CONSTELLATION ENERGY CORPORATION (CEG)
Outperform

Contracting Scaled Fivefold in a Quarter, but the Fourth-Quarter Guide Says Most of the Beat Was Timing: Maintaining Outperform

Published: By A.N. Burrows CEG | Q3 2026 Earnings Analysis

Key Takeaways

  • The contracting channel became a contracting program. Constellation signed about 4,560 megawatts with three customers in the quarter, five times the 920-megawatt tranche that ended the pause in August. The pieces are a 20-year, 690-megawatt agreement with Amazon at Calvert Cliffs, a 20-year Google agreement that funds roughly 890 megawatts of uprates across 11 units plus a separate 15-year, 2,700-megawatt load-following agreement, and a first 270-megawatt Peak+ contract. Contracted clean baseload output by 2032 moves to about 45% from about 30% three months ago, and 91% of the nuclear counterparties are rated AA- or better. This is the conviction trigger we set in August. Pricing is still undisclosed, and none of this year's contracts enters the base modeling disclosures until the year-end call.
  • The $1.16 beat is mostly a timing beat. Adjusted operating EPS of $5.06 cleared the $3.90 consensus by 29.7%, yet the guidance midpoint rose only $0.65, to a range of $12.50 to $12.80. That range implies a fourth quarter near $2.30, flat on a year ago and $0.74 below the $3.04 the Street carried. Against the $12.20 full-year consensus the real beat is $0.45, or 3.7%. Calpine's summer earnings landed in the third quarter, and the CFO said this year's commercial outperformance is not "transferable to all years."
  • A roughly 25% rise in hub power prices produced 2.5% growth in the legacy nuclear regions. Weighted by each region's nuclear output, day-ahead prices at the hubs we match to the four regions rose about 25%, led by PJM West at 49.6% for Mid-Atlantic and ComEd at 20.4% for Midwest. Mid-Atlantic, Midwest, New York and ERCOT grew revenues net of fuel to $2,716 million from $2,649 million, after 6.2% growth in the second quarter. Nuclear production tax credit revenue fell $131 million as prices rose through the phase-out band, and the Illinois carbon mitigation credit went to negative $26.54 per MWh from negative $17.11. Calpine's $1,927 million of segment revenues net of fuel, 37% of the total, carried the quarter. Bear-1 in our standing thesis moves back to emerging.
  • One disclosure moved against the story without comment, another was not restated, and we correct one of our own. The appendix resets the 2030 production tax credit strike to $49.88 per MWh, which reverses the $0.30 of 2030 base earnings the August call announced. The two free-cash-flow ranges are absent from the deck and the transcript; the CFO referred to the March free-cash-flow outlook only to note that it excluded further premium deals and the $5 billion buyback. Separately, our August note called the $325 clearing price for 2028/2029 a capacity step up worth $0.51 to $0.65 per share. Against the $333 at which 2027/2028 cleared, it is flat, and we withdraw the figure.
  • Rating: Maintaining Outperform. We raise 2026E to $12.75 and lower 2027E to $13.00 from $13.25, because management pointed analysts back to the March 2027 outlook and this year's enhanced earnings do not carry forward. Fair value moves to $286 to $325, or 22 to 25 times 2027E, with a midpoint near $306 that is 10.7% above the $276.09 reaction close. That close is where the shares stood in August. Conviction rises to 7 from 6 on the contracting trigger.

Results vs. Consensus

Q3 2026 Scorecard

MetricActualConsensusBeat/MissMagnitude
Adjusted operating EPS$5.06$3.90Beat+$1.16 / +29.7%
Operating revenues$9,550M$9,210M–$9,233MBeat+3.4% to +3.7%
Adjusted operating earnings$1,792Mn/an/a+88.2% YoY
Revenues net of fuel$5,150Mn/an/a+71.5% YoY
Operating income (GAAP)$2,492Mn/an/a+129.5% YoY
Diluted EPS (GAAP)$5.19n/an/a+74.7% YoY
FY2026 adjusted operating EPS guidance$12.50–$12.80$12.20Above+$0.45 / +3.7% at midpoint
Implied Q4 2026 adjusted operating EPS~$2.30 at midpoint$3.04Below-$0.74 / -24%

Every pre-print read of the quarter we found sat between $3.88 and $3.91 except one at $3.72. We use $3.90, the most widely circulated figure, which makes the beat $1.16 or 29.7%. It is the largest of the three beats we have covered, after 7.0% in May and 8.1% in August, and it is also the least informative, for a reason the guidance makes plain: the company raised its full-year range by far less than it beat the quarter.

The revenue line remains not decision-useful. The two most widely cited reads put consensus at $9.21 billion and $9.23 billion, which the reported $9,550 million beats by 3.4% to 3.7%, but other reads sat as low as $8.56 billion. An 8% spread between providers on a single quarter's top line is the same symptom we flagged in May and August: operating revenues for a power marketer include unrealized marks on economic hedges and the amortization of acquired Calpine contracts, and models differ on whether to carry either. The 10-Q's unaudited pro forma comparison, which assumes the acquisition closed on January 1, 2025, is the cleaner test. On that basis third-quarter revenue was $9,550 million against $9,446 million, growth of 1.1%, and pro forma net income was $1,835 million against $1,550 million, growth of 18.4%. In August both lines shrank.

