Financebotresearch desk研究台

CEG › analyze

CEG · Analyze

WATCH Energy

Price at analysis: $282.41 | Mkt cap $100.06B | 2026-08-27

0. Knowledge-base check (done before research)

python .mcp/kb.py find CEG returned one live note: Knowledge/Themes/nuclear-uranium.md (swept 2026-06-10, refined 2026-07-31). Its standing finding: CEG sits in the generation layer of the nuclear fuel-cycle stack — "contracted cash flow, 20yr PPAs, the lowest-risk layer" — as against enrichment (LEU, held) and SMR (SMR, held), which the household already owns and which carry materially higher risk. No prior /analyze report existed for CEG, so this is a from-scratch pass, not a differential one.

Pitfalls checked before trusting any vendor number: pitfall-vendor-forward-eps-is-the-wrong- fiscal-year, pattern-net-margin-above-operating-margin-is-a-tripwire, pitfall-single- quarter-fcf-read-as-ttm, pitfall-yahoo-snapshot-fcf-field-diverges-from-quarter-sum. All four fired or nearly fired on this name — see §1 and Data Traps below.

1. Data-quality gate — three traps fired on this name before any framework work could start

CEG is a merchant generator that closed a transformational acquisition (Calpine, ~$26.6B, closed within Q1 2026) mid-TTM-window, so every vendor "trailing twelve month" field is a Frankenstein of two different companies. Nothing in §2–§4 below uses an un-reconciled vendor aggregate; all are rebuilt from the four quarters Q3 2025 – Q2 2026.

(a) TTM reconstruction, quarter by quarter:

Quarter Revenue Net income (to common) Op. income (as reported) OCF Capex FCF (OCF−capex)
Q3 2025 $6.570B $930M $1,086M $1,848M $390M $1,458M
Q4 2025 $6.074B $432M $598M $805M $986M −$181M
Q1 2026 (Calpine closes) $11.122B $1,590M $2,332M $425M $1,275M −$850M
Q2 2026 $7.504B $513M $580M $1,128M $1,246M −$118M
TTM sum $31.270B $3,465M $4,596M $4,206M $3,897M $309M

Revenue and net income tie exactly to Yahoo's snapshot totalRevenue/netIncomeToCommon — those two vendor fields are trustworthy. Two are not:

(b) The vendor freeCashflow snapshot field is broken — a second confirmed instance of [[pitfall-yahoo-snapshot-fcf-field-diverges-from-quarter-sum]]. Yahoo's get_stock_info reports freeCashflow: -$6.625B. The reconciled quarter-sum above — OCF minus capex, which also ties exactly to each quarter's own pre-computed "Free Cash Flow" cell in quarterly_cashflow — gives +$309M. A $6.9B gap. The likely mechanism here (unconfirmed, same caveat as the DPZ case): the snapshot aggregate appears to be pulling in Calpine acquisition financing/investing cash flows (Q1 2026 alone shows a $2.537B "Purchase Of Business" investing outflow, plus $7.7B of debt issuance against $6.75B of repayment in the same quarter) into what should be an operating-only figure. Never quote CEG's FCF from the snapshot field — use the quarter-sum. This report recommends promoting CEG onto the existing pitfall note as a second, confirming ticker (see Pitfall Notes below).

(c) The tripwire in [[pattern-net-margin-above-operating-margin-is-a-tripwire]] nearly fired, and Yahoo's own operatingMargins field is internally inconsistent with its own statement data. Yahoo's snapshot shows operatingMargins: 8.66% against profitMargins: 11.08% — net above operating, the tripwire's exact shape. Reconciling from the quarterly income statements using the company's own as-reported operating income line (not Yahoo's derived "Operating Income" = revenue − total expenses, which swings negative in Q2 2026 because a large non-operating item gets folded into "Total Expenses" that quarter) gives TTM operating income of $4.596B on $31.27B revenue = 14.7% operating margin — comfortably above the 11.1% net margin, as it should be. The tripwire is explained, not a genuine anomaly: it is a vendor statement-line-selection defect, not evidence of below-the-line inflation. Worth adding as a documented exception/refinement to the pattern note (see Pitfall Notes).

(d) The genuine below-the-line item worth flagging on its own: TTM net income carries roughly $1.28B of "Gain on Sale of Security" (Q3 2025 $19M, Q4 2025 −$48M, Q1 2026 $84M, Q2 2026 $1,226M), most of it landing in a single quarter. This looks like nuclear-decommissioning- trust or hedge-book mark-to-market rather than core generation margin — exactly the "mark-to- market of energy hedges" risk this report was asked to isolate. Core TTM operating earnings power is the $4.596B operating-income figure in (a), not the $3.465B GAAP net income figure, which is flattered by roughly a third of a billion dollars per quarter of below-the-line marks that will not repeat on the same schedule.

