Constellation Energy

Phase A — Understand the business

Lens 1 · Company Overview

Constellation is the largest competitive (merchant) power producer in the United States, built around the country's largest nuclear fleet and — as of January 2026 — a large natural-gas and geothermal fleet acquired from Calpine. It was spun out of Exelon in February 2022 as a pure-play generation + competitive-supply company (Exelon kept the regulated wires utilities).

How it makes money — two engines:

1. Generation (the economic core). ~32.4 GW of nuclear plus the Calpine gas/geothermal fleet. Nuclear runs near-baseload at a ~92–94% capacity factor (Q1 2026: 92.3%, vs 94.1% a year earlier), producing low-marginal-cost MWh sold into wholesale markets (PJM, MISO, NYISO, ERCOT, ISO-NE) or under bilateral PPAs.
2. Customer-facing competitive supply. Sells electricity and gas to ~2 million customers — utilities, municipalities, co-ops, C&I, government, and residential — in competitive retail markets. Calpine materially expanded this retail platform.

Contract structure & key terms. The strategic pivot is from merchant price exposure → long-dated firm PPAs with credit-worthy hyperscalers:

• Microsoft / Crane Clean Energy Center (ex-Three Mile Island Unit 1): 20-year virtual PPA for all output (~835 MW), restart targeted late 2027. Morgan Stanley pegged the deal at ~$1.70 EPS / ~$700M EBITDA / ~$450M FCF annually.
• Meta / Clinton Power Station (IL): 20-year PPA beginning 2027.
• Nuclear PTC floor (IRA §13105): a federal price floor of $43.75/MWh on nuclear output through 2032 — downside-protected, upside-uncapped. This is the structural backstop under the whole fleet.

Plain-terms thesis: CEG owns ~30 GW of already-built, already-permitted, zero-carbon baseload that costs ~$30–50/MWh to run and is now signing 20-year contracts at premium prices to power AI data centers. The bet is 20 years of structural margin expansion on a depreciated asset base. Calpine adds dispatchable gas to firm-up that clean offering and a Texas/California footprint.

Lens 2 · Supply Chain

Map: fuel & equipment → CEG generation → grid/PPA → end customer.

• Upstream — nuclear fuel cycle (the real chokepoint):
• Uranium → conversion → enrichment → fuel fabrication. The Western enrichment bottleneck runs through Urenco and Orano; Centrus Energy (LEU) is the only US-based enricher scaling HALEU. Russia (Rosatom/Tenex) historically supplied ~20%+ of US enrichment — displaced post-2024 import ban, tightening the Western supply chain.
• Westinghouse (owned by a Brookfield / Cameco JV — not by CEG) supplies fuel assemblies and reactor services to the CEG fleet. This is a key single-vendor-class dependency for fuel fabrication and outage services.
• Equipment / restart vendors: the Crane restart requires NRC-licensed refurbishment — turbine, generator, transformer, and the 760 MW of Capacity Interconnection Rights transferred from the Eddystone plant (FERC waiver granted Jun 1, 2026). A pending Susquehanna River Basin Commission water-intake permit (~73.2M gal/day) remains a named open item.
• The company: ~55–60 GW combined fleet post-Calpine — nuclear (PJM/MISO/NYISO/ERCOT via STP), gas (Calpine: ~10 GW TX, ~10 GW PJM/NE, ~8 GW CA/West), geothermal (The Geysers, CA — world's largest geothermal complex), plus solar/wind/storage.
• Midstream — the grid: PJM is the dominant exposure (Midwest ~11.9 GW + Mid-Atlantic ~10.5 GW pre-Calpine). Grid interconnection timing and FERC co-location rules (Lens 10) are the chokepoint between CEG's MWh and the data centers that want them.
• Downstream — end customers: hyperscalers (Microsoft, Meta), ~2M retail/C&I customers, and merchant wholesale buyers across five ISOs.

Chokepoints, named: (1) enrichment / fuel fabrication (Westinghouse-Brookfield/Cameco + the Western enrichment oligopoly); (2) grid interconnection & FERC co-location rules (PJM); (3) NRC licensing throughput for restarts and 80-year life extensions.