Year-Over-Year Comparison

$ millions except per shareQ3 2026Q3 2025Change
Operating revenues9,5506,570+45.4%
Purchased power and fuel4,4003,567+23.4%
Revenues net of fuel5,1503,003+71.5%
Operating and maintenance1,9911,511+31.8%
Depreciation and amortization454241+88.4%
Taxes other than income taxes225165+36.4%
Total operating expenses7,0705,484+28.9%
Gain on sales of assets12n/an/a
Operating income2,4921,086+129.5%
Interest expense, net(217)(134)+61.9%
Other, net(44)443n/m
Income before income taxes2,2311,395+59.9%
Income tax expense405466-13.1%
Net income attributable to common shareholders1,835930+97.3%
Adjusted operating earnings1,792952+88.2%
Diluted shares outstanding (millions)354313+13.1%
Diluted EPS (GAAP)$5.19$2.97+74.7%
Adjusted operating EPS$5.06$3.04+66.4%
Margin and rateQ3 2026Q3 2025Change
Revenues net of fuel margin53.9%45.7%+8.2pp
Operating margin (GAAP)26.1%16.5%+9.6pp
Adjusted operating earnings margin on revenue18.8%14.5%+4.3pp
Effective tax rate18.2%33.4%-15.2pp

Unlike August, the reported and adjusted statements point the same way. In August they diverged by $1.13 per share. This quarter GAAP earnings of $5.19 sit $0.13 above the adjusted $5.06, and the bridge is mostly two items that cut against each other: $0.36 of unrealized hedge gains and a $0.30 deferred-tax benefit from a change in forecast state apportionment are both removed, while $0.20 of Calpine merger and integration costs, $0.16 of decommissioning-related losses, $0.11 of acquired-contract amortization and $0.07 of pension non-service costs are added back. The 18.2% effective tax rate is flattered by that apportionment adjustment and by lower qualified trust income, and neither reaches the adjusted result, so the low GAAP rate is not a source of the beat.

Nine-Month Comparison

$ millions except per share9M 20269M 2025Change
Operating revenues28,17619,459+44.8%
Revenues net of fuel13,4018,376+60.0%
Operating income5,4042,488+117.2%
Net income attributable to common shareholders3,9381,887+108.7%
Diluted EPS (GAAP)$11.06$6.02+83.7%
Adjusted operating EPS (sum of the three reported quarters)$10.35$7.09+46.0%
Net cash from operating activities4,9793,432+45.1%
Capital expenditures(3,461)(1,963)+76.3%
Repurchases of common stock(3,001)(400)n/m
Dividends paid(459)(365)+25.8%

Nine-month adjusted operating EPS of $10.35 is the sum of $2.74, $2.55 and $5.06, against $2.14, $1.91 and $3.04 a year ago. It is already 81.8% of the new $12.65 midpoint, against a 75.5% nine-month share of the full year in 2025. The third quarter alone is 40.0% of the guided year, where a year ago it was 32.4%. That shift in seasonality is the single most important fact in the print, and it is the reason the scorecard's last row reads the way it does.

Sequential Comparison

We show the sequential view because a reader will otherwise build it. It is seasonally distorted in the opposite direction from last quarter: the second quarter carried 86 planned refueling outage days and the third carried 20, and the acquired gas fleet earns most of its year in the summer.

$ millions except per shareQ3 2026Q2 2026Change
Operating revenues9,5507,504+27.3%
Revenues net of fuel5,1503,481+47.9%
Operating income2,492580+329.7%
Adjusted operating earnings1,792920+94.8%
Diluted EPS (GAAP)$5.19$1.42+265.5%
Adjusted operating EPS$5.06$2.55+98.4%

Quality of the Beat

Revenue: inorganic, but no longer negative underneath. Reported revenue rose $2,980 million and the Calpine segment contributed $3,384 million, so the legacy book again shrank on a gross basis. The shrinkage is concentrated in Other Power Regions, where lower New England load volumes and lower Western prices took $444 million out of revenue and $361 million out of purchased power. Unlike August, the pro forma comparison grew: combined revenue rose 1.1% and combined net income rose 18.4%.

Margins: mix, and a summer gas fleet. The revenue net of fuel margin rose 8.2 points to 53.9%. Calpine's segment earned a 56.9% margin on $3,384 million of revenue in its peak season, above Mid-Atlantic's 54.7% and within three points of Midwest and ERCOT. Inside the legacy fleet the direction split: Mid-Atlantic's margin expanded 4.1 points while Midwest's compressed 8.1 points and ERCOT's 7.8 points. Operating and maintenance rose $480 million, of which $261 million was labor, contracting and materials including incentives, $62 million was Calpine integration and $163 million was administrative cost largely from the acquisition.

EPS: a beat on the quarter, a smaller beat on the year. Adjusted operating earnings rose 88.2% while adjusted EPS rose 66.4%; the gap is the 13.1% higher diluted share count from the Calpine consideration, partly offset by 10.9 million shares repurchased this year. The $1.16 quarterly beat converts into $0.45 against the full-year consensus because the guide implies a fourth quarter roughly $0.74 below where the Street had it. Most of the quarter's surprise is a shift in when the year's earnings arrive.

Cash: the strongest quarter of the year. Third-quarter operating cash flow was $3,426 million, the difference between the nine-month $4,979 million and the half-year $1,553 million, against $940 million of capital expenditure. The pass-through drag we tracked through the first half is nearly gone: about $1,030 million has been refunded or offset this year and the remaining payable is about $155 million, from $270 million at June 30 and $1,190 million at year-end.

Segment Performance

Constellation's chief operating decision maker allocates resources on revenues net of fuel by segment, a table that appears only in the 10-Q. For the third consecutive quarter it is where the thesis is tested, and this quarter it gives a less comfortable answer than August's.

