The Contracting Pause Ended and PJM Capacity Cleared at Its Cap in the Same Quarter: Upgrading to Outperform

Results vs. Consensus

Q2 2026 Scorecard

Two independent pre-print reads put adjusted operating EPS consensus at $2.29 and $2.36. We use $2.36, the more widely circulated figure, which makes the beat $0.19 or 8.1%. On the lower read the beat is $0.26 or 11.4%. The conclusion does not change: this was a clear beat on the metric management is judged by.

Year-Over-Year Comparison

The single most striking feature of this table is that the GAAP and the adjusted lines point in opposite directions. Adjusted operating earnings rose 53.6% and adjusted operating EPS rose 33.5%. Reported operating income fell 39.0% and reported earnings per share fell 46.8%. The bridge between them is $1.13 per share wide this quarter and runs the opposite way from a year ago, when adjusted earnings sat $0.76 below GAAP. Four items carry it: a $340 million after-tax unrealized loss on fair-value adjustments added back, a $221 million decommissioning-related credit subtracted, $149 million of acquired-commodity-contract amortization added back, and $84 million of Calpine merger and integration costs added back. None of the four is unreasonable as an exclusion. Together they mean that a reader who follows only the reported statements would conclude the quarter was poor.

Half-Year Comparison

Half-year adjusted operating EPS of $5.29 is the sum of the two reported quarters, $2.74 and $2.55, against $2.14 and $1.91 a year ago. That $5.29 is 44.1% of the new $12.00 guidance midpoint, against a 43.1% first-half share of the full year in 2025, so the year is running modestly ahead of its own seasonal pace.

Sequential Comparison

The sequential comparison is seasonally distorted and we present it only because a reader will otherwise build it. Q1 is Constellation's strongest weather-driven demand quarter and carries the January storm; Q2 is the heaviest refueling quarter and carried 86 planned outage days against 99 in Q1. No run-rate inference should be drawn.

Quality of the Beat

Revenue. Inorganic, and negative underneath. Reported revenue rose $1,403 million. Calpine contributed $1,907 million of operating revenues in the quarter, which is more than the entire increase. The legacy book therefore shrank on a gross basis, and the pro forma table confirms it: combined revenue would have been $8,479 million a year ago against $7,504 million now. The decline is concentrated in Other Power Regions, where wholesale and retail load revenue fell $205 million on lower New England volumes, and in the unallocated line, where $205 million of acquired-commodity-contract amortization runs through revenue with no prior-year comparative. Neither is an operating deterioration in the fleet. Both mean the top line tells you nothing.

Margins. The revenue net of fuel margin fell 2.3 points to 46.4%, which is mix rather than pricing: the acquired retail and wholesale book carries a lower gross margin than the legacy nuclear regions. Within the legacy fleet the direction is the opposite and it is the good news of the quarter. Mid-Atlantic's revenue net of fuel margin expanded 9.5 points to 63.5% and ERCOT's expanded 5.3 points to 63.7%. Midwest went the other way, down 5.6 points to 62.4%, on the Illinois zero-emission credit recognition timing. Operating and maintenance rose 39.3%, or $636 million, against a fleet that roughly doubled in capacity and ran 45 more planned refueling outage days. Management named performance-related compensation as an offset to the guidance raise without quantifying it, which is the one cost line worth watching next quarter.

EPS. Adjusted operating earnings grew 53.6% while adjusted operating EPS grew 33.5%. The 20-point gap is the 46 million additional diluted shares, up 14.6% year over year, of which 50 million were issued to Calpine's owners in January and roughly 8 million have since been repurchased. The effective tax rate rose 9.6 points to 44.2%, which the 10-Q attributes to higher qualified nuclear decommissioning trust income taxed at a higher rate. Because that income is itself removed in the non-GAAP bridge, the GAAP tax rate is a poor guide to the adjusted result, and the company does not disclose a separate adjusted rate.

Cash. Half-year operating cash flow of $1,553 million was flat against $1,584 million a year ago while adjusted earnings rose 30.6%, and it sits against $2,521 million of capital expenditure. The reconciliation is disclosed and it is nearly exhausted: the company refunded or offset approximately $920 million during the half against state-program compensation relating to nuclear production tax credits recorded in 2024, and the associated payable has now fallen to roughly $270 million from $1,190 million at year-end. That is a genuine cash drag that will not repeat at anything like this magnitude. What it does not explain is that the buyback was funded partly with short-term borrowings, which rose to $5,226 million from $1,650 million at year-end.

Segment Performance

Constellation reports six segments, and the metric its chief operating decision maker uses is revenues net of fuel, defined as operating revenues less purchased power and fuel expense. That table appears only in the Form 10-Q. It is where the quarter is actually decided, and it is where last quarter's central criticism of this business is tested.

Revenues Net of Fuel by Segment

Segment Revenue and Revenue Net of Fuel Margin

Mid-Atlantic

The quarter's best segment and the direct answer to last quarter's bear case. Revenue rose 7.4% and revenues net of fuel rose 26.3%, so purchased power and fuel expense actually fell to $567 million from $666 million while the region sold more. Nuclear generation rose 3.4% to 12,676 gigawatt-hours and purchased power volumes fell 18.3%, meaning the region self-generated a larger share of a higher-priced load. PJM Eastern Mid-Atlantic capacity reference prices rose 130.4% to $289.67 per MW-day and PJM West energy prices rose 21.1%. The segment's disclosed revenue driver is favorable retail load revenue of $160 million on higher contracted energy prices, partly offset by $90 million of unfavorable realized economic hedges.

This is what capture looks like. It also happened against a headwind: New Jersey's zero-emission credit program concluded in May 2025, so the year-ago quarter still carried two months of a $10.00 per MWh credit that is now zero. Mid-Atlantic grew 26.3% while absorbing that.

Assessment: The region with the most direct exposure to a PJM data-center contract and to the Crane restart is now also the region converting merchant strength into segment earnings. The hedge roll we said would take two quarters took one. The offset, visible in the driver detail, is that realized hedges are still a $90 million drag, which is the mechanical price of having been protected on the way up.

Midwest

The only legacy segment to decline, and the decline is entirely an accounting recognition artifact rather than an economic one. Revenue rose 2.9% while revenues net of fuel fell 5.5% and the margin compressed 5.6 points to 62.4%. The named drivers are favorable retail load revenue of $110 million and favorable net capacity revenue of $65 million, offset by $135 million of unfavorable zero-emission credit program revenue. The mechanism is specific: Constellation recognizes banked Illinois zero-emission credits from prior planning years in the second quarter once it knows the current year will not breach the annual cap. It recognized $85 million this quarter against $201 million a year ago. The 2026/2027 planning year is the final year of the program, at a $1.02 per ZEC price against a $228 million cap. Illinois zero-emission credit reference prices fell 83.1% to $1.12 per MWh.