Lens 3 · Competitive Advantages (moats)

• Irreplaceable asset base (the core moat). You cannot build a new US nuclear reactor on a relevant timeline — Vogtle 3 & 4 ran years late and billions over. CEG's ~30 GW of operating, permitted nuclear is a replacement-cost moat: the only way to get more zero-carbon baseload at scale this decade is to buy CEG's output or restart a mothballed unit (which only CEG, Holtec/Palisades, and now Vistra/Perry are even attempting). This is the durable edge.
• Cost-curve position. Existing nuclear marginal cost ~$30–50/MWh vs PPAs at $80–120/MWh = ~20 years of locked margin on contracted volume. The IRA PTC floor ($43.75/MWh) removes most downside.
• Scale & operating excellence. Largest US operator → best-in-class capacity factors (>92%), regulatory relationships, refueling-outage know-how. Scale also funds the uprate/life-extension capex that smaller operators can't.
• Bargaining power. Shifting decisively toward CEG: hyperscalers need firm, clean, 24/7 power and there are very few sellers of it. The Microsoft/Meta deals show CEG can extract premium, long-dated, take-the-output terms. Against suppliers (Westinghouse, fuel) CEG has less leverage — it is one of several buyers of a concentrated input.
• Where the moat is weaker than bulls claim: the gas fleet from Calpine is not moated — gas peakers/CCGTs are buildable and Vistra (Cogentrix, +5.5 GW), NRG, and developers are adding them. The moat is the nuclear, not the 55 GW headline.

Lens 4 · Segments

Hard-requirement caveat: segments.csv does not exist in the research layer. No segment figure here can be -sourced; all are . Treat segment granularity as low-confidence until a 10-Q is ingested.

CEG historically reported a single integrated generation-and-supply business sliced by region, not by clean product lines. Pre-Calpine regional capacity:

• Midwest (western PJM + most of MISO): ~11.9 GW
• Mid-Atlantic (eastern PJM): ~10.5 GW
• NYISO: ~3 GW
• ERCOT: ~3.6 GW
• Other (NE, South, West, Canada): ~3.3 GW

Post-Calpine the mix shifts hard toward gas and toward Texas/California. Calpine adds ~10 GW ERCOT, ~10 GW PJM/NE, ~8 GW CA/West. Trend & cause: the revenue base roughly doubled overnight (Q1 2026 rev $11.1B vs $6.8B — see Lens 5) because the Calpine retail + gas volumes consolidated. The margin quality mix is the watch item: nuclear is high-margin/contracted; merchant gas is lower-margin/more cyclical. The 4.4 GW PJM-gas divestiture to LS Power (Lens 10) trims the gas overlap. Net direction: accelerating revenue, diluting average margin %, but adding dispatchable firming capability that makes the clean-PPA product more saleable.

---

Phase B — Measure performance

Lens 5 · Earnings Result (Q1 2026, reported May 11, 2026)

• Vs consensus: adj op EPS $2.74 beat the ~$2.59 estimate (+$0.15 / ~+6%); revenue $11.12B vs ~$9B expected. Stock +4.1% pre-market on the print.
• What drove it: the year-on-year revenue double is overwhelmingly the Calpine consolidation (closed Jan 7, 2026) — gas generation + the expanded retail book. The GAAP EPS jump to $4.49 also includes acquisition-accounting/mark-to-market noise; adjusted EPS ($2.74) is the cleaner read of underlying earnings power.
• Margins: the apparent margin compression (net income up ~13x but revenue only ~1.6x) reflects gas/retail's lower margin % vs nuclear — exactly the mix-shift flagged in Lens 4. The Q1 net-income spike is flattered by non-recurring items.
• Guidance & tone: reaffirmed FY2026 adjusted operating EPS of $11.00–$12.00 on ~361M shares. Tone positive — CEO Dominguez: "We had a good first quarter thanks to our folks. We will continue to strive through the balance of 2026".
• Balance-sheet flags (material): total assets jumped to $96,911M from $57,249M at YE2025; long-term debt rose to $16,994M, short-term borrowings to $5,102M; the deal added ~$18,481M PP&E and ~$11,107M goodwill. Goodwill of this size is now a real impairment-watch item (Lens 10).
• Unusual vs own history: the company's entire scale and capital structure reset in one quarter. Pre-Calpine ND/EBITDA ~0.8x → pro forma ~2.25x. This is a different-risk company than the 2022–2025 pure-nuclear CEG.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts/ in the research layer — assessed from web summaries of the Q1 2026 call + 2024–2025 context. Lower confidence than a transcript-grounded read.