Revenues Net of Fuel by Segment

Segment ($M)Q3 2026Q3 2025YoY% of total
Mid-Atlantic1,027892+15.1%19.9%
Midwest932943-1.2%18.1%
New York412399+3.3%8.0%
ERCOT345415-16.9%6.7%
Other Power Regions260343-24.2%5.0%
Calpine1,927n/an/a37.4%
Total reportable segments4,9032,992+63.9%95.2%
Other (unallocated, includes hedge marks and acquired-contract amortization)24711n/m4.8%
Total consolidated5,1503,003+71.5%100.0%
Memo: four legacy regions (Mid-Atlantic, Midwest, New York, ERCOT)2,7162,649+2.5%52.7%

The Legacy-Region Test, Quarter by Quarter

Four legacy regions, revenues net of fuel ($M)20262025Growth
Q12,2842,274+0.4%
Q22,6412,486+6.2%
Q32,7162,649+2.5%
Nine months7,6407,408+3.1%

In August the arithmetic produced 6.2% and we moved Bear-1 to contained on the strength of it. In November it produced 2.5% in a quarter when day-ahead power at the hub we match to each region rose about 25%, weighted by nuclear output. PJM West, Mid-Atlantic's hub, averaged $69.97 per MWh against $46.77, a rise of 49.6%; ComEd, Midwest's hub and the home of more than half the nuclear output, rose 20.4%; NYISO Central rose 3.2% and ERCOT North fell 1.8%. PJM capacity prices were 22.0% higher. Nine months of evidence now reads 3.1%. The mechanism is visible in two lines the 10-Q quantifies: nuclear production tax credit revenue fell $131 million to $44 million because higher prices pushed most units through the phase-out band, and revenue from state programs subject to refund or pass-through fell a further $60 million, to negative $280 million. Together that is a $191 million pre-tax swing against the legacy fleet. If all of it sits in the four regions, which the filing does not confirm, underlying growth was near 9.7%. That is the point of Bear-1: the price rose and much of it was handed back by design.

Segment Revenue and Revenue Net of Fuel Margin

SegmentRevenue Q3 2026 ($M)Revenue Q3 2025 ($M)YoYRNF margin Q3 2026RNF margin Q3 2025
Mid-Atlantic1,8771,763+6.5%54.7%50.6%
Midwest1,5621,390+12.4%59.7%67.8%
New York569558+2.0%72.4%71.5%
ERCOT592628-5.7%58.3%66.1%
Other Power Regions1,0991,543-28.8%23.7%22.2%
Calpine3,384n/an/a56.9%n/a
Total reportable segments9,0835,882+54.4%54.0%50.9%

Mid-Atlantic

The region where price became earnings, for the second quarter running. Revenue rose 6.5% and revenues net of fuel rose 15.1%, so the margin widened 4.1 points to 54.7%. The disclosed drivers are $160 million of favorable retail load revenue on higher contracted energy prices, partly offset by $70 million of unfavorable realized hedges on the revenue side, while purchased power and fuel actually fell $21 million. Nuclear generation was flat at 13,659 gigawatt-hours and purchased power volumes fell 8.5%, so the region again self-supplied a larger share of a higher-priced load. Eastern Mid-Atlantic capacity cleared at $329.17 per MW-day against $269.92, and New Jersey's zero-emission credit program ended in May 2025, so there is no state clawback in the comparison.

Assessment: Mid-Atlantic is the clean control in the experiment: a nuclear region with no live state credit and full exposure to PJM energy and capacity. It converted. The region also holds Calvert Cliffs, which the Amazon agreement supports through relicensing and for which the 10-Q's decommissioning note now assumes a later retirement date. The segment that most directly tests whether the fleet is levered to power prices says it is, where nothing claws the price back.

Midwest

The region where price was handed back. Revenue rose 12.4% but revenues net of fuel fell 1.2% and the margin compressed 8.1 points to 59.7%. On the revenue side the 10-Q names $175 million of favorable net generation and wholesale load revenue and $130 million of favorable retail load revenue, offset by $105 million of lower nuclear production tax credit revenue and $55 million of lower carbon mitigation credit revenue, both "due to higher energy and capacity prices." On the cost side, purchased power and fuel rose $183 million, of which $175 million is attributed to purchased power to supply load, net of generation, "primarily due to higher net transmission costs." Adding back the two credit items alone puts Midwest revenues net of fuel at $1,092 million, growth of 15.8%.

The Illinois carbon mitigation credit is the formula clawback we said in August was "getting worse." It is. The credit price equals a $34.50 per MWh bid less energy and capacity index prices, and when that is negative Constellation pays ComEd. The average was negative $26.54 per MWh this quarter against negative $17.11 a year ago, and negative $19.54 for nine months against negative $6.53. ComEd energy prices rose 20.4% to $51.44 and capacity rose to $329.17, which is the arithmetic. The Illinois zero-emission credit is in its final planning year at $1.02 per ZEC.

Assessment: This is Bear-1 in a single segment. The largest nuclear region by volume, 24,333 gigawatt-hours and up 2.9%, sold more power at higher prices and earned less from it. The drag has an end date: the carbon mitigation credit program runs through May 2027, and the company's modeling table shows no carbon mitigation credit volume after 2027 while its production-tax-credit-floor volume rises. Until then, every dollar of ComEd price strength is partly a liability. The $175 million of higher net transmission cost is the line we would most like explained, and the call did not touch it.

New York

Revenue rose 2.0% with no individually significant driver disclosed, and revenues net of fuel rose 3.3% to $412 million on a margin of 72.4%. Nuclear generation fell 7.4% to 6,177 gigawatt-hours. The New York zero-emission credit held at $14.76 per MWh, while NYISO Rest of State capacity fell 15.9% to $162.67 per MW-day and Central energy rose 3.2%.