"As we've discussed on prior second quarter calls, Illinois ZEC revenue timing can vary year-to-year. This quarter, we recognized $85 million of bank credits compared with $200 million last year. This timing item was already reflected in our 2026 guidance and has no impact to full year results."
— Shane Smith, EVP and Chief Financial Officer

Adjust for that $116 million and Midwest revenues net of fuel grew about 5.7%, in line with Mid-Atlantic's underlying trajectory. The less comfortable Midwest disclosure sits elsewhere in the filing. The Illinois carbon mitigation credit price is set by subtracting energy and capacity index prices from a bid price of $34.50 per MWh, and when the result is negative Constellation pays ComEd. The average realized carbon mitigation credit price was negative $5.91 per MWh this quarter against negative $0.42 a year ago, and negative $16.05 for the half against negative $1.23.

Assessment: Two things are true at once. The reported Midwest decline is noise that reverses on a full-year view, exactly as management said. And the carbon mitigation credit is now a live, growing clawback that scales with the same capacity and energy prices the bull case depends on, in the segment holding 9,800 megawatts of the nuclear fleet's PJM capacity. When the Illinois zero-emission credit program ends in May 2027, Clinton and Quad Cities revert to merchant pricing plus the production tax credit floor. Nobody asked how that bridge is modelled.

New York

Revenue rose 5.4% with no individually significant driver disclosed, while revenues net of fuel fell 1.8% and the margin compressed 5.1 points to 69.1%. Nuclear generation fell 4.5% to 6,336 gigawatt-hours. New York zero-emission credit reference prices were flat year over year at $14.76 per MWh, and NYISO Rest of State capacity prices rose 47.2%. The segment's forward news was regulatory rather than financial: subsequent license renewal applications were filed for both Ginna and Nine Mile Point Unit 1, which if approved extend both units to 2049.

Assessment: Still the most contracted and least optionality-bearing region, and still best modelled as a regulated asset. The license extensions are the substantive development: two units carried to 2049 converts a decade of terminal-value uncertainty into a disclosed operating life, and it happened because a state chose to extend its credit program rather than because a market cleared.

ERCOT

Revenue fell 3.9% while revenues net of fuel rose 4.8% and the margin expanded 5.3 points to 63.7%, so the decline is volumetric and the economics improved. Nuclear generation fell 19.0% to 2,036 gigawatt-hours on outage timing, while gas, oil and renewables generation rose 16.7% to 3,742 gigawatt-hours. ERCOT North energy reference prices fell 9.7% and Houston fell 11.7%. Management was direct that the soft ERCOT tape was anticipated and hedged.

Assessment: Unchanged from last quarter. The legacy nuclear economics in this region are small and shrinking; the ERCOT story for this name is the acquired gas fleet, the co-location template, and the price a buyer will pay for the assets. On the last of those the quarter produced a hard number, and it was better than the market's: $1,420 per kilowatt for Brazos Valley in a soft market.

Other Power Regions

Revenue fell 18.2% while revenues net of fuel rose 58.6% and the margin nearly doubled to 29.8% from 15.4%. The driver detail is unusually informative: revenue fell $214 million on lower New England load volumes, while purchased power and fuel fell more, helped by $150 million on lower load volumes in New England and the West, $120 million associated with environmental product sales, and $55 million of favorable realized economic hedges. Total supply and sales in the region fell 16.6%.

Assessment: A $106 million contribution from shrinking a low-margin book. That is good portfolio management and it is 8.2% of consolidated revenues net of fuel. It does not change the thesis in either direction, but it is the second consecutive quarter in which this segment has out-delivered its size, and it belongs to the same commercial function management is crediting for the guidance raise.

Calpine

The second full quarter of ownership produced segment revenue of $2,147 million and revenues net of fuel of $1,028 million, a 47.9% margin that again sits above every legacy region on a gross-revenue basis except the two contracted nuclear regions. Calpine generated 24,914 gigawatt-hours of gas, oil and renewable output and purchased a further 2,728 gigawatt-hours. Total operating revenues attributable to Calpine across all line items were $1,907 million for the quarter and $5,043 million for the half.

And for the second consecutive quarter the 10-Q states that "it is impracticable to determine Calpine's earnings since the Acquisition Date," now attributing the impracticability to the commencement of integration activities and the consolidation of financing.

Assessment: The margin evidence keeps validating the transaction logic, and the divestiture proceeds now validate the price: management put the DOJ-mandated sales at approximately $5.9 billion of gross proceeds at nearly $1,200 per kilowatt against a $960 per kilowatt implied purchase price for the Calpine assets. That is a real, third-party mark on roughly a fifth of what was bought, and it came in above cost. What has not improved is measurability. The roughly $2 per share of full-year accretion embedded in guidance still has no quarterly marker, two quarters in, and the reason given has shifted from purchase accounting to integration. The longer that persists, the more the accretion figure functions as an assertion rather than a disclosure.

Operating Statistics

The 1.8-point capacity-factor decline is entirely planned: 86 refueling outage days against 41, with six refuelings completed and a turbine upgrade at Byron Unit 1 inside one of them. Non-refueling outage days actually improved. The number that matters more for the thesis is not in this table, because it is a single day rather than a quarter.

"During the mid-Atlantic heat wave ahead of the 4th of July holiday, our employees delivered, achieving a nuclear capacity factor above 99%, while safely managing our dispatchable generation fleet through challenging operating conditions as the new organization came together effectively."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

The one metric moving the wrong way is the acquired thermal fleet's equivalent forced outage factor, at 6.2% against 4.5% in the first quarter of ownership. Summer is when a gas fleet is dispatched hardest and forced outages are most expensive, so a single-quarter step is not a trend. It is worth a third data point before assuming the acquired fleet runs at legacy Constellation availability.

Market Reference Prices

These are the prices the company publishes as context for its own results. Read against the segment table above, the argument of last quarter's report has partly resolved itself and the argument of this one is in the capacity row.

PJM capacity at $289.67 per MW-day is now within 12% of the $325 administrative cap, and as the next section shows, the 2028/2029 auction has already closed that gap entirely.

Key Topics & Management Commentary

Overall Management Tone: The most confident call of the four we have heard from this company, and the shift is specific rather than atmospheric. Where the prior call deferred to regulatory timelines the company does not control, this one opened by claiming those timelines had been beaten and closed by listing what had already been signed and approved. Prepared remarks ran long on values, community and awards before reaching the numbers, which is the posture of a management team that believes the numbers will speak for themselves. Q&A drew no pushback of any kind, and the least convincing moments were the two flat refusals to price the new contracts and to characterize where within a $20 to $50 per MWh disclosed range they landed.