• Persistent themes (multi-quarter): "firm, clean megawatts," "strategic partner," premium contracting with hyperscalers, fleet life-extension (80-year licenses), nuclear PTC as a floor. Management has consistently framed CEG as the seller of scarce 24/7 clean power.
• Tone trajectory: 2024 = euphoric (Microsoft/TMI deal, the stock +20% in a day, a wave of PT hikes). Early 2025 = defensive (DeepSeek shock forced management to argue AI power demand was real and durable). 2026 = confident-operational — the message moved from "the demand is real" to "we now have the scale (Calpine) and the contracts to serve it; execute." The Q1 2026 tone was steady and reaffirming, not promotional.
• What they stopped saying: less emphasis on pure-nuclear identity; more on "America's leading producer of clean and reliable (i.e., dispatchable gas) energy" — a deliberate reframing to justify Calpine.
• Watch: whether forward calls quantify Calpine synergy capture and put numbers on the FERC co-location outcome, rather than describing demand qualitatively.

Lens 7 · Comps

Read: CEG trades at a clear premium to Vistra (~26x vs ~18x fwd P/E; ~13x vs ~10x EV/EBITDA). The premium is defensible on asset quality and balance sheet — CEG's nuclear is more contracted/clean and its leverage (2.25x pro forma) is far below Vistra's ($19.6B net debt). But ~26x forward for a power generator only makes sense if you underwrite the AI-power growth as durable. The comp gap is the debate: bulls say CEG deserves the premium for the best assets; bears say a generator at 26x is priced for a secular story that a single FERC ruling or AI-capex pause can re-rate.

Lens 8 · Stock-Price Catalysts (>5% moves, ~5yr)

The tape tells you exactly what this stock reacts to — AI-power deal news and AI-capex sentiment, not quarterly operating beats:

• Sep 2024: +20% in a day on the Microsoft/TMI 20-year PPA — the defining re-rate that turned CEG into "the nuclear AI trade".
• Jan 27, 2025: −21% on the DeepSeek shock (alongside VST −28%, NVDA) — the market briefly decided efficient AI models would need less power. Pure sentiment, no change to CEG fundamentals.
• Jan 2025: Calpine deal announced ($16.4B) — strategic re-rate of the business mix.
• 2026 YTD: −18% to −25% from the 52-week high — the AI-power trade cooled, "Wall Street's fascination with nuclear faded," and the stock de-rated despite reaffirmed guidance.
• Jun 1–2, 2026: positive on the FERC Eddystone CIR waiver de-risking the Crane restart.

Pattern: CEG is a high-beta expression of the AI-infrastructure thesis. It moves on (1) hyperscaler PPA announcements, (2) macro AI-demand sentiment (DeepSeek), and (3) nuclear-specific regulatory milestones (FERC, NRC) — far more than on its own EPS line. That makes it a narrative stock as much as a cash-flow stock, which cuts both ways.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO — Joseph Dominguez (President & CEO since the Feb 2022 spin). Career insider: President & CEO of Exelon Generation (2021–22), senior roles at ComEd, regulatory/policy background. He is the architect of the post-spin strategy: lobbied for the IRA nuclear PTC, signed the Microsoft/Meta PPAs, and drove the Calpine and South Texas Project acquisitions.
• Track record (quantified): completed the South Texas Project stake purchase (NRG's 44%, 2,645 MW dual-unit) for $1.75B ahead of schedule (NRC approved Oct 30, 2023); closed Calpine Jan 7, 2026; secured the first hyperscaler nuclear-restart PPA in US history (Microsoft/TMI). On his watch the equity went from spin-out to a ~$90B+ cap — a genuinely strong value-creation record, though much of it rode the AI-power wave.
• Skin in the game / insider activity: no insider purchases by Dominguez in the prior ~18 months. Not a red flag per se (large-cap utility CEOs rarely open-market buy), but no recent conviction-buy signal either. Insider ownership level: n/a (would need the proxy).
• Capital allocation: the defining question. Bought STP cheaply; Calpine at 7.9x EV/EBITDA is full but defensible for the scale/dispatchability. Authorized a $5B buyback and guides $3.9B growth capex at "double-digit returns". The risk: paying ~$26.6B net for a gas fleet right as the equity's whole premium rests on the nuclear/clean story — a mix-dilutive use of balance sheet if AI-power demand under-delivers.
• Founder vs professional manager: professional manager / insider operator (not a founder). Implication: disciplined, regulatory-savvy, M&A-led growth — appropriate for a regulated-adjacent generator, but the equity story now depends on his M&A and contracting judgment, not on an owner's long-horizon conviction.