(e) The forward-P/E pitfall fires again, textbook shape. Yahoo shows forwardPE: 21.17x on forwardEps: $13.34. But epsCurrentYear: $12.12 / priceEpsCurrentYear: 23.31x is the true current-year read — and it lines up almost exactly with the company's own raised FY2026 guidance of $11.50–$12.50 adjusted operating EPS (midpoint $12.00), confirmed 2026-08-05. The $13.34 "forward" figure is FY2027 consensus, one year out. True current-year P/E is 23.3x, not the 21.2x the vendor snapshot shows — an 8-9% understatement, in the direction that always makes the stock look cheaper than it is.

2. General Health (Fundamentals)

Metric Value Read
TTM revenue $31.27B +23% YoY per vendor, but this is a step change from Calpine consolidation, not organic growth — treat organic growth separately (see below)
TTM operating income (reconciled) $4.596B (14.7% margin) Healthy for the sector; the vendor's 8.66% figure is a statement-line artifact, see §1(c)
TTM net income to common $3.465B (11.1% margin) ~$1.28B is non-operating securities/hedge marks — see §1(d)
TTM interest expense ~$783M Up sharply from FY2025's $511M annual figure — reflects ~$12.7B of Calpine debt assumed at close
FY2025 FCF (annual, pre-Calpine) $1.29B First positive FCF year after three straight negative years (2022 −$4.04B, 2023 −$7.72B, 2024 −$5.03B) — heavy capex on uprates and the Crane/TMI restart
TTM FCF (reconciled quarter-sum, spans the merger) $309M Thin — the Calpine close itself (large one-time investing/financing flows) and elevated capex during integration are compressing free cash flow in the transition window
Debt/Equity (TTM, post-Calpine) 76.4% Materially higher than the FY2025 annual Debt/Assets of 15.7% fin.py shows — that figure predates full Calpine debt consolidation and should be treated as stale
Current ratio / Quick ratio 1.46 / 0.47 Normal for a capital-intensive generator carrying fuel/materials inventory
Shares outstanding 354.31M, up from 312.29M (FY2025) The increase is the ~50M shares issued to Calpine sellers in the acquisition, not organic dilution — frame it as acquisition currency, not float creep
ROE / ROA 15.1% / 3.9% Leverage-assisted ROE; unremarkable ROA for the asset base
Dividend $1.71/yr, yield 0.61%, payout ratio 16% Low yield, very low payout — large headroom to grow the dividend, "boring by design" management framing per recent coverage

Capital allocation (FY2025, pre-Calpine baseline): capex $2.95B (the dominant use of cash), debt issuance/repayment roughly netting to modest deleveraging, buybacks $400M, dividends $486M. Post-Calpine the mix shifts hard toward debt service and integration capex; buybacks continue (company guidance cites "accretive share repurchases" as a contributor to the Q2 beat) but at a reduced pace relative to the balance-sheet load taken on.

Read: the underlying nuclear generation business is healthy and improving (first positive annual FCF in four years, operating margin ~15% once correctly reconciled), but the Calpine acquisition has reset every leverage and cash-flow metric mid-stream. The combined entity's "clean" run-rate FCF and leverage profile will not be visible until Q4 2026/Q1 2027 prints a full four quarters of consolidated Calpine data.

3. Moat & Competitive Advantage

Quantitative base: ROE 15.1%, gross margin 22%, operating margin ~15% (reconciled) — solid but not spectacular relative to software or platform moats; this is an asset-heavy generation business and should be judged against utility/IPP peers, not tech comparables.

The moat itself is close to unimpeachable at the asset level. Constellation is the largest nuclear fleet owner in the United States (now ~31.7GW of generating capacity pre-Calpine, ~55GW combined post-close per company disclosure). New nuclear capacity cannot be built at scale on any timeline that matters to a hyperscaler planning data centers for 2027-2030 — existing licensed reactors are close to a fixed, non-replicable asset. That scarcity is the entire thesis:

  • Direct hyperscaler offtake. The Microsoft 20-year PPA underwriting the Crane Clean Energy Center (Three Mile Island Unit 1) restart is the marquee deal, but management disclosed an additional ~920MW of long-term nuclear PPAs (avg. 18.5-year duration) in Q2 2026 alone, bringing roughly 30% of the clean baseload fleet under contract. Each new PPA converts merchant-price-exposed capacity into contracted, lower-variance cash flow — a real, compounding de-risking of the earnings base.
  • Policy tailwind, with a built-in governor. The Section 45U nuclear production tax credit floors revenue when merchant power prices are low, but the credit phases out as market prices rise — a mechanism that caps the combined upside from PTC-plus-merchant-price but also means CEG rarely gets the worst of both worlds simultaneously.
  • Capacity-market strength. The PJM 2028/2029 base residual auction cleared 18,875MW of Constellation capacity (15,700MW nuclear) at $325/MW-day — a historically elevated clearing price directly reflecting the AI-driven demand growth thesis, and revenue CEG collects regardless of whether the associated MWh are ever delivered to a specific customer.
  • Crane/TMI restart is tracking ahead of the original 2028 schedule, with a 2027 return to service now flagged as plausible by management, and a June 2026 FERC waiver from PJM interconnection rules removing one procedural obstacle. NRC approval remains outstanding — named explicitly as a risk below, not yet cleared.

Adversarial stress-test: "You are a well-funded rival — how do you attack this?" You cannot build a competing nuclear fleet inside the relevant time horizon; the only levers are (1) build enough gas/battery/renewable capacity to erode CEG's pricing power in PJM capacity auctions, (2) out-bid CEG for the next hyperscaler PPA, or (3) wait for policy to turn against nuclear. None of the three is easy, but (1) is exactly what Calpine buys CEG protection against — it converts CEG from a pure-nuclear bet into a diversified nuclear-plus-gas fleet that can also monetize gas-fired flexibility as a firming asset for intermittent renewables and data-center load-following, which is a real, if less differentiated, moat extension.

Disruption forecast (5-10yr): the clearest long-run threat is not a rival generator but policy — a future administration less favorable to nuclear could weaken 45U, though the current administration's posture (a $1B DOE loan specifically for the TMI/Crane restart) points the other way for now. SMR proliferation in the 2030s is a distant, second-order threat to utility-scale nuclear economics, not a near-term one.

Evergreen assessment: high. Baseload, carbon-free, dispatchable power at this scale is a structurally scarce asset independent of any single AI-capex cycle; even a slowdown in hyperscaler capex would leave CEG's existing PPA book and capacity-market revenue largely intact, because those contracts are already signed.

4. Valuation

Applied conditionally per §3 of analysis_notes.md: CEG is a low-yield growth-infrastructure name, not a value/asset play, so Graham is down-weighted to a sanity floor and DYT/DDM carry almost no weight given the 0.61% yield. Bogle and peer/FCF-based approaches carry the load.

Model Output Weight & why
Graham IV — √(22.5 × $10.23 × $90.00) $143.93 vs $282.41 spot Low weight. CEG is not a cheap, asset-backed value name; Graham was designed for exactly that case and reads CEG as ~2x overvalued. Useful only as a floor sanity check, not a target.
Peer multiple (true current-year P/E) CEG 23.3x vs Vistra (VST) ~15.75x fwd, Talen (TLN) ~14.78x fwd, IPP industry median ~15.6x Medium-high weight. CEG trades a ~45-50% premium to the closest "AI-power" comparables. Some premium is earned (largest, purest nuclear fleet; direct Microsoft PPA; now diversified via Calpine), but this is a large gap to defend indefinitely.
Bogle expected return Dividend yield 0.6% + normalized EPS growth (~13-15%/yr on management's coupon-clipping framework, excluding the one-time Calpine step-change) ± multiple normalization If the 23x current-year multiple mean-reverts toward peer levels (~17-18x) over 2-3 years, expected annual return runs roughly flat to high-single-digits — a reasonable but unexciting outcome at today's price. If the multiple holds, expected return is closer to the mid-teens EPS growth rate plus the token dividend.
DYT / DDM N/A-weight 0.61% yield and 16% payout ratio make yield-based models nearly meaningless here; noted only that payout has substantial room to rise.

Fair value range: $255-330, built from (a) a current-year P/E band of ~21-27x applied to the $12.00 guided midpoint EPS (giving $252-324) and (b) a light upward adjustment toward the analyst consensus mean target of $348 (20 analysts, average rating 1.55/"Buy," targets ranging $290-441, several recently trimmed through 2026 even as guidance was raised — a signal that the Street itself is not fully confident the premium multiple holds).

Entry $225-250 — roughly 19-21x the guided current-year EPS, close to the 52-week low ($228.63) and to where CEG's premium to VST/TLN narrows to a more defensible ~25-30% rather than today's ~45-50%.