1. The Contracting Pause Ended, and the Price of Ending It Is Undisclosed

The single most important disclosure of the last call was a negative one: management confirmed that some hyperscaler counterparties had stopped advancing negotiations pending PJM's rules. This quarter the same management reported deals.

"Since the last call, we have signed approximately 920 megawatts of long-term nuclear deals that are consistent with our view of long-term value. These contracts have an average duration of 18.5 years and are with investment-grade customers."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

The deck adds the structure: 890 megawatts of existing generation plus a 30 megawatt uprate at Dresden enabled by a 176 megawatt agreement with Walmart announced on June 23. The contracts begin between 2029 and 2031 and are fully ramped by 2032, and they take contracted clean baseload output to roughly 30% by 2032. What management would not supply is the number that determines what any of it is worth.

"While I can't disclose pricing, what I can say is that the deals recognize the value of existing clean and reliable nuclear energy as a premium product."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: This is the event we named in May as the upgrade trigger, and it happened. Two qualifications keep it from being a larger event than it is. The 890 megawatts of existing generation is roughly 6% of the 15,700 megawatts of nuclear capacity Constellation cleared in PJM's most recent auction, so this is a first tranche rather than a repricing of the asset base. And the counterparty mix is broader than the narrative assumed: a retailer signing its first nuclear power purchase agreement is a genuinely new customer category and it is not a hyperscaler. Constellation is selling firm clean power to whoever will pay for it, which is a better business than selling it to four buyers, and it is a slower one.

2. The Capacity Market Hit Its Ceiling, and the Call Did Not Mention It

On July 14 Constellation filed a standalone Form 8-K under Item 8.01 to report the results of PJM's 2028/2029 Base Residual Auction. Every plant it submitted cleared, and every zone cleared at the same price.

The result cuts both ways, and the second half is the more useful half for an owner. Constellation has locked 18,875 megawatts at $325 per MW-day from June 1, 2028. Against the $289.67 reference price that flowed this quarter, that is an illustrative annualized capacity revenue step of roughly $243 million pre-tax, or about $0.51 per share after tax at the 25.5% marginal rate on the revised 357 million share count. Against the $279.80 six-month average the step is nearer $311 million and $0.65. Those are arithmetic on two disclosed numbers rather than a company estimate, they ignore hedging, and they are reduced in Illinois by the carbon mitigation credit clawback, which grows as capacity prices grow. The direction is not in doubt: a visible, contracted, high-single-digit percentage step in base earnings arriving in the middle of 2028.

One further detail from the same filing matters for how much of that step survives. "Capacity revenues for nuclear units are included in the gross receipts calculation for the Production Tax Credit." Since the company already records no material production tax credit because gross receipts exceed the phase-out band for most units, higher capacity revenue no longer costs anything in forgone credit. The floor is already surrendered; the upside above it now flows.

Assessment: This is the cleanest example we have seen of a company declining to discuss a disclosure that is simultaneously the confirmation of a bear point and the crystallization of a bull one. Bear-2 in our standing thesis moves from emerging to materializing: the capacity re-rate that drove the 2024 and 2025 narrative is administratively finished. But it finished at the top of the band, not the bottom, and that is a materially better outcome than a collar implies in the abstract.

3. Crane Cleared Both of Its Binary Gates

The two items we identified as the most tradeable events in the story both resolved favorably inside the quarter.

"During the quarter, the NRC approved the Crane new fuel licensing amendment request, clearing the path for the receipt of new fuel and representing another significant milestone towards returning the facility to service in the second half of 2027. In addition, FERC granted the waiver request to transfer the capacity injection rights from our Eddystone facility, which is slated to close, to Crane."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

The waiver was requested in March with a decision sought by June 1; at the Q1 call the general counsel guided to a "June, July" response and FERC delivered in June. The company expects the transfer to clear many of the transmission contingencies PJM's Phase I study had flagged with in-service dates extending as late as December 2030.

Assessment: The distinction management drew in May was that the plant restarts regardless and only the full capacity credit was in FERC's hands. FERC granted it. In a market clearing at $325 per MW-day, the difference between a 2027 and a 2031 full-deliverability date for a restarted nuclear unit is worth real money, and that binary is now closed in the company's favor. The remaining Crane risk is execution on a second-half-2027 restart, which is an operational question rather than a regulatory one, and this fleet's operational record is the strongest part of the story.

4. The Guidance Raise: How Much Is Business and How Much Is Share Count

Guidance moved to $11.50 to $12.50 from $11.00 to $12.00, the first raise since the range was set at the March business and earnings outlook.

"Strong commercial execution combined with the benefits of our disciplined capital allocation gives us the confidence to raise the midpoint of our full year guidance by $0.50 per share just halfway through the year. We will revisit our full year outlook in the Q3 call as we put the summer behind us."
— Shane Smith, EVP and Chief Financial Officer

The deck names three components and quantifies none: commercial outperformance through higher realized customer margins and portfolio optimization, accretive share repurchases, and an offset from higher operating and maintenance expense tied to performance-related compensation. One of the three can be sized from the footnotes. Initial guidance assumed 361 million average diluted shares; revised guidance assumes 357 million. On the original $11.50 midpoint, that 1.1% reduction is worth approximately $0.13, or roughly a quarter of the increase.

Assessment: Roughly $0.37 of the raise is operational, and its source is the commercial and trading function rather than the generation fleet. Higher realized customer margins and "portfolio optimization during periods of volatility" are the least durable earnings line this company has, and they are being capitalized into a full-year guide halfway through the year. That is not a criticism of the raise, which is real and was earned. It is a caution against extrapolating it: the fleet is a contracted asset, the commercial book is a market-conditions business, and this raise came from the second one.

5. The State-Credit Clawback Is Two-Thirds Discharged, and the Illinois Half Is Getting Worse

Last quarter's central finding was a $395 million pre-tax swing from nuclear production tax credit pass-through obligations that management named in one clause and never quantified. The 10-Q now shows the mechanism working itself out.

The second-quarter swing is $65 million against $395 million in the first, and the balance-sheet position tells you why: the estimated payable for refunds and pass-throughs has fallen to approximately $270 million from $1,190 million at year-end, with roughly $920 million refunded or offset during the half. The obligation created by the 2024 credit year is nearly settled. The filing still says the company "did not record a material nuclear PTC benefit as the estimate of full year gross receipts exceeded the phase-out for annual gross receipts per MWh for most units," which is the correct reading of a credit designed to pay when prices are low.

