering data-center demand to new build while debt peaks.
5. Monitor: PJM auction and backstop outcomes, contract MW and pricing signals, gas, and debt against the end-2027 target.