Trim 30x ttm (implies ~$307 at today's $10.23 TTM EPS, rising toward the mid-$300s as the Calpine-inclusive quarters fully populate the trailing twelve months over the next two quarters). Set on ttm, not fwd, because §1(e) confirms the vendor forward field is on the wrong fiscal year for this name — using it would set the trim on an artificially cheap-looking multiple, the exact failure mode the pitfall describes.

5. Sentiment & Intelligence

  • Earnings momentum is genuinely strong and repeatedly beating. Q2 2026 GAAP EPS $1.42, adjusted EPS $2.55; full-year 2026 adjusted guidance raised to $11.50-12.50 (from $11.00-12.00) on 2026-08-05, driven by Calpine accretion, firmer PJM capacity prices, and commercial optimization gains.
  • Analyst coverage skews bullish but has been cutting targets through 2026 even as guidance rose — Barclays $358→$324, Wells Fargo $384→$362, TD Cowen $381→$368, Citigroup $348→$297, Keybanc a steep $417→$321 — while Scotiabank ($441), UBS ($380), and JP Morgan ($400→ still >$300) remain more constructive. The pattern (raising the earnings number while cutting the price target) is itself informative: analysts are not disputing the fundamentals, they are compressing the multiple they're willing to pay for them.
  • The stock is down ~31.6% from its 52-week high ($412.70 → $282.41) and down 12.5% over the trailing year against an S&P +18.1% — a meaningful de-rating from an earlier "nuclear-for- AI" euphoria peak, consistent with the multiple-compression signal above.
  • Insider activity is quiet and mildly positive. A director bought 1,500 shares on 2026-08-11 at $278.62 (near current levels) — a modest but recent open-market purchase, not a scheduled grant. The bulk of other Form 4 activity is routine director/officer stock-award grants; no notable insider selling pattern surfaced.
  • Recent news flow (last week of Aug 2026) is uniformly constructive on the underlying business — "new power deals piling up," guidance raise coverage, dividend-growth framing — with the recurring market question being whether the stock price already reflects it.

6. Type-Specific Overlay — none of REIT/BDC/dividend-grower frameworks are the primary lens

CEG is not a REIT or BDC. It is a marginal dividend payer (0.61% yield) so the Dividend Grower overlay applies only lightly: the FCF payout ratio is currently distorted by the merger-year FCF compression (§1), but on the pre-merger FY2025 base ($1.29B FCF vs $486M dividends paid) the underlying payout ratio was a conservative ~38% — plenty of room, consistent with the "dividend growth is boring by design" framing in recent coverage.

7. Synthesis & Weighted Verdict

Is this a good business? Yes, clearly. The largest US nuclear fleet is a structurally scarce, policy-favored, increasingly contracted (PPA-backed) asset that is difficult for any rival to replicate on a relevant timeline, now diversified by the Calpine gas fleet into a broader dispatchable-power platform. Fundamentals (§2) confirm real, improving operating profitability once the vendor data traps in §1 are corrected.

Has the market already priced that in? Largely, yes. CEG trades at a true current-year P/E of 23.3x against peer AI-power comparables (VST, TLN) at ~15-16x — a 45-50% premium — and the Street's own behavior (raising EPS estimates while cutting price targets through 2026) reads as a market that believes the earnings but is unwilling to keep paying an expanding multiple for them. Graham's floor test ($144) underscores that there is no value-investor margin of safety here; this is a growth/quality name priced for growth/quality, not a bargain.

Tension worth naming explicitly: Moat analysis argues for a structurally justified premium (scarce, irreplaceable, increasingly contracted assets); Valuation argues the premium has already run ahead of the peer set by a wide margin, and the leverage taken on for Calpine (D/E now ~76%, TTM FCF barely positive through the transition) adds real execution risk that a pure nuclear-scarcity thesis doesn't price in. Weighting: for an infrastructure name mid- integration on a large debt-funded deal, Fundamentals + Valuation should carry more weight than Moat optimism until at least one full post-Calpine quarterly cycle confirms the combined entity's clean cash-flow and leverage trajectory — hence WATCH rather than ACCUMULATE at today's price, with a defined entry band for a pullback rather than a chase.

Verdict: WATCH | conviction 6.0/10. A quality, moat-backed compounder in a structurally favored niche, but priced today at a premium to peers that leaves little room for integration missteps, NRC delay on Crane, or any softening of the AI-capex narrative. Revisit on the Q3 2026 print (first quarter of overlapping Calpine-inclusive data giving a cleaner leverage/FCF read), on the NRC's Crane/TMI restart decision, or on a pullback into the $225-250 entry band.