Lens 10 · Forensic Red Flags

All figures `` (no filings ingested). A filings-grounded refresh should re-run this lens against the actual 10-Q/10-K.

• Goodwill (top watch item): the Calpine deal created ~$11.1B of goodwill on a now-$96.9B asset base. If AI-power demand or merchant gas economics disappoint, this is the impairment exposure. Track the goodwill line and any segment-level impairment testing.
• GAAP vs adjusted divergence: GAAP EPS $4.49 vs adjusted $2.74 in Q1 2026 — a wide gap driven by acquisition accounting and mark-to-market on the commodity/derivative book. Nuclear-decommissioning-trust gains/losses and hedge MTM routinely swing CEG's GAAP earnings; adjusted operating EPS is the right metric, but verify the add-backs are non-recurring, not recurring costs dressed as one-offs.
• Leverage step-change: ND/EBITDA ~0.8x → ~2.25x pro forma; short-term borrowings $5.1B. A ratings downgrade would trigger up to ~$2,972M of collateral postings on derivatives — a real liquidity-cliff risk if investment grade is lost. Currently IG.
• Revenue recognition / commodity book: as a merchant generator + retailer, CEG carries a large derivative/hedge book; gross revenue ($11.1B/qtr) is not a clean read of economic activity. Watch for receivables/collateral swings with power-price volatility.
• Nuclear decommissioning trusts & ARO: asset-retirement obligations and the funded status of decommissioning trusts are a perennial nuclear-operator estimate risk (discount-rate and return assumptions). Not quantified here — flag for filings review.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. Verified via SEC EDGAR EFTS (LR + AAER) over 2021-06-17 → 2026-06-17 — regulatory/regulatory-findings.md shows 0 SEC findings [research file written 2026-06-17].
• Item 3 (Legal Proceedings): n/a — most recent 10-K not in filings/ (not ingested per WAVE constraint). Flag for a filings-grounded refresh.
• Non-SEC enforcement (web): No material enforcement hits surfaced for Constellation Energy (the 2026 corporate entity) in this pass. The relevant regulatory activity is structural, not enforcement — the FERC/DOJ antitrust conditions on the Calpine merger (the 4.4 GW LS Power divestiture, see below) and the FERC co-location docket (Lens 12/13), which are deal/market-design matters, not fraud or misconduct.
• Bottom line: No material regulatory or legal integrity findings — verified via SEC EDGAR EFTS (LR, AAER) and web search as of 2026-06-17. 10-K Item 3 not checked (filing not ingested). The live regulatory exposure is market-design risk (FERC co-location), covered in Phase D.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (EPS, FY2026 → FY2028)

Built bottom-up from company guidance. Output ``; arithmetic shown. Brier forecast NOT logged — this is the unattended --watchlist breadth loop (per skill rules, forecast.ts create is skipped here).

Anchor: management guides FY2026 adjusted operating EPS $11.00–$12.00, on ~361M shares, and targets base EPS growth >20% from 2026→2029 with a floor commitment of ≥10% over any rolling 3-year window.