Assessment: Still the most contracted region and still best modelled as regulated. It is also the template management keeps citing for utility procurement of existing nuclear, and the CEO urged PJM states on the call to "follow the lead of New York." That is advocacy for a policy that does not yet exist outside New York.

ERCOT

Revenue fell 5.7% and revenues net of fuel fell 16.9% to $345 million, with the margin down 7.8 points to 58.3%. No individually significant driver was disclosed on either line. Nuclear generation was flat at 2,508 gigawatt-hours and gas, oil and renewable generation rose 7.0% to 4,626 gigawatt-hours, while ERCOT North energy fell 1.8% and Houston was flat. The forward curve moved against the region between the second and third quarter reference dates: the 2027 ERCOT North around-the-clock spark spread in the company's modeling appendix fell to $17.26 per MWh from $21.85.

Assessment: In August ERCOT delivered the operating evidence while PJM carried the valuation. This quarter it did neither. The decline is small in dollars, $70 million, and it is consistent with the sequencing management described in August, in which storage reaches the grid ahead of the data-center load it is meant to serve. The Calpine gas fleet in Texas sits inside the Calpine segment, so this segment understates the company's ERCOT exposure.

Other Power Regions

Revenue fell 28.8% and revenues net of fuel fell 24.2% to $260 million. The revenue drivers are $185 million of lower retail load revenue on lower Western prices and lower New England volumes, $165 million of lower wholesale load revenue on lower New England volumes, and $90 million from ISO hourly energy netting in the West. Purchased power fell $361 million on the same volumes. Total supply in the region fell 17.9%.

Assessment: After two quarters of out-delivering its size, the segment gave back $83 million. Nine-month revenues net of fuel are still up 13.9%, so this is a quarterly reversal in a low-margin book rather than a trend. It is worth noting because the Rhode Island State Energy Center, a 609-megawatt combined-cycle plant in ISO New England, will sit in this region once the acquisition closes.

Calpine

The quarter belonged to the acquired fleet. Segment revenue of $3,384 million and revenues net of fuel of $1,927 million, a 56.9% margin, compare with $1,126 million and $1,028 million of revenues net of fuel in the first two quarters of ownership. Calpine generated 38,387 gigawatt-hours of gas, oil and renewable output and purchased a further 1,890 gigawatt-hours. Nine-month segment revenues net of fuel are $4,081 million.

"Accretion from Calpine is the key driver of higher year-over-year Q3 results. As expected, the summer months are some of the strongest for the Calpine portfolio, and this year was no exception." — Shane Smith, EVP and Chief Financial Officer

For the third consecutive quarter the 10-Q states that "it is impracticable to determine Calpine's earnings since the Acquisition Date," with the same integration-and-financing reason it gave in August.

Assessment: The margin evidence keeps validating the transaction and the seasonality explains the beat: an acquired gas fleet that earns 37% of the company's revenues net of fuel in its peak quarter will make the third quarter look exceptional and the fourth ordinary. What has still not improved is measurability. Three quarters in, the roughly $2 per share of full-year accretion in the guide has no earnings marker, only a segment margin.

Operating Statistics

MetricQ3 2026Q3 2025Change
Total nuclear generation (GWh)46,67746,477+0.4%
Nuclear fleet capacity factor (excl. Salem and STP)97.2%96.8%+0.4pp
Nuclear capacity factor, June to August99.2%n/an/a
Planned refueling outage days2023-3 days
Non-refueling outage days65+1 day
Natural gas, oil and renewables generation (GWh)44,9826,174n/m
  of which Calpine (GWh)38,387n/an/a
Purchased power (GWh)17,01517,984-5.4%
Total supply and sales (GWh)108,67470,635+53.9%
Equivalent forced outage factor (gas, oil, pumped storage)5.9%n/a6.2% in Q2, 4.5% in Q1
Renewable energy capture96.4%96.8%-0.4pp

The second-highest quarterly capacity factor in the company's history, with no unplanned outage days during the summer months when PJM set peak load records. The gas, oil and pumped-storage fleet's forced outage factor improved to 5.9% from 6.2%, which is the third data point we said we wanted before assuming the acquired fleet runs at legacy availability. It is not yet there, but it moved the right way in the quarter that mattered most. The nine-month capacity factor of 94.2% against 95.3% reflects 205 planned refueling days against 152, front-loaded into the first half.

Market Reference Prices

Reference priceQ3 2026Q3 2025Change
PJM West energy ($/MWh)69.9746.77+49.6%
PJM ComEd energy ($/MWh)51.4442.72+20.4%
NYISO Central energy ($/MWh)51.0749.51+3.2%
ERCOT North energy ($/MWh)34.4135.05-1.8%
ERCOT Houston energy ($/MWh)36.0535.90+0.4%
ISO-NE Southeast Massachusetts energy ($/MWh)54.0550.43+7.2%
CAISO NP15 energy ($/MWh)43.1841.35+4.4%
PJM Eastern Mid-Atlantic capacity ($/MW-day)329.17269.92+22.0%
PJM ComEd capacity ($/MW-day)329.17269.92+22.0%
NYISO Rest of State capacity ($/MW-day)162.67193.33-15.9%
Illinois zero-emission credit ($/MWh)1.021.17-12.8%
New York zero-emission credit ($/MWh)14.7614.760.0%
Illinois carbon mitigation credit, average realized ($/MWh)(26.54)(17.11)n/m
Henry Hub natural gas ($/MMBtu)2.873.03-5.3%

Production Tax Credit and State-Program Lines

Impact to operating revenues ($M)Q3 2026Q3 2025Change
Nuclear production tax credit revenue44175(131)
State-sponsored programs net revenue (programs subject to refund or pass-through)(280)(220)(60)
Combined(236)(45)(191)
Impact to operating revenues ($M), nine months9M 20269M 2025Change
Nuclear production tax credit revenue69220(151)
State-sponsored programs net revenue (programs subject to refund or pass-through)(555)(30)(525)
Combined(486)190(676)

Two different things are happening in this table and they should not be confused. The 2024-vintage pass-through obligation that depressed the first half is nearly discharged: the payable is about $155 million against $1,190 million at year-end. The state-programs line nevertheless went to negative $280 million in the quarter, because it includes the Illinois carbon mitigation credit, which pays ComEd whenever energy plus capacity exceeds $34.50 per MWh. And the production tax credit is behaving as designed: a floor that pays when prices are low and disappears when they are high. A year ago the credit was worth $175 million in the quarter. This year, with PJM West near $70, it was worth $44 million.