Key Risks

  • Valuation premium to the "nuclear-for-AI" peer set (~45-50% over VST/TLN) is the single largest risk — it requires CEG to keep out-executing peers indefinitely, not just to execute well.
  • Calpine integration and leverage. D/E jumped to ~76% (TTM), ~$12.7B of assumed debt, reconciled TTM FCF of only $309M through the transition window — a name that must show de-leveraging and FCF recovery over the next few quarters to justify the premium multiple.
  • Regulatory/counterparty risk on Crane (TMI) restart — NRC approval is still outstanding despite the FERC/PJM waiver already secured; PJM's own Market Monitor has publicly opposed aspects of the waiver process, a live friction point.
  • Earnings-quality noise — roughly $1.28B of TTM net income came from below-the-line securities/hedge gains concentrated in one quarter (Q2 2026); GAAP EPS should not be taken at face value without separating operating from mark-to-market results, exactly as this report's guard anticipated.

Data Traps Hit

  1. [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]] — fired again. Vendor forwardPE 21.17x (on FY2027 consensus $13.34 EPS) understates the true current-year multiple of 23.3x (on priceEpsCurrentYear/guided FY2026 EPS ~$12.00). Add CEG to the ticker list.
  2. [[pitfall-yahoo-snapshot-fcf-field-diverges-from-quarter-sum]] — a second confirming instance (after DPZ). Vendor snapshot freeCashflow −$6.625B vs reconciled quarter-sum +$309M, a $6.9B gap, on a name mid-acquisition. Raises confidence toward the "vendor aggregate absorbs investing/financing cash flows around an M&A close" hypothesis. Add CEG.
  3. [[pattern-net-margin-above-operating-margin-is-a-tripwire]] — tripwire nearly fired (Yahoo shows operatingMargins 8.66% < netMargin 11.08%) but resolved as a new, seventh cause worth naming: Yahoo's derived "Operating Income" field (revenue minus total expenses) goes negative in a quarter where a large non-operating gain gets folded into "Total Expenses," while the company's own as-reported operating-income line stays solidly positive and above net income across all four TTM quarters. This is a vendor statement-line-selection defect on the operating-margin field itself, not a below-the-line contamination of net income — worth documenting as distinct from causes 1-6 already in the note, because the standard remedy ("find what's below the operating line") does not apply; the fix is to use the company's as-reported operating income line, not Yahoo's derived one.

Pitfall Notes for Orchestrator

Amend pitfall-yahoo-snapshot-fcf-field-diverges-from-quarter-sum: add CEG as a second ticker. CEG snapshot freeCashflow = −$6,625,124,864; reconciled TTM (OCF minus capex, summed Q3 2025–Q2 2026, cross-checked against each quarter's own pre-computed FCF cell) = +$309M. CEG is mid-acquisition (Calpine, closed Q1 2026, ~$2.537B "Purchase of Business" outflow plus large debt issuance/repayment that quarter) — worth checking whether the vendor aggregate is pulling in investing/financing cash flows around the M&A close specifically, which would explain both this case and DPZ's smaller-magnitude gap under a common mechanism (or ruling that out if DPZ had no M&A event in its window). Upgrade confidence from medium toward high if the M&A mechanism is confirmed, or keep at medium and broaden scope to "any name with a same-window M&A close or major debt refinancing" if not.

Amend pitfall-vendor-forward-eps-is-the-wrong-fiscal-year: add CEG. forwardPE 21.17x on FY2027 forwardEps $13.34; true current-year (FY2026) read is priceEpsCurrentYear 23.31x on epsCurrentYear $12.12, which lines up almost exactly with the company's own raised guidance ($11.50-12.50, midpoint $12.00, confirmed 2026-08-05). Another case where the current-year field and the company's own guidance agree to the cent while forwardPE is silently one fiscal year out.

New refinement candidate for pattern-net-margin-above-operating-margin-is-a-tripwire: a potential cause 7 — the tripwire can fire (or nearly fire) purely from a vendor statement-line-selection defect on the operating-margin field, independent of any real below-the-line contamination of net income. On CEG, Yahoo's derived "Operating Income" (revenue − total expenses) swung negative in one TTM quarter because a large one-time gain got folded into "Total Expenses" that period, dragging the trailing operatingMargins field below the true as-reported operating-income margin. Reconciling from the company's own as-reported operating-income line (available in the same quarterly_income_stmt response, field "Total Operating Income As Reported") resolved it cleanly — TTM operating margin 14.7% vs net margin 11.1%, the normal relationship. Worth adding as an explicit exception so the next analyst who sees this tripwire nearly-fire checks which operating-income field the vendor derived margin used before hunting for a below-the-line cause that isn't there.

Sources