Key input lines (all `` unless noted):

• Buyback: $5B authorized on ~$90B cap → ~5.5% of shares over time → ~1–2%/yr EPS tailwind.
• Crane/Microsoft: ~$1.70 incremental EPS once online (~late 2027) — phases into 2028+, not 2026.
• Nuclear PTC floor ($43.75/MWh through 2032) caps downside on ~30 GW of output.
• Risk to the model: merchant gas margin (Calpine) is cyclical and not in the contracted base — the bear path is mostly a gas-margin + FERC-outcome story.

Implied valuation check: at ~$254–271 on base FY2026 $11.50 → ~22–24x current-year; on base FY2028 $15.20 → ~17x two years out. Not cheap, not absurd if the growth compounds. The whole call rests on whether the >15% EPS CAGR is real.

Lens 12 · Bull vs Bear

Bull case. CEG owns the single scarcest asset in the AI build-out: ~30 GW of operating, permitted, zero-carbon baseload that cannot be replicated this decade. Hyperscalers will pay premium, 20-year, take-the-output prices for firm clean power (Microsoft, Meta already signed); the IRA PTC floor removes downside; Calpine adds the dispatchable gas that makes the clean product saleable 24/7 and a 2M-customer retail channel. Life-extension (80-yr licenses) and uprates add low-cost GW with no new build. Management guides >20% base EPS growth to 2029, backed by a $5B buyback and double-digit-return growth capex. If even half the announced AI-power demand materializes, today's ~26x forward will look cheap in three years. Contrarian view the market is missing: the 2026 de-rate (−20%+) treated CEG as if the AI-power trade is over — but the structural scarcity of clean baseload is independent of any single quarter's AI-capex headlines; the contracts are 20 years, the asset is irreplaceable, and the buyer set is only getting more desperate for power.

Bear case (2–3 permanent-impairment risks).

1. FERC co-location / behind-the-meter ruling (the binary). On Dec 18, 2025 FERC declared PJM's co-location tariff "unjust and unreasonable" and ordered new rules (compliance filings Feb–Apr 2026; FERC to act on the large-load docket by June 2026). FERC already rejected the Talen-Amazon-Susquehanna behind-the-meter expansion in Nov 2024. If the final framework forces co-located data centers to pay full grid/transmission charges (front-of-the-meter economics), the premium in the hyperscaler-PPA thesis compresses — the entire re-rate rationale weakens. This is the risk bulls most underweight.
2. AI-capex air-pocket. CEG is a high-beta proxy for hyperscaler capex (DeepSeek proved a sentiment shock = −21% in a day). A genuine slowdown — efficiency gains, a capex-digestion pause, or a single hyperscaler pullback — de-rates the whole complex regardless of CEG's contracts.
3. Mix dilution + leverage. CEG paid ~$26.6B net for a gas fleet while the equity premium rests on nuclear/clean. Pro forma ND/EBITDA ~2.25x (from 0.8x) and ~$11.1B goodwill mean less balance-sheet slack and real impairment exposure if gas margins or AI demand disappoint.

Pre-mortem (18 months out, thesis broke): FERC's mid-2026 co-location framework lands restrictive; one hyperscaler pauses data-center power procurement amid an AI-capex digestion; merchant gas spreads compress; the Crane restart slips past 2027 on the water permit or NRC. The stock, having entered at ~26x on a >15% growth story, re-rates to ~16x on ~10% growth → a 30–40% drawdown even with EPS still rising.

Are multiples too high? At ~26x forward, yes relative to peers (VST ~18x) and to a generator's historical range — the premium prices in durable AI-power growth. It is not too high if you underwrite the 20-year contracted clean-baseload thesis as structural. The honest answer: the easy money (the 2024 re-rate) is gone; from here it's a fundamentals-must-deliver story.