Key Topics & Management Commentary

Overall Management Tone: The most assertive of the three calls we have covered, and the first in which management argued with the market rather than with the regulator. Prepared remarks opened with the election and closed by saying the market had not "adequately valued the business," with a capital-allocation response promised by the fourth-quarter call. Management was specific on contract volume, counterparty quality and uprate cost, and as unwilling as in August on contract price. Q&A again drew no pushback, and the two changes least favorable to the story, the implied fourth quarter and the 2030 production tax credit reset, went undiscussed.

1. Contracting Became a Program

In May management confirmed that some hyperscaler counterparties had paused negotiations pending PJM's rules. In August it reported a 920-megawatt first tranche. In November it reported 4,560 megawatts in a single quarter with three customers, about 5,500 megawatts since the March business update and more than 8,500 megawatts so far this year.

"Last quarter, we announced 920 megawatts of long-term nuclear agreements. Since then, we have signed 3 additional long-term agreements totaling about 4,560 megawatts, nearly 3,500 megawatts support our existing fleet and roughly 1,100 megawatts support new nuclear capacity through our uprate program." — Joseph Dominguez, Chairman, President and Chief Executive Officer

The deck converts the volume into duration. Contracted clean baseload generation reaches 60 million MWh in 2029, 71 million in 2030, 83 million in 2031 and 91 million in 2032, about 45% of the clean baseload fleet, against about 30% in the August presentation. The company says it has signed 43 million annual MWh since March and still has 113 million MWh of 2032 output available, 65 million of it at Northern Illinois Hub, 39 million at PJM West and 9 million in ERCOT. Of the counterparties supporting the nuclear agreements, 91% are rated AA- or better and the remaining 9% between BBB+ and A+.

"This is a platform heretofore unseen where so much EPS growth and cash flow is divorced from the day-to-day movement in forwards through long-term contracts and the PTC price floor." — Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: This is the trigger we wrote in August: a second tranche at materially larger scale. It fired, at five times the first, and it raises our conviction. Two qualifications travel with it. None of this year's agreements is yet in the company's base modeling table, whose contracted-clean volumes "do not include PPAs signed in 2026," so the earnings it adds are a year-end disclosure rather than a number anyone can model today. And on a rough comparison of the company's two tables, only about 6 million of the 2029 MWh are new relative to that modeling table; most of the volume arrives between 2030 and 2032.

2. What the Amazon and Google Agreements Actually Are

The headline count mixes instruments, and they are worth separating. The September 30 Amazon agreement is a 20-year, 690-megawatt power purchase agreement with a related retail supply agreement, which will fund more than $3 billion of investment at Calvert Cliffs including a 190-megawatt expansion due between 2030 and 2032. The October 6 Google announcement has two parts: a 20-year power purchase agreement that funds more than $4.3 billion of uprates across 11 units in Illinois, Pennsylvania and New Jersey for about 890 megawatts, the first expected in 2028, and a separate 15-year "financial energy load following agreement" for 2,700 megawatts in PJM. The third deal is the 270-megawatt Peak+ contract with an unnamed investment-grade customer, which starts in 2028.

The 10-Q adds the caveat that the release omits: "Realization of the anticipated investments and capacity expansions remains subject to regulatory approvals, successful completion of the planned uprates, continued operation and relicensing of the applicable facilities, and other conditions."

Assessment: The 2,700-megawatt Google agreement is the largest component of the quarter's 4,560 megawatts and it is a financial instrument, not a unit-contingent sale of a plant's output. It may well carry premium economics, and the CFO said the contracted clean megawatts are all long-duration, but a 15-year load-following financial agreement behaves differently from a 20-year physical power purchase agreement in a stress case. We would count the physical agreements as the core of the program and the financial one as its largest unknown.

3. Uprates: 1,165 Megawatts for About $4.3 Billion

The CFO sized the organic investment behind the contracts for the first time. The committed uprates total 1,165 megawatts at ownership: 110 megawatts in 2027 to 2028, 605 megawatts in 2029 to 2030 and 450 megawatts in 2031 to 2032. Total project cost is about $4.3 billion, of which $1.3 billion was already in the March growth-capital plan and about $3.0 billion is incremental growth capital from 2028 through 2032.

"In fact, the capital cost of our nuclear uprates is not materially outside the range of current CCGT estimates. And unlike new gas generation, the uprates come with little to no incremental fuel or operating and maintenance costs." — Shane Smith, EVP and Chief Financial Officer

The arithmetic is about $3,690 per kilowatt, against the $3,000 per kilowatt combined-cycle cost the deck uses for its own new-build comparison. Asked why the cost came in lower than analysts expected, the CEO said earlier third-party estimates had combined license-extension spending with uprate spending. The conditional $5 billion loan commitment from the DOE's Office of Energy Dominance Financing would cover up to 80% of eligible costs for 30 years at Treasuries plus 37.5 basis points, which the CFO put at about $40 million of annual pre-tax interest savings if fully used.