The Illinois carbon mitigation credit is the piece moving the other way. Its price is a bid price of $34.50 per MWh less energy and capacity index prices, so a rising capacity price mechanically drives it negative and Constellation pays ComEd. The average realized price was negative $5.91 per MWh this quarter against negative $0.42 a year ago.

Assessment: Bear-1 in our standing thesis was that hedging, state credit clawbacks and the production tax credit phase-out permanently damp the transmission from merchant prices to nuclear earnings. Half of that has now been answered: the pass-through drag was a 2024-vintage obligation working through the cash statement, not a permanent feature, and it is nearly exhausted. The other half is not a timing item. The carbon mitigation credit clawback is designed to grow with exactly the prices the bull case is built on, and it applies to the ComEd zone that holds 9,800 megawatts of cleared nuclear capacity. Net, we move Bear-1 from emerging to contained, with the carbon mitigation credit as the residual we keep watching.

6. $2.25 Billion of Buyback, and the Lock-Up That Never Came Up

"Since the first quarter call, we returned just under $2 billion of capital to our owners through share repurchases. Together, with the $335 million we shared on the Q1 call, we have allocated about $2.2 billion to accretive share repurchase in the 4 months since our business and earnings outlook at the end of March. We will continue to be opportunistic as we deploy the remaining $2.8 billion of available authorization."
— Shane Smith, EVP and Chief Financial Officer

The 10-Q supplies what the call did not. Roughly 7.1 million shares were repurchased in the second quarter for about $2.0 billion, and a further one million shares in July for about $250 million, an implied blended price near $278 across 8.1 million shares, or 2.2% of the shares outstanding at March 31. It also explains part of the mechanism: "In June 2026, the lock-up restrictions were waived for 11 million of these shares to permit their sale in an underwritten secondary public offering, which included two million shares that were repurchased by the Company. We did not sell any shares or receive any proceeds from the offering."

Assessment: This is what we said would happen, and it happened cleanly. The authorization was sized partly to absorb a block, a block came, and the company took two million shares of it inside an underwritten offering. The subsequent price action confirms the second half of the May read as well: the shares made a 52-week closing low of $236.50 in the weeks after the lock-up before recovering into the print. Buying back stock through a supply event at an average near $278, with the low print roughly 15% below that, is competent rather than brilliant. What is notable is that neither the lock-up, the secondary, nor the identity of the seller was mentioned once on the call, in a quarter when the company deployed 2.2% of its market capitalization.

7. Brazos Valley Closes the DOJ File, and at a Premium to What Was Paid

"Today, after a very competitive process, we announced an agreement with LS Power to sell the Brazos Valley Energy Center for $860 million or about $1,420 per kilowatt. It was great to see such a high level of interest for the asset in spite of recent ERCOT weakness."
— Shane Smith, EVP and Chief Financial Officer

The DOJ resolution required definitive agreements on five PJM assets, one ERCOT asset and a minority interest within 240 days of closing, by September 4. The Gregory minority interest went in January, the five PJM assets to LS Power for $5.0 billion in March, and the ERCOT plant, which had no announced buyer at the time of the Q1 call, now has one with four weeks to spare.

"In total, the assets that were required to be divested by the DOJ are expected to generate approximately $5.9 billion in gross proceeds, which at a nearly $1,200 per kW basis reflects a healthy premium to the implied $960 per kilowatt purchase price of the Calpine assets."
— Shane Smith, EVP and Chief Financial Officer

Assessment: Two of the three residual integration risks we flagged in May are now closed. The one remaining is a condition rather than a question: Brazos is collateral under the Calpine Construction Finance Company first-lien term loan, and the 10-Q says lenders may require consent to release it as guarantor depending on how net proceeds are applied. The parallel item on the CDHI revolver has already been resolved, with lenders releasing four plants from collateral in June after two capacity reductions. The $1,200 per kilowatt realization against a $960 per kilowatt implied purchase price is the first independent valuation mark on the Calpine transaction, it covers roughly a fifth of the acquired capacity, and it came in above cost. That is the strongest single fact in the bull case this quarter.

8. Nuclear Operations Absorbed 45 Extra Outage Days and a Turbine Upgrade

"Our 23-day average refueling outage duration in the quarter included the successful implementation of our turbine upgrade at Byron Unit 1. And even with that longer outage, the team still outperformed the industry average duration by 40%."
— Shane Smith, EVP and Chief Financial Officer

Six refuelings, 86 planned outage days against 41, and a 93.0% capacity factor. Non-refueling outage days fell to 20 from 22. Total nuclear generation of 44,160 gigawatt-hours was 2.2% lower on a fleet that spent more than twice as many days offline by plan.

Assessment: This is the pillar of the thesis that has never been in question and still is not. It is also the mechanism by which the New York license renewals and the Dresden uprate become credible: a fleet that refuels 40% faster than the industry is a fleet whose regulator and whose states are willing to extend. The acquired thermal fleet is the part of the operating story still being established, and its forced outage factor moved the wrong way this quarter.

9. The Production Tax Credit Floor Was Marked Up, Quietly

"Following the IRS publication of the 2025 inflation adjustment, we updated our forward PTC strike price assumptions, incorporating a 2.8% adjustment for 2025 and continuing to assume 2% inflation annually in 2026 and beyond, the projected 2030 PTC strike price will increase from $49.88 to $50.88 per megawatt hour. This change will increase our view of 2030 base earnings by approximately $0.30 per share."
— Shane Smith, EVP and Chief Financial Officer

The credit is indexed to the GDP price deflator and phases out between $26.00 and $44.75 per MWh of gross receipts for 2025 and 2026, with both thresholds inflating. A one-dollar move in the 2030 strike is worth $0.30 of base earnings, which implies substantial operating leverage in the index itself. The deck's own sensitivity carries the same figure the other way: production tax credit inflation at 3% rather than 2% is worth $100 million of free cash flow before growth and $0.30 of earnings per share in 2029.

Assessment: An inflation-linked floor under roughly 147 million megawatt-hours of nuclear output is a genuinely unusual asset, and it is the least discussed part of this equity. It is also the part that improves quietly, without a contract or a regulator. In a quarter where the merchant ceiling was capped by tariff, a durable inflation-indexed floor moving up by a dollar deserved more than the appendix.