Lens 13 · Devil's Advocate (short-seller)

• What structurally breaks the model: the value rests on hyperscalers paying a premium for clean firm power and being allowed to co-locate efficiently. FERC can legislate that premium away. The Dec-2025 order + the Talen-Amazon rejection show the regulator is actively reshaping co-location economics. If behind-the-meter loads must pay full transmission/grid charges, the clean-power premium narrows toward merchant economics — and a merchant generator does not deserve 26x.
• Revenue concentration: the growth thesis is concentrated in a handful of hyperscaler counterparties (Microsoft, Meta) and one restart (Crane). If one walks, slips, or renegotiates, a disproportionate share of the incremental story evaporates. The base fleet is diversified; the upside is not.
• Most dangerous competitor bulls underestimate — Vistra. VST is doing the same playbook (Perry nuclear restart, 3,800 MW Meta/AWS PPAs, Cogentrix +5.5 GW gas) at ~18x vs CEG's ~26x. If the market decides nuclear-IPP exposure is fungible, CEG's premium compresses toward VST — a de-rate independent of operations. Talen (Susquehanna/AWS) is a third credible clean-baseload seller.
• Worst capital-allocation move: spending ~$26.6B net on Calpine's gas fleet at a full 7.9x while the equity's entire premium is the nuclear/clean story — and levering up 0.8x→2.25x to do it. If gas margins normalize down, that's value-dilutive at the worst possible multiple.
• Assumptions that must hold for today's price: (1) FERC co-location lands favorably (or at least not restrictively); (2) AI-power demand stays on its announced trajectory for 20 years; (3) Crane restarts ~on time at the contracted economics; (4) merchant gas margins hold; (5) >15% EPS CAGR compounds. Break any one of (1)–(3) and the multiple is too high.
• −20–30% growth scenario: if EPS growth disappoints to ~8–10% (vs >15% baked in), the multiple compresses to a peer-like ~17–18x → the stock re-rates 30%+ lower even with rising EPS.
• Single scenario that permanently impairs: a restrictive FERC co-location framework combined with a genuine AI-capex slowdown — turning CEG from "scarce premium clean baseload" back into "a levered merchant generator with a gas fleet bought at the top." Plausibility: moderate — the FERC piece is a live, dated catalyst (mid-2026); the AI-capex piece is sentiment-driven and has already happened once (DeepSeek).

Lens 14 · Management Questions (ordered by information value)

1. What is your base, bull, and bear read on FERC's mid-2026 PJM co-location framework, and quantify the EPS at risk if behind-the-meter loads must pay full transmission charges.
2. How much of the >20% 2026→2029 base EPS growth depends on the Crane restart and new hyperscaler PPAs vs the existing contracted/merchant base?
3. What is the realistic Crane Clean Energy Center restart date given the pending Susquehanna River water-intake permit and NRC final approval — and what's the contingency if it slips past 2027?
4. Walk through the Calpine purchase math: at 7.9x EV/EBITDA for a gas-heavy fleet, what merchant-margin assumptions justify it, and what's the impairment trigger on the ~$11B of goodwill?
5. Pro forma leverage moved from ~0.8x to 2.25x ND/EBITDA. What is your hard ceiling, and at what point do collateral-posting triggers ($3B on a downgrade) constrain the buyback?
6. How fungible is your hyperscaler premium versus Vistra and Talen — what stops a buyer from contracting with whoever is cheapest, and why do you deserve a ~26x multiple vs their ~18x?
7. How much additional fleet capacity will you commit to long-dated PPAs vs keep on merchant exposure, and what's the optimal contracted/merchant mix?
8. What is the customer concentration of your contracted growth book by counterparty, and how is each PPA protected against an early-termination or AI-capex pullback?
9. What does the gas fleet's margin profile look like through-cycle, and how much of FY2026 guidance rests on current (possibly peak) gas spreads?
10. What is your appetite for further nuclear-restart or asset acquisitions (e.g., Palisades if Holtec struggles), and your disciplined-return hurdle?
11. How exposed is the fuel supply chain to enrichment/fabrication concentration (Westinghouse-Brookfield/Cameco), and what's your HALEU/fuel-security plan?
12. What are the realistic economics and timeline of 80-year license extensions and uprates — how many incremental GW at what cost?
13. How do you think about the IRA nuclear PTC post-2032, and what's the plan for the cliff?
14. What capital-return framework balances the $5B buyback, growth capex, and maintaining investment-grade through the higher leverage?
15. Which assumption in your own plan are you least confident in, and what would change your mind on the AI-power demand durability?

---