Assessment: On a per-kilowatt basis the uprates cost about 23% more than the gas benchmark and earn their keep through zero fuel cost, a 20-year contracted offtake and a federal loan at a 37.5-basis-point spread. That is a sound trade, and $40 million of interest savings is worth roughly $0.08 per share. It is also a reminder that the earnings from the uprate half of the program arrive between 2029 and 2032 and come with $3.0 billion of capital that the March plan did not contain.

4. Peak+, and the Answer to the Claim That Uprates Were the Only Path

The CEO addressed directly a concern circulating since August: that once the uprates were sold, the existing fleet had no further contracting path.

"I know that there's been some chatter that once we sold the uprates, our ability to contract the existing fleet would somehow be exhausted. It is not. First, as a factual matter, we have not finished exploring uprates at plants. Design and engineering work continues. And second, while uprates are valuable and certainly in very high demand, they were never the only pathway for contracting." — Joseph Dominguez, Chairman, President and Chief Executive Officer

Peak+ pairs unit-contingent energy, capacity and attributes from existing nuclear with new capacity for peak hours: demand response from the retail book, third-party storage, or peaking capacity at owned sites. The CEO argued the bundle can be priced "well within" the $110 to $130 per MWh levelized cost of a new combined-cycle plant while still earning a premium over the production tax credit floor. The first contract is 20 years and 270 megawatts.

Assessment: Peak+ is the most strategically important idea on the call, because it converts the company's argument about the grid ("America has a peak capacity problem, not an energy problem") into a product that sells existing megawatts. It is also the least proven: one contract, an unnamed counterparty, and a stated willingness to accept components that individually miss the 10% unlevered hurdle so long as the bundle clears it. The CFO said Peak+ volumes will appear with other contracted clean megawatts at year-end, which will be the first chance to see what the product earns.

5. The Guidance Raise, and the Fourth Quarter That Pays for It

Guidance moved to $12.50 to $12.80 from $11.50 to $12.50. The new low end equals the old high end, and the new midpoint of $12.65 is $0.65 above August's and $1.15 above the original March midpoint.

"We are raising the midpoint and narrowing our full year adjusted operating earnings guidance range to $12.50 per share to $12.80 per share. With this latest increase, the low end of our revised range is now at the top end of the guidance we provided at Q2 and $0.50 higher than the top end of our original guidance provided in March." — Shane Smith, EVP and Chief Financial Officer

Subtract the nine-month $10.35 and the range implies a fourth quarter of $2.15 to $2.45, $2.30 at the midpoint, against $2.30 a year ago and a $3.04 consensus. In dollars the midpoint implies about $819 million of fourth-quarter adjusted earnings against $719 million in the fourth quarter of 2025, an increase of 14% for a company whose third-quarter generation rose 74% on a year ago. The deck names the drivers of the raise as "commercial outperformance in a volatile market," partly offset by higher operating and maintenance expense tied to that outperformance.

Assessment: Either the fourth quarter is unusually burdened or the range is conservative, and the evidence supports a bit of both. The burdens are real: Brazos Valley left the portfolio in October, the performance-incentive accrual rose, the acquired gas fleet's winter quarter is weaker than its summer, and the production tax credit is out of the money at these prices. But a company that says 2026 will be its fifth consecutive year above its original guidance midpoint, and that just reported a third quarter worth 40% of the year, is not guiding to a flat fourth quarter because it expects one. The useful conclusion is narrower. The Street's $3.04 for the fourth quarter was never consistent with the company's own range, and the true surprise in this print is $0.45 on the year, not $1.16 on the quarter.

6. Enhanced Earnings: What Management Said Does Not Repeat

The company splits earnings into a base, built from contracts, the production tax credit floor, minimum fossil results and ten-year average commercial margins, and an enhanced remainder that captures prices and volatility above those assumptions. This year the enhanced remainder did the work, and the CFO was explicit about how far to extrapolate it.

"In terms of what I spoke to you today, I was trying to communicate that the enhanced bucket really performed well this year. Our commercial team did a great job through the volatility. And also, we had the benefit of higher generation performance, like I highlighted, in the summer months. And so those 2 both contributed to this year, but I wouldn't say that is transferable to all years." — Shane Smith, EVP and Chief Financial Officer

The modeling appendix quantifies it. The 2026 enhanced-only adjusted gross margin rose to $3,200 to $3,300 million from $2,825 to $3,025 million in August, while the 2027 figure stayed at $2,150 to $2,550 million. The performance-incentive charge applied against 2026 enhanced earnings rose to $175 million from $125 million, about $0.36 per share, against $50 million for 2027. And the CFO pointed modellers to a specific anchor.

"For those looking to calibrate your models before the call, I would draw you back to our March disclosures that included our 2027 outlook, which is still a good starting point, and remind you that the tremendous deals we have struck this year ramp starting in 2028 and beyond." — Shane Smith, EVP and Chief Financial Officer

Assessment: This is the most decision-useful paragraph management delivered, and it points down for 2027. The March outlook put 2027 base earnings at $7.60 to $7.70 and base at 60% to 65% of the total, which grosses up to roughly $11.70 to $12.80, or $12.10 to $13.25 after the $0.25 of base earnings the CFO says has been locked in since March. Against a 2026 midpoint of $12.65, that is flat to modest growth. Our August 2027E of $13.25 sat at the top of that range, and the Street's 2027 reads sit above it: $13.41 in one updated two days before the print and $13.31 in one published on the afternoon of the print. We move to $13.00, and we read the statement as Bear-5 materializing in management's own words.