10. What the Regulator Actually Did, and What Is Still Open

"In June, we received strong validation from FERC, which made clear its desire to move more quickly in establishing new pathways for serving large loads. FERC ordered every RTO to justify how their existing tariffs provide for the just and reasonable interconnection of large loads to the grid or propose revisions to their tariffs."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

PJM has since released proposals for the reliability backstop procurement and the Interim Resource Adequacy Service, set a 6.8 gigawatt target for the backstop, and opened a bilateral matchmaking process intended to pair customers with new supply before the central procurement. The auction is proposed for this fall with results expected by year-end. Co-location rules run on a slower track: management expects PJM to respond in about November after 90-day extensions, and a FERC order in the first to second quarter of 2027, against a PJM process that had previously been targeting 2029.

Assessment: The regulatory clock that defined last quarter's Hold has been pulled forward by roughly two years on co-location and by six months on the backstop, and the company has already signed contracts that the commercial lead confirmed are not contingent on either outcome. Bear-3 moves from materializing to emerging. What has not changed is that the largest prize, a repricing of the whole fleet through premium co-located offtakes, still sits behind a FERC order that will not land before 2027.

11. Leverage Went Up to Buy Stock

Net debt rose $2,337 million in the quarter while the company spent $1,971 million on repurchases and $154 million on dividends. Short-term borrowings tripled from year-end. Credit ratings are unchanged at BBB+ and Baa1, and the collateral that would be required on a downgrade below investment grade, which needs three notches, rose to $3,441 million from $2,670 million.

Assessment: Buying back 2.2% of the company with short-term debt while holding $5.7 billion of assets under contract for sale is a bridge, not a leverage decision, and the roughly $5.9 billion of expected divestiture proceeds closing by year-end is the repayment. That sequencing is defensible and it is worth watching, because if the LS Power closings slip past December the bridge stays outstanding into a year in which capital expenditure is running 60% above the prior year.

Guidance & Outlook

Implied ramp to the guide. First-half adjusted operating EPS of $5.29 is 44.1% of the new $12.00 midpoint, against a 43.1% first-half share of the full year in 2025. Reaching the midpoint requires $6.71 across the second half, versus $5.34 in the same period of 2025, growth of 25.7%. The low end requires $6.21, growth of 16.3%. The high end requires $7.21, growth of 35.0%. Given that the year-over-year first-half growth was 30.6% with the acquisition annualizing through both quarters, the midpoint looks reachable and the high end requires either a hot summer or another quarter of commercial outperformance.

Street at. Consensus entering the print sat at $11.72, above the old $11.50 midpoint and below the new $12.00 one. A post-print composite reads $12.04 against a range of $11.33 to $12.39 across nine contributors. The Street has effectively adopted the new midpoint and no more, which for a company that raised halfway through the year and explicitly said it would revisit at Q3 is a conservative place to sit.

Guidance style. Constellation set its range in March, affirmed it in May after a $0.18 beat, and raised it in August after a $0.19 beat. That is a company that absorbs the first beat and passes through the second, and it is the pattern of a management team that prefers to raise once, in size, with a reason it can name. The commitment to "revisit our full year outlook in the Q3 call" is unusually explicit and should be read as a second raise being available rather than promised.

Analyst Q&A Highlights

Seven questioners, no pushback, and the shortest Q&A of the four calls we have covered. The two questions that mattered both went at pricing, and both were declined.

What the 920 Megawatts Is Actually Worth

The dominant unanswered question of the call. Management disclosed volume, duration and counterparty quality but not price, and the deck's own March sensitivity had given a range of $20 to $50 per MWh as a premium to the production tax credit floor. Asked where inside that range the executed deals landed, the answer was that the range stands and the specifics stay private.

Q: "Okay. And then one more question. Just your range of the potential value of contracts, I mean, I think it's $20 to $50 a megawatt hour, pretty wide. Can you give us any sense of how things are trending within that range?"
— Steven Fleishman, Wolfe Research

A: "Steve, it's Shane. I mean, again, to the point of trying to protect customer sensitivity around the exact pricing, we're going to keep that range as it is. I mean I think it is important to note that when we provided that disclosure in March, we were obviously talking with a number of different counterparties and the transactions all fit that profile. So I think we're comfortable with keeping that sensitivity as it is."
— Shane Smith, EVP and Chief Financial Officer

Assessment: The refusal is defensible on customer-confidentiality grounds and it is also the entire valuation question. "The transactions all fit that profile" is the one piece of information given, and it is worth something: it says the signed deals sit inside a range whose low end is $20 per MWh above a floor the company has already told us is roughly $50 per MWh in 2030. A 2.5-to-1 spread between the low and high end of that range, applied to roughly a gigawatt, is the difference between $0.40 and $1.00 of 2029 earnings on the company's own table. Until the range narrows, modellers should take the low end.

Whether the Signed Contracts Are Exposed to Rules Not Yet Written

The structural concern behind the announcement: if PJM's Connect and Manage or Interim Resource Adequacy Service rules land differently than expected, does a contract signed today unwind or reprice? Management answered with the demand-side logic rather than the contract terms, arguing that whatever compliance path a customer chooses for its peak hours, it still has to buy energy for the rest of the year.

Q: "And just I think there have been some concern that with the Connect and Manage or IRAS like whether customers can contract for existing like whatever contracts you have, have they kind of incorporated kind of whatever outcomes could come out of that?"
— Steven Fleishman, Wolfe Research

A: "Let's suppose that in the context of Connect and Manage, a customer is deciding to use batteries or further rely on backup generation. They could do that and comply with what we see as the proposed rules going forward. But they'll still need to buy power, they'll still need to buy energy really for all of the other hours they're not managing, the other 99%. And so that's where the existing resources are going to be able to provide quick solution sets."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: The argument is the strongest structural case this company makes and it is genuinely hard to rebut: peak management and energy supply are different products, and only one of them is solved by a battery or a backup generator. It is also an answer about the market rather than about the contracts. The direct answer on the contracts came later in the call, from the commercial side, and it is the single most useful sentence in the transcript for an owner: the signed deals are not contingent on the FERC or PJM outcomes. That converts the 920 megawatts from a conditional announcement into contracted revenue and removes the obvious bear rejoinder that the company front-ran a rulemaking.

When PJM Co-Location Rules Actually Land

The largest remaining regulatory item, and the one with the longest tail. Management supplied a two-step timeline with named quarters, which is more than the prior call produced, and contrasted it against where the process had been headed at the start of the year.

