Constellation Energy

Refresh delta (since 2026-06-17)

What changed

1. Guidance up. FY26 adjusted operating EPS guidance was raised to $11.50–12.50 from $11.00–12.00 at the Q2 print [8-K 2026-08-06]. H1 delivered $5.29 (Q1 $2.74 + Q2 $2.55), up 31% on H1-25's $4.05. Q2 adjusted operating earnings were $920M against $599M a year ago. Management credited Calpine, higher PJM capacity prices and retail margins [8-K 2026-08-06; call Q2-26].
2. The co-location binary broke CEG's way. The June dossier named FERC's PJM co-location framework as the risk bulls were underweighting. In June 2026 FERC ruled on the paper hearing and "pressed PJM to implement changes to accommodate co-located load prior to PJM's proposed June 2029 effective date" [10-Q Q2-26, MD&A]. Management now expects "rule clarity in the first to second quarter of '27" [call Q2-26]. CEG is still arguing the details: it protested PJM's Interim Resource Adequacy Service and large-load registry proposals in September.
3. Capacity revenue locked at the cap through May 2029. Every CEG plant in PJM cleared the 2028/29 auction at $325/MW-day, the price cap, on 18,875 MW of ownership volume [8-K 2026-07-14]. That is about $2.2B a year of capacity revenue for that delivery year. PJM's own uncapped simulation cleared at ~$555/MW-day, so the cap is now holding CEG's price down, not up.
4. Crane is on schedule. The NRC approved Crane's new-fuel licence amendment and FERC granted the Eddystone capacity-rights transfer [8-K 2026-08-06]. Management kept the second half of 2027 as the restart target [call Q2-26]. The $1.0B DOE-guaranteed loan is still undrawn [10-Q Q2-26, Note 13].
5. More contracts, and the first number on how much is contracted. CEG signed 920 MW of new 15–20-year nuclear PPAs with investment-grade buyers, starting 2029–2032. They include 176 MW with Walmart from Dresden, 30 MW of which is a new uprate [8-K 2026-08-06]. The CEO said "roughly 30% of our clean baseload output" is now under long-term contract [call Q2-26]. No new hyperscaler name was announced in the window.
6. The stock went nowhere. CEG closed at $267.17 on 2026-06-17 and $261.62 on 2026-09-24 (−2%). In between it closed at $236.50 on 2026-07-01. That was the day after the Calpine lock-up released half of the 50M acquisition shares and PJM declared a grid emergency, and the day Citi cut its target to $297 from $348. Market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. is ~$92.7B.

June dossier corrections

• Q1 revenue did not "double on Calpine". Of Q1's $11.1B, about $1.3B was unrealized hedge gains in the unallocated "Other" line. H1 unrealized revenue gains were $995M after a $320M Q2 loss [10-Q Q2-26, Note 5]. Calpine added $3.1B of Q1 revenue [10-Q Q1-26, Note 2]. On a pro forma basis Q1 revenue rose 22% ($11.35B vs $9.32B), not 64%.
• The nuclear PTC "floor" is not in play at today's prices. CEG "did not record a material nuclear PTC benefit" in H1 because gross receipts were above the phase-out for most units [10-Q Q2-26, Note 6]. The PTC protects the downside, and management now puts the 2030 strike at $50.88/MWh (was $49.88), worth ~$0.30 of 2030 base EPS [call Q2-26].
• Goodwill and downgrade collateral are larger than June's figures. Goodwill is $11.5B, not $11.1B, and collateral due on a downgrade below investment grade is $3.44B, not ~$2.97B [10-Q Q2-26, Note 12].

What held. The structural thesis stands. The nuclear fleet ran at a 93.0% capacity factor in Q2 (94.8% a year ago, on more refuelling days) [8-K 2026-08-06]. Nobody can build a new US reactor this decade, and the buyer set has widened from hyperscalers to Walmart and CyrusOne. The supply chain is unchanged.