7. The Production Tax Credit Floor Moved Down for 2030, Quietly

In August the CFO announced that the 2.8% inflation adjustment for 2025 lifted the projected 2030 strike to $50.88 from $49.88 per MWh, worth about $0.30 of 2030 base earnings. The November appendix reverses that under the same 2% base-case assumption. The 2030 "Remaining Units – PTC w/2% Inflation" price in the detailed modeling table is $49.88, the footnote attributes the change to "revisions for re-baselining of the 2023 index, impacting the 2% scenario only," and the table of base-EPS impacts against the August outlook shows negative $0.30 for 2030.

On the call the topic came up twice, both times in the other direction. In prepared remarks the CFO listed "higher inflation that boosts our base EPS with PTC certainty" among the opportunities still ahead, and in Q&A the CEO described the floor this way.

"Remember, we're using the PTC floor as the launching point for this. And that is also going up over time, especially as interest has continued to clock or inflation has continued to clock at a higher rate than expected before." — Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: The amount is small, about $144 million of pre-tax 2030 base earnings, and it affects one year of the schedule. The disclosure pattern matters more than the amount. A $0.30 increase was delivered in prepared remarks in August; the same $0.30 going away was delivered in an appendix footnote in November. We said in August that the inflation-indexed floor was the least discussed part of this equity. It still is, in both directions.

8. Bear-1 Returns: The Price Rose, the Credits Fell

The legacy-region test above is the quarter's most important negative for the thesis. With prices at the matched hubs up about 25% on a nuclear-output weighting, and PJM West up 49.6% in Mid-Atlantic, the four legacy regions grew revenues net of fuel 2.5%, Midwest declined, and the 10-Q's two named credit lines took $191 million out of revenue year over year. The release summarized the year-over-year driver as "favorable market and portfolio conditions and addition of Calpine, partially offset by unfavorable interest and nuclear PTC portfolio."

Assessment: Bear-1 in our standing thesis said that hedging, state-credit clawbacks and the production tax credit phase-out damp the transmission from merchant prices to nuclear earnings. In August one quarter of 6.2% growth persuaded us it was contained. Nine months at 3.1% say it is not. We move it back to emerging rather than materializing for two reasons: Mid-Atlantic, the one region with no live state credit, grew 15.1%, so the transmission works where nothing claws it back, and the Illinois carbon mitigation credit, the largest live clawback, ends in May 2027. The strategic response is the contracting program itself: a fleet moving toward 45% long-term contracted is deliberately trading merchant upside for duration.

9. Calpine's Summer, and a Third Quarter Without an Earnings Figure

The acquired fleet produced $1,927 million of segment revenues net of fuel, 37.4% of the company's total, on 38,387 gigawatt-hours of generation. Calpine's attributable operating revenues were $3,489 million for the quarter and $8,531 million for nine months. The required divestitures progressed: Brazos Valley closed in October for $860 million before adjustments, with an expected gain of about $250 million in the fourth quarter, and the five PJM assets sold to LS Power for $5.0 billion are still expected to close by year-end.

Assessment: The question we keep asking is whether the roughly $2 per share of accretion in the guide can be verified. The segment table says the business earns well; the 10-Q says, for the third time, that its earnings cannot be determined. With integration under way and financing consolidated, that may simply be permanent, in which case the accretion figure will remain a management assertion supported by segment margins. We stop expecting the disclosure and grade Bull-4 on the segment evidence and the divestiture marks instead.

10. Balance Sheet: The Bridge Is Being Repaid

$ millionsSep 30, 2026Jun 30, 2026Dec 31, 2025
Cash and cash equivalents6036973,641
Short-term borrowings3,9895,2261,650
Total debt23,38724,7008,992
Net debt22,78424,0035,351
Goodwill11,49911,527420
Assets held for sale5,6615,743126
Nuclear decommissioning trust funds20,38220,49219,336
Total assets99,18598,25357,249
Total equity33,03932,32214,853

Net debt fell $1,219 million in the quarter and short-term borrowings fell $1,237 million, while the company repurchased $1,030 million of stock and paid $150 million of dividends. In October it repaid and terminated the CDHI revolver, and Brazos Valley was released from the Calpine Construction Finance Company collateral on its sale, which closes the lender-consent question we raised in August. Credit ratings are unchanged at BBB+ and Baa1, about $7.2 billion of credit-facility capacity was available at September 30, and the collateral required on a downgrade below investment grade rose to about $3.6 billion from $3.4 billion.

"With the success we've enjoyed in the visible free cash flow now on the horizon, Constellation strongly believes in the compelling value of buybacks to Constellation and to its owners, and you should expect us to update our buyback plan on or before the fourth quarter call as is our usual practice." — Joseph Dominguez, Chairman, President and Chief Executive Officer

The quarter's repurchases were 3.79 million shares at an average of $271.62: about 1.0 million in July at $248.33, 2.1 million in August at $277.09 and 0.7 million in September at $288.52. About $2.0 billion of authority remains. The CEO's "$800 million in the third quarter" counts the period after the August call, since July's roughly $250 million was already disclosed then. Separately, the company agreed to buy the 609-megawatt Rhode Island State Energy Center from Shell Energy North America for $715 million, about $1,174 per kilowatt, described as immediately accretive and credit neutral.

Assessment: The sequence we called defensible in August is playing out as described: buy stock on the short end, repay with divestiture proceeds. One closing is done and the larger one is seven weeks from its year-end target. Two things deserve watching. The company bought more stock in August and September at $277 to $289 than in July at $248, so the buyback has not been especially price-sensitive. And the CEO's closing message that the market has not "adequately valued the business" reads as notice of a larger authorization, which on a balance sheet still carrying $22.8 billion of net debt should be funded from the LS Power proceeds rather than from new short-term borrowing.