Q: "Right. Congratulations on a good quarter and announcements here. I was wondering, what do you guys see as the next step in the co-location process in the PJM? Will there be like a definitive document coming out of PJM that will be the final and authoritative, I guess, document in this process as we establish the final clarity for everyone and when will that be?"
— Sophie Karp, KBCM

A: "In short, 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. So this is an area where FERC has put some good pressure on PJM and the RTOs to provide some clarity."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: A two-year pull-forward on the item that gates the biggest prize in the story, with a PJM response expected around November and a FERC order in the first half of 2027. That is real progress and it is also a reminder of the timescale: the premium co-located repricing of the fleet is a 2027 conversation at the earliest, and revenue from anything signed under it lands later still. The idea management floated, modelling load and generation together at a common point of interconnection with batteries added at existing plants, is the most commercially interesting thing said on the call and it exists today only as advocacy.

Bilateral Contracting Versus the Reliability Backstop

A recurring line of questioning tested whether the company's contracting now runs through PJM's forthcoming backstop procurement or around it. Management was explicit that the backstop is one channel of several and that customers have been transacting bilaterally for years before the process existed.

Q: "Look, a couple of clarifications here. Starting with bilateral, if you can. I know you said it was confidential, but can you speak a little bit to the timing, right, in as much as, obviously, the RBP is somewhat specific on when it goes down. How do you think about the time line here inasmuch as where are you in those negotiations vis-a-vis this potential RBP coming up in the next couple of months. Is it lagged by a couple of months here, thereafter, or is it even prior?"
— Julien Dumoulin-Smith, Jefferies

A: "I think the RBP is one avenue, but it is not the exclusive avenue for customer conversations. And indeed, I think customers will consider whether they want to participate in that or do something completely separate, bilaterally separate as they've done now for a couple of years before this PJM process was even dreamed up."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

A: "Contracts today, those are signed contracts. So if the question was, are we dependent upon something on the FERC decisions or PJM decisions. These deals are not dependent upon those outcomes."
— Daniel Eggers, Constellation Energy

Assessment: Useful because it decouples the story from a single dated event, and because the follow-on answer from the commercial side settles the contingency question outright. If the fall backstop auction disappoints, that is not the same as contracting stalling, and the 920 megawatts already signed outside the process, on terms not dependent on FERC or PJM outcomes, is the proof. The corollary is that the backstop auction result, expected by year-end, will be a poor read on Constellation's own contracting, and investors treating it as one will misprice the reaction either way.

What the New Capital-Allocation Floor Represents

The 2029 sensitivity table gained a floor this quarter where it previously had only an upside figure. The question asked whether anything new sat behind it on the growth-investment side. The answer was that the floor is entirely retrospective.

Q: "Just want to turn to Slide 13, if I could. And looking at the earnings and free cash flow before growth opportunities in '29. And I think the capital allocation is bolded on the table. I think that might be new this quarter. I saw the share repurchases there. But just wondering on the growth investment side, is there anything kind of new to think about with regards to this table here?"
— Jeremy Tonet, JPMorgan Securities

A: "So in the initial disclosure back in March, we gave -- we didn't give a floor, we kind of gave a $0.50 plus as an upside. And so what the $0.20 as a floor is meant to represent is the progress we've made in the last 4 months from those accretive $2.2 billion of buybacks. So I wanted to establish that floor. Naturally, that raises kind of our view of the higher end and we'll see what's beyond. From a growth angle, obviously, we continue to explore, but there's nothing that we've disclosed that would directly inform that range at this time."
— Shane Smith, EVP and Chief Financial Officer

Assessment: A clean, honest answer that quietly resets what the sensitivity table means. The $0.20 floor is not optionality, it is arithmetic on shares already retired, which makes it the only line on that table an investor can bank. Everything else on it remains contingent. The more informative half is the last sentence: nothing has been disclosed that would inform the growth-investment range, which after a $21.8 billion acquisition is a reassuring silence rather than a worrying one.

Why ERCOT Forwards Stayed Weak

The same contradiction that opened the prior quarter's Q&A returned: an enormous announced data-center pipeline against soft forward prices. Management's answer this time was more concrete and less hedged, attributing the softness to build-out sequencing rather than to demand doubt.

Q: "And curious if you could also maybe give your view on the outlook for just on the power market side of things for ERCOT outlook for power prices and spark spreads from here. We know the market has been under pressure. And curious your view on that and maybe also just the battery storage, how that's impacting the market from here, too."
— David Arcaro, Morgan Stanley

A: "What we're seeing in Texas is what we fully anticipated. The battery storage and other things you're talking about started earlier and they're arriving on the grid earlier than the load is. And so we talk a lot about data centers in Texas. But if you really take a look at where the construction of that build-out lies, you'd find that the vast majority of the data centers that are anticipated are still at some stage of construction and not on the grid."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: The sequencing argument is correct and it is testable: storage interconnects in months and data centers in years, so a period of storage-led softness before load arrives is the expected shape rather than a failure of the thesis. The claim that the outcome was "completely expected" and positioned for is corroborated by the segment data, where ERCOT's revenue net of fuel rose 4.8% and its margin expanded 5.3 points in a quarter when regional energy prices fell about 10%. It is also the second consecutive quarter in which ERCOT delivers the operating evidence while PJM carries the valuation.

Whether New Nuclear Construction Is Close

An adjacent question probed whether the buyback is crowding out organic investment and whether new nuclear economics work anywhere in the country yet. The answer was the most deflating on the call and also the most disciplined.

Q: "And then just a higher level question. You've been allocating capital to share buybacks quite a bit, and that makes sense given the variation and the opportunistic nature of that. How do you think about jump starting the cycle of investing organically into maybe new builds in addition to the -- let's say, nuclear start you're working on. Is there a place in the U.S. that you think is where the economics work for that right now?"
— Sophie Karp, KBCM

A: "We're having some good conversations in New York about the future of nuclear. But there's nothing right now that I would describe as imminently on the horizon for investment in new nuclear. What you're seeing us do and what you'll see us continue to do here is to prepare our sites for new nuclear development because that work's got to get done anyway. The permitting, the early site permits that are needed. But it's not yet going to show up on financial disclosures for capital. It's not that imminent."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: Two consecutive quarters of the same message with different specifics: the uprates contribute nothing before 2030, and new nuclear is not on the capital plan at all. Any model carrying new-build contribution inside the guidance horizon should carry zero. The discipline is the right call given a balance sheet that just absorbed $21.8 billion, and it also means the growth beyond 2029 rests on contracting the existing fleet at premium prices rather than on adding megawatts.

What They're NOT Saying

Last quarter the gap between the filings and the call was the most striking feature of the print. It narrowed this quarter, mostly because the news was good and management wanted to deliver it. What remains unsaid is more concentrated and, in one case, larger.