Phase A — Understand the business

Lens 1 · Overview: carried from the June dossier, with these filing-sourced updates:

• Fleet: "55 gigawatts of capacity from nuclear, natural gas, oil, geothermal, hydro, wind and solar" and "approximately 2.5 million customer accounts" [8-K 2026-08-05 Ex. 99.1]. After the required sales, Calpine brings ~23 GW and a retail book of ~62 TWh a year [10-Q Q2-26, Note 2].
• Calpine price: $21.8B of consideration, made up of 50M CEG shares valued at $17.6B and $4.3B of cash. CEG also assumed $12.6B of Calpine debt [10-Q Q2-26, Note 2].
• The divestitures are nearly done. Five PJM gas plants (4.4 GW) go to LS Power for $5.0B, and Brazos Valley (606 MW, ERCOT) for $860M. Both are expected to close by year-end [10-Q Q2-26, Note 2]. Management puts total gross proceeds at ~$5.9B, "nearly $1,200 per kW" against an implied ~$960/kW paid for Calpine [call Q2-26].
• New growth assets: Pin Oak Creek (460 MW ERCOT gas peaker) reached commercial operation on 2026-04-30. The 105 MW Pastoria solar project came online on 2026-04-16 [8-K 2026-05-11]. At Freestone (Texas), a 380 MW CyrusOne data-center agreement is signed and a 380 MW Phase 2 is exclusive. That is on top of the 400 MW at Thad Hill [10-Q Q2-26, MD&A].
• Licence extensions: Ginna and Nine Mile Point 1 have filed to run to 2049 [8-K 2026-08-06]. Clinton and Dresden renewals were approved earlier [8-K 2026-02-24].

Lens 2 · Supply chain: carried, unchanged since the June dossier.

Lens 3 · Moats: carried. One addition: the Brazos Valley sale shows that even the gas fleet the June dossier called "not moated" commands a premium. Management said private equity paid "over $1,400 a kW for Texas assets in a soft ERCOT market" [call Q2-26]. The moat is still the nuclear, but the gas assets were not bought at the top as June's bear case feared.

Segments (re-run: Calpine is now its own segment)

CEG reports six segments. Five are regions, and Calpine is reviewed standalone. The segment metric is RNF: revenue net of purchased power and fuel [10-Q Q2-26, Note 5].

[10-Q Q2-26, Note 5]

Read: Calpine is ~30% of segment revenue and ~26% of segment RNF. It earns a lower margin per revenue dollar (48%) than the regional nuclear book (57% blended), as June predicted. Mid-Atlantic RNF is up 26% on PJM capacity prices: the EMAAC reference price was $289.67/MW-day against $125.71 a year ago [8-K 2026-08-06, Statistics]. Midwest is down 5% on outages and ComEd energy prices ($29.32 vs $31.09/MWh). The unallocated line swung RNF by −$0.78B year on year, mostly hedge mark-to-market. That is a timing item that does not show up in adjusted EPS.

Phase B — Measure performance

Earnings (Q2-26, reported 2026-08-06)

• GAAP versus adjusted. The $1.13/share gap in Q2 runs the other way from Q1. Unrealized hedge losses (+$0.94), Calpine contract amortisation (+$0.41) and merger costs (+$0.23) add back. Decommissioning-trust gains (−$0.61) come out [8-K 2026-08-06, reconciliation]. The Q2 effective tax rate was 44.2%.
• Versus consensus: consensus FY26 EPS is ~$11.62 against the new $12.00 midpoint, so the Street sits below the raised guide.
• Balance sheet at 2026-06-30 [10-Q Q2-26]:
• Total assets were $98.3B (Dec-25: $57.2B).
• Goodwill was $11.5B, with its allocation to reporting units still provisional.
• Assets held for sale were $5.7B.
• Debt totalled $24.7B: short-term borrowings of $5.2B, current long-term debt of $0.4B and long-term debt of $19.1B.
• Cash was $0.7B.
• Gross debt was $9.0B at Dec-25, so it rose by $15.7B in six months. Of that, $12.6B is assumed Calpine debt, part of which was refinanced.
• Capital returned: ~7.1M shares were repurchased in Q2 for ~$2.0B, including 2M bought out of the June secondary offering. Another 1M shares were bought in July for ~$250M. Remaining authority is ~$2.8B [10-Q Q2-26, Note 16]. Shares outstanding went from 362M at March 31 to 355M at June 30 [10-Q Q1-26; 10-Q Q2-26].

Earnings call (Q2-26)

Tone: confident, and more political than in June. The CEO spent part of his prepared remarks on the case for data centers in general. He compared public opposition to data centers with the opposition to restarting Three Mile Island: "if we can restart Crane at Three Mile Island and earn overwhelming political and public support, then we can certainly earn the public support to build a 21st century data economy" [call Q2-26].

• Deal pipeline: "our transactional pipeline for future deals is both robust and active" [call Q2-26]. As in June, there was no number for the pipeline and no new hyperscaler name.
• Co-location: "we think that we're going to see rule clarity in the first to second quarter of '27 in terms of co-location, far ahead of where PJM might have been before where they were targeting 2029" [call Q2-26]. The CEO is now pitching "hybrid" deals that combine batteries, existing generation and new capacity to serve peak-constrained load.
• Capital allocation: the 2029 sensitivity to capital deployment now has a $0.20/share floor, reflecting buybacks already done, with ">$0.75" of upside [call Q2-26].
• ERCOT: analysts pushed on soft Texas prices. Management called them "completely expected" and weather-driven, and said they expect the market to tighten as data centers connect [call Q2-26].
• What changed from June: the "is AI demand real?" defence is gone. The message is now about regulatory plumbing (backstop auctions, interim transmission services, EPA backup-generator limits) and buybacks.

Comps

Read: the premium to Vistra is ~7 turns, against ~8 in June. The gap narrowed because CEG's earnings rose while its price didn't. Talen's forward multiple could not be sourced.

Stock-price catalysts since 2026-06-17

• Jun 23: the Walmart PPA was announced. There was no large move.
• Jun 30 – Jul 1: −11.5% from the June 17 close to $236.50 (intraday low $228.63). Three things landed together: half of the 50M Calpine acquisition shares came off lock-up on June 30, PJM declared a reliability emergency the same day, and Citi cut its target to $297 on July 1 [10-Q Q2-26, Note 2; web: TheStreet, 2026-07-01]. The rest of the lock-up (~25M shares) releases on 2027-06-30 [10-Q Q2-26, Note 2].
• Jul 14: the PJM 2028/29 auction cleared at the cap [8-K 2026-07-14].
• Aug 6–7: Q2 beat and raise. The stock fell 1.5% on the day ($261.10) and rose 3.4% the next ($269.89). New director Roger Crandall bought 1,500 shares at $278.62 [Form 4, filed 2026-08-13].
• Since mid-August: the stock drifted back to ~$262. One officer, Andrew Novotny, filed Rule 144 notices to sell ~29.7k shares worth ~$8.0M (August 21 and September 17) [Form 144s on EDGAR]. That is small against 354M shares outstanding.

Pattern: since June the stock has moved on supply (the lock-up), grid-politics headlines and sector targets, not on its own earnings. That fits the June read of CEG as a narrative stock, except the narrative has moved from "AI demand" to "PJM rules and balance sheet".

Phase C — Judge people & books

Management (re-run: governance change)

• Joe Dominguez became Chair as well as CEO on 2026-08-04, when Robert Lawless retired from the board. Charles Harrington became Lead Independent Director with an extra $50k retainer. Roger Crandall (CEO of MassMutual) joined the board on 2026-08-05 [8-K 2026-08-05]. Combining the two roles concentrates power. The mitigant is a named lead independent director. Crandall's open-market purchase is the only insider buy in the window.
• Capital allocation: since the March outlook CEG has spent ~$2.2B on buybacks, at $282 average in Q2 [call Q2-26; 10-Q Q2-26, Note 16], while gross debt stands at $24.7B. The buyback is being run alongside the leverage step-up, not after it. The Brazos Valley and PJM sale proceeds ($5.9B) are the bridge. See Lens 10.

Forensic red flags (re-run against the 10-Q)

• Cash conversion. H1 operating cash flow was $1.55B, flat on H1-25 ($1.58B). Over the same period adjusted earnings rose ~31% per share and revenue grew 44% [8-K 2026-08-06, cash flow]. The drags were a −$1.28B cut in payables, −$0.36B of collateral posted and −$0.81B of other working capital. H1 capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. was $2.52B, so free cash flowFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. was −$0.97B before the $2.5B Calpine cash payment and the $2.0B of buybacks. CEG funded those with short-term debt: net short-term borrowing and loans added +$3.6B. This is the watch item that replaces the FERC binary.
• Downgrade collateral. $3.44B would be due at BB+/Ba1 (Dec-25: $2.67B). Cash collateral posted is $2.46B (Dec-25: $1.40B). CEG is rated BBB+/Baa1, three notches above junk [10-Q Q2-26, Note 12 and Liquidity]. Liquidity is $7.0B of undrawn facilities plus $0.7B of cash, so a downgrade is survivable but not free.
• Goodwill of $11.5B is not yet allocated to reporting units. The allocation is still provisional [10-Q Q2-26, Note 2]. No impairment test is meaningful until that is done. Check again at the 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes..
• The GAAP-to-adjusted gap swings both ways. Hedge mark-to-market flattered Q1 GAAP and depressed Q2 GAAP. Revenue has become a poor guide to activity (Lens 4). Use RNF and adjusted operating earnings.
• Litigation: accruals were $28M current plus $122M non-current, up from $15M plus $113M [10-Q Q2-26, Note 15]. The Winter Storm Uri tort claims against generators are being dismissed: the Texas Supreme Court denied rehearing in July 2026. This clears a legacy CEG and Calpine exposure [10-Q Q2-26, Note 15].
• Regulatory findings: EDGAR full-text search, re-run 2026-09-25, found no SEC Litigation Releases or AAERs over 2021-09-25 → 2026-09-25. No non-SEC enforcement was found in the web pass. The live regulatory exposure is still market design (PJM, FERC, ERCOT), not misconduct.