1. That the 2028/2029 capacity auction cleared at the administrative cap. Every one of the 18,875 megawatts Constellation submitted cleared at $325 per MW-day, the exact ceiling FERC approved in April. It was reported in a standalone Form 8-K on July 14 and restated in the 10-Q. It appeared in neither the prepared remarks, the deck's highlights, nor any of the seven analyst exchanges. This is simultaneously the confirmation that the merchant capacity re-rate is administratively finished for two delivery years and the crystallization of a visible base-earnings step from June 2028. Not discussing either is a choice.
2. What the 920 megawatts is priced at, and that as executed it contributes nothing to 2029. The company's own sensitivity table values one gigawatt of nuclear offtake at $0.40 to $1.00 of 2029 earnings per share. It signed 0.92 gigawatts, and the same presentation says the agreements "have later start dates and are not expected to materially impact 2029 earnings." Both statements are in the same deck, three slides apart, and the reconciliation between them was never drawn: the sensitivity assumes contracts running through 2029, and these begin in 2029 and ramp to 2032.
3. That Calpine's earnings still cannot be determined. Second consecutive quarter with the same conclusion and a different reason. In Q1 it was purchase accounting; now it is "the commencement of integration activities for certain functions and the consolidation of financing activities." The roughly $2 per share of full-year accretion remains the load-bearing number in the guide, and there is still no quarterly marker against it.
4. The composition of the guidance raise. Three drivers were named and none quantified. The footnotes let an outsider size one of them: the assumed share count fell from 361 million to 357 million, worth approximately $0.13 on the original midpoint, or roughly a quarter of the $0.50 increase. Whether the remaining $0.37 is retail margin, wholesale optimization or trading was not addressed, and those three have very different durability.
5. That reported operating income fell 39.0% and GAAP earnings per share fell 46.8%. Neither figure was mentioned on the call, in the deck's highlights or in the release's summary of drivers, which described the quarter as reflecting "the addition of Calpine and favorable market and portfolio conditions, partially offset by unfavorable nuclear outages." All of that is true of the adjusted result and none of it explains the reported one.
6. The lock-up, the secondary offering and who sold. Eleven million shares had their lock-up waived in June to permit an underwritten secondary, and the company bought two million of them. In the same window the shares set a 52-week closing low. The company deployed 2.2% of its market capitalization. None of the lock-up, the offering or the seller was raised in prepared remarks or in Q&A, in a quarter where capital allocation was one of the three named reasons for a guidance raise.
7. The Illinois carbon mitigation credit clawback, and what replaces the zero-emission credit in 2028. The realized carbon mitigation credit price fell to negative $5.91 per MWh from negative $0.42, and to negative $16.05 for the half from negative $1.23. It is a formula clawback that grows as energy and capacity prices grow, in the zone holding 9,800 megawatts of cleared nuclear capacity. Separately, management named the May 2027 end of the Illinois zero-emission credit program but did not size what Clinton and Quad Cities earn in 2028 once they revert to merchant pricing plus the production tax credit floor.
8. What the higher operating and maintenance expense is. Operating and maintenance rose $636 million, or 39.3%. The guidance bridge names "higher O&M due to performance related compensation expenses" as the offset to the raise, with no figure attached. For a quarter in which the adjusted result cleared consensus by 8%, an unsized compensation accrual that partially offsets the guidance increase is a number worth having.

Market Reaction

• Pre-print setup: Closed at $265.12 on Wednesday, August 5, down 25.0% year to date against an S&P 500 up 12.8%, down 21.7% over twelve months, and up 10.6% over the trailing thirty days from a July 7 low of $239.71. The 52-week closing range entering the print was $236.50 to $403.95, so the shares sat 34.4% below the high and had already made a new low in the weeks after the June lock-up expiry.
• Reaction session (before-the-open reporter, same day): Opened at $277.92, a gap of 4.8%, and traded as high as $280.00, up 5.6%. It then fell all session to a low of $260.68, down 1.7%, and closed at $261.10, down 1.5% or $4.02. The peak-to-close give-back was 6.8%.
• Volume: 4.9 million shares against a thirty-day average of 3.2 million, 1.5 times normal.
• Relative: The S&P 500 fell 0.2% and the utilities sector ETF fell 0.6%. Among independent power peers, Talen rose 2.6%, Vistra rose 0.6% and NRG fell 1.4%. The fade was again specific to Constellation.
• Since the print: The shares recovered to $282.50 by August 14, gave it back to $266.83 on August 18 in a broad sell-off across nuclear names, and closed at $276.75 on August 28, 6.0% above the reaction-day close and 4.4% above the pre-print level.

The shape of the session was almost identical to the prior quarter's: a beat and a gap up, surrendered in full by the close, on above-average volume, while a direct peer finished higher. That pattern twice in a row from a stock down 25% on the year is not a verdict on either print. It is a positioning statement about a shareholder base that has been wrong for three quarters and sells strength.

What makes the second occurrence different is that the content justified the gap and the market still faded it. In May the fade was legible: a beat that traced to an acquisition, a six-week-old guide merely affirmed, and a catalyst that had gone backwards. In August the company raised guidance by $0.50, announced 920 megawatts of long-term contracts, cleared its final DOJ obligation at a premium price, and reported that both Crane regulatory gates had been granted. A tape that sells that is either exhausted or focused on something else. The candidate for something else is on the July 14 filing rather than in the release: the capacity auction cleared at its ceiling, which removes the mechanism through which most owners of this stock have expected to be paid.

The three weeks since have partly resolved the question. The shares are 6.0% above the reaction close and above where they entered the print, having traded with the nuclear complex rather than against it. The August 18 drawdown was sector-wide, not name-specific. The market took roughly two weeks to price the quarter it faded on the day, which is the more common outcome for a print whose good news is dated 2029 and whose bad news is dated 2028.

Street Perspective

Debate: Is the Contracting Breakthrough Priced Too Far Out to Matter?

Bull view: The bull case being made on the Street is that the first tranche is the hard one. Constellation went from confirming that customers had paused negotiations to signing 920 megawatts at an 18.5-year average duration with investment-grade counterparties in a single quarter, took contracted clean baseload output to roughly 30% by 2032, and did it with a pipeline management describes as robust and active. On this reading the start dates are a feature: contracts beginning in 2029 mean the fleet is being sold forward at premium prices into a decade when the production tax credit floor is inflating and PJM capacity is capped, and the earnings arrive precisely when the current guidance horizon ends.

Bear view: The bear camp contends that a company whose entire equity story is AI-driven power demand has, after two years of narrative, contracted roughly 6% of its PJM nuclear capacity at an undisclosed price for revenue that starts three years out and does not move the 2029 number by management's own statement. The sensitivity table says one gigawatt of nuclear offtake is worth $0.40 to $1.00 of 2029 earnings; the gigawatt signed is worth approximately nothing in 2029. On that view the announcement is a milestone in a process rather than a change in the earnings algorithm.