Phase D — Project & stress-test

Forward projection

Anchor: the new FY26 guide is $11.50–12.50 [8-K 2026-08-06]. Consensus is ~$11.62 for FY26 and ~$13.59 for FY27. No Brier forecast is logged because this was an unattended run.

Logic:

• 2027: the full PJM 2026/27 capacity year at the cap is ~+13%. Buyback accretion from ~8M shares retired since March, against ~360M, adds ~+2%.
• 2028: Crane runs for the full year (Morgan Stanley's 2024 estimate is ~$1.70 EPS at full output, phased at half here for ramp risk). Capacity stays flat at $325. Contracts starting 2029–2032 do not help yet.
• Bear case: ERCOT stays soft, the PJM cap is politically extended or tightened, and Crane slips into 2028.
• Bull case: a hyperscaler co-location deal lands after rule clarity in H1-27, and buybacks continue at the ">$0.75" end.

Valuation: at $261.62 the stock is ~21.8x the base FY26, ~19.2x FY27 and ~17.0x FY28. In June it was ~23x the same FY26 base ($267.17 on $11.50) and ~26x forward on web data. The multiple came down while the estimates went up.

Bull vs bear

Bull, updated. The one dated regulatory risk broke CEG's way. FERC pushed PJM to speed up co-location services, and management expects rule clarity by mid-2027. Capacity revenue is locked at the $325 cap through May 2029, a price below what PJM's own model says the market would clear at. Crane is on schedule with fuel and grid rights in hand. The guide went up, and the stock is ~19x next year against ~26x in June. The contrarian point: the market treated the June–July drawdown as a verdict on AI power. It was driven by share supply (the lock-up) and a sell-side target cut, not by fundamentals.

Bear, updated. Three risks replace June's.

1. Balance sheet and cash conversion. Gross debt is $24.7B, H1 free cash flow was negative, buybacks were funded with short-term debt, and $3.4B of collateral is due on a downgrade. If the $5.9B of divestitures slip past year-end, or power prices fall, the buyback stops before the leverage comes down.
2. PJM politics. The $325 cap exists because governors and the White House's energy council pushed for it. A reserve shortfall of ~6.8 GW means more intervention is coming, possibly a "reliability backstop" procurement that adds supply outside the market. The cap limits the upside, and the backstop could cut prices after 2029.
3. Stock overhang. ~25M Calpine-holder shares unlock on 2027-06-30, and the July precedent shows the stock can fall ~10% on a supply event.

Pre-mortem, 18 months out: co-location rules land in mid-2027 with transmission charges heavy enough that hyperscalers choose front-of-meter PPAs at lower premiums. ERCOT stays oversupplied with batteries, and Crane slips to 2028. Buybacks pause to protect the rating. EPS still grows ~8–10%, but the stock trades at ~15x.

Devil's advocate

• June's short thesis is dead, and a new one has replaced it. FERC did not take the co-location premium away. The live short is now "a levered merchant generator buying back stock with commercial paper at the top of a capped capacity cycle."
• The "30% contracted" disclosure cuts both ways. Seventy percent of clean baseload output is still merchant or short-dated. After two years of AI-power deal flow, most of the fleet still sells into markets whose capacity price is politically capped.
• The most dangerous competitor is still Vistra. It runs the same playbook at ~14x against CEG's ~21x FY26. That the gap narrowed is a sign the market treats them as fungible.
• Assumptions that must hold at today's price: (1) the divestitures close by year-end at ~$5.9B; (2) co-location rules arrive in H1-27 without prohibitive charges; (3) Crane restarts in H2-27; (4) the rating holds at BBB+/Baa1. Break (1) or (4) and the buyback, and with it the EPS bridge, stalls.

Lens 14 · Management questions: carried from the June dossier, with two added at the top:

• What is the leverage target and timeline once the ~$5.9B of divestiture proceeds arrive, and will buybacks pause if they slip past year-end?
• Why was H1 operating cash flow flat year on year when revenue grew 44%, and how much of the $1.28B payables reduction reverses in H2?