Our take: The bears have the arithmetic and the bulls have the direction. Nothing signed this quarter changes an estimate inside the guidance horizon, and any model that marked up 2027 or 2028 on this news is wrong. What changed is the probability distribution beyond it, and specifically the removal of the tail risk that motivated our Hold: the possibility that the counterparties simply would not transact on terms this company would accept. They transacted, at a duration that only makes sense if both sides think the price is right, and the commercial lead confirmed the deals do not depend on the FERC or PJM outcomes. That is worth a rating change and it is not worth an estimate change.

Debate: Does the $325 Cap Bound the Story or Floor It?

Bull view: A growing view on the Street is that a collar clearing at its ceiling is a gift disguised as a constraint. Constellation has locked 18,875 megawatts at $325 per MW-day from June 2028 with no auction risk, against a price that was $109 to $126 per MW-day two years ago. On the company's published reference prices the step from what flowed this quarter is worth roughly $243 million of annualized pre-tax revenue, and it requires no contract, no regulator and no customer. The 2029/2030 auction sits under the same collar with a floor near $175, so the downside is bounded too.

Bear view: The skeptics answer that the entire 2024 and 2025 re-rating of this sector was a capacity-price re-rating, that it is now administratively over for the delivery years inside the plan, and that a tariff which caps price at $325 in a shortage market is a political instrument that can be tightened as easily as it was extended. FERC accepted the extension citing reliability and data-center load growth; the same reasoning supports intervention in the other direction if retail bills rise. Meanwhile the Illinois carbon mitigation credit mechanically claws back a share of the increase in the zone holding more than half the cleared nuclear megawatts.

Our take: Both sides are right about different time horizons, and the bulls are right about the one that matters for a twelve-month view. A contracted, auction-cleared revenue step arriving in mid-2028 is a higher-quality earnings input than the merchant upside it replaced, and the market has not obviously priced it, given that the company declined to mention it. Over a longer horizon the bear point about political durability is the serious one: a price cap that binds is a policy choice being made annually, and Constellation's largest revenue line now sits inside it.

Debate: Is Buying Back Stock With Short-Term Debt the Right Use of This Balance Sheet?

Bull view: The bull argument is that management has been the best-informed buyer in the market. It deployed roughly $2.25 billion at a blended price near $278 across four months, including through a supply event it had pre-positioned for, retiring 2.2% of the company. The buyback is being bridged with short-term borrowings against approximately $5.9 billion of divestiture proceeds expected to close by year-end, so the leverage is temporary by construction, and $2.8 billion of authority remains.

Bear view: The bear camp notes that net debt rose $2.3 billion in a quarter to $24.0 billion, that short-term borrowings tripled from year-end to $5.2 billion, that goodwill of $11.5 billion sits behind the equity, and that half-year operating cash flow of $1.6 billion did not cover half-year capital expenditure of $2.5 billion. Repurchasing shares on a revolver ahead of asset-sale closings that require DOJ and lender approvals is a timing bet, and the same $2.25 billion could have retired acquisition debt at a certain return.

Our take: The bridge is sound and the disclosure around it is not. Management characterized $2.2 billion of repurchases as "accretive" and "opportunistic" without once addressing that it was funded on the short end against sales that have not closed. On the substance we side with the bulls: buying 2.2% of a fleet whose replacement cost the company puts at more than three times enterprise value, at a blended price 21% below where the shares started the year, is a defensible use of proceeds that are contracted and near. The condition to watch is the calendar. If the LS Power closings slip past December, the bridge becomes leverage and the argument changes.

Model Update & Valuation

We initiated coverage last quarter with a framework rather than a model. This is the first revision. The estimate changes are modest; the valuation change comes mostly from the price.

Valuation. At the August 28 close of $276.75 and 354.3 million shares outstanding, the market capitalization is approximately $98.1 billion. Adding net debt of roughly $24.0 billion gives an enterprise value near $122.1 billion, before any adjustment for the $20.5 billion nuclear decommissioning trust and the $12.6 billion asset retirement obligation, which broadly offset one another. The shares trade at 23.1 times the $12.00 guidance midpoint, 24.1 times the low end and 22.1 times the high end. Against our estimates that is 22.8 times 2026 and 20.9 times 2027. Annualizing the quarter's $154 million of dividends paid gives a yield of roughly 0.6%, so the return case remains growth and multiple rather than income.

The de-rating did most of the work. Three months ago the same framework read 26.1 times a $11.50 midpoint at $299.69. Today it reads 23.1 times a $12.00 midpoint at $276.75. The multiple has come in three turns and the earnings base has moved up $0.50, while five discrete risks have been retired. Nothing about the underlying growth algorithm has improved, and that is the point: we are not paying for a better algorithm, we are paying less for a more certain one.

Fair value framework. We hold the multiple range constant at 22 to 25 times our 2027 estimate, deliberately, so that the rating change comes from the price and the risk profile rather than from a re-rating we have assumed into existence. On $13.25 that is $292 to $331 with a midpoint near $311, approximately 12.5% above the August 28 close, or roughly 13% including the dividend. Applying management's base-earnings algorithm to the total instead would produce a 2029 figure near $20.70 against the roughly $17 that the company's own two disclosed ranges imply, which is the size of the gap between the headline and the arithmetic. We underwrite the arithmetic.

Thesis Scorecard Post-Earnings

The scorecard grades the eight pillars established at initiation, unchanged, against what this quarter's print, call and filings revealed.

Overall: the thesis is strengthened, and the reason is de-risking rather than growth. Six of the eight pillars improved or held; the one that deteriorated, the capacity collar, deteriorated into a contracted revenue step rather than into a loss. The quarter did not raise our view of what this business earns in 2029. It materially raised our confidence that it earns what management says it will earn in 2027 and 2028, and it removed the specific risk that motivated the initiation at Hold, which was that the counterparties would not transact.

Action: upgrade to Outperform with a fair value range of $292 to $331. We said in May that we wanted either a signed offtake of scale or a lower entry price. The quarter delivered a signed offtake and the intervening three months delivered a lower price, and refusing to act on both would be moving the goalposts. We would downgrade on a slip in the LS Power closings past year-end that leaves the buyback bridge outstanding, on evidence that the commercial margin outperformance driving the guidance raise is cyclical rather than structural, or on any tightening of PJM's capacity price collar below $325. We would raise conviction on a disclosed price for the signed nuclear agreements, on a third quarter that lets Calpine's earnings become determinable, or on a second contracting tranche at materially larger scale.