11. The Regulatory Clock Slipped, and Management Said It Helped

PJM's one-time reliability backstop procurement, expected in August to run this fall, was accepted by FERC in September but suspended to become effective February 28, 2027. PJM refiled on October 29 with an auction window of March 1 to March 28, 2027 and results on May 10, 2027. PJM's Interim Resource Adequacy Service proposal, which would have exposed new large loads above 50 megawatts to curtailment unless they contracted for new capacity, was largely rejected by FERC on state-jurisdiction grounds. On co-location, PJM made its compliance filing in August and the company expects FERC action in the first quarter of 2027.

"So I actually -- I shouldn't say this, but I actually like a little bit of the regulatory delay because it's allowed us to have better conversations with customers and share with them what we're able to do in this space." — Joseph Dominguez, Chairman, President and Chief Executive Officer

The election framed the rest. The CEO opened by saying the market had overreacted to "fear that Democrats won't let data centers get built," pointed to existing data-center development in Illinois and Virginia, and argued that labor's interest in construction jobs makes a halt unlikely.

Assessment: The most important regulatory fact of the quarter is not in the docket: 4,560 megawatts were signed with no final PJM rules in place, which is direct evidence that the company's contracting does not wait for the regulator. That lowers the weight of Bear-3. The election raises it in a different way, through state siting and local opposition rather than through FERC, and the 6.7% fall on November 4, with Vistra down 13.8% on a day the broad market rose, shows the market pricing that risk across the group. Net, Bear-3 stays emerging.

12. Correcting Our Own Arithmetic: The 2028/2029 Capacity Price Is Flat

In August we wrote that clearing 18,875 megawatts in the 2028/2029 Base Residual Auction at the $325 per MW-day cap was a contracted step up from June 2028, worth roughly $243 million pre-tax, or $0.51 per share, against the $289.67 second-quarter reference price, and about $311 million, or $0.65, against the $279.80 six-month average. That comparison was wrong on both bases. The $289.67 was a blended second-quarter average that still included two months of the 2025/2026 planning-year price of about $270, and the $279.80 included five. The relevant comparison is the planning year immediately before June 2028, and the company's own capacity table, in both the August and November decks, shows 2027/2028 PJM nuclear at 15,525 megawatts and $333 per MW-day.

PJM cleared capacity2026/20272027/20282028/2029
Nuclear (MW)15,55015,52515,700
Fossil and other (MW)6,6002,9503,175
Clearing price, PJM units outside the ComEd CMC block ($/MW-day)329333325

On those figures the move into June 2028 is about negative $25 million a year for the nuclear fleet and positive $18 million for fossil and other, essentially flat. Nuclear capacity revenue also counts in the production tax credit's gross-receipts test, and revenue above the credit value on the ComEd carbon-mitigation units is returned to customers until May 2027.

Assessment: We withdraw the $0.51 to $0.65 per share figure. It was never in our 2027 estimate, so the correction does not move 2027E, but it was part of the offsetting argument for Bear-2, and that offset is gone. The cap binding at $325 is now simply a ceiling: capacity revenue for the delivery years inside the plan is flat at the top of the collar rather than rising into it.

Guidance & Outlook

MetricMarch 2026August 2026November 2026Change this quarter
FY2026 adjusted operating EPS$11.00–$12.00$11.50–$12.50$12.50–$12.80Midpoint +$0.65; range narrowed
Assumed average diluted shares361M357M356M-0.3%
Base EPS growth, 2026–202920%+20%+20%+Maintained; refresh through 2030 at year-end
Long-term rolling three-year base EPS growth10%+10%+10%+Maintained
Free cash flow before growth, 2026–2027 / 2028–2029$8.4B / $11.5B–$13.0B$8.4B / $11.5B–$13.0BNot restatedAbsent from the deck
2029 capital-allocation upside to EPS$0.50+$0.20–$0.75+$0.25–$0.80+Floor +$0.05
2030 PTC strike, 2% inflation case ($/MWh)49.8850.8849.88-$1.00; -$0.30 to 2030 base EPS
2026 enhanced-only adjusted gross marginn/a$2,825M–$3,025M$3,200M–$3,300M+$325M at midpoint
2027 enhanced-only adjusted gross marginn/a$2,150M–$2,550M$2,150M–$2,550MUnchanged
2026 performance incentive adjustmentn/a($125M)($175M)-$50M
Buyback authority remaining~$4.7B (May)~$2.8B~$2.0BUpdate due by the Q4 call

What the Range Implies for the Fourth Quarter

Guidance pointFY2026Implied Q4 (less $10.35)vs Q4 2025 ($2.30)vs Q4 consensus ($3.04)
Low$12.50$2.15-6.5%-$0.89
Midpoint$12.65$2.300.0%-$0.74
High$12.80$2.45+6.5%-$0.59

The figures are on a sum-of-quarters basis, which can differ from full-year EPS on average shares by a few cents. Even the top of the range sits $0.59 below what the Street carried for the fourth quarter.

Implied ramp. The first nine months are 81.8% of the midpoint against a 75.5% nine-month share in 2025, and the third quarter alone is 40.0% against 32.4%. On the second half, the Street had $6.94 ($3.90 plus $3.04) and the guide implies $7.36 at the midpoint, a second-half beat of $0.42.

Street at. Full-year consensus entering the print was $12.20, below the new low end. The 2027 reads we found were $13.41 two days before the print and $13.31 on the afternoon of the print. The full-year figure has to rise at least $0.30; the fourth quarter has to fall by $0.60 to $0.75; and the CFO's pointer to the March 2027 outlook puts pressure on the 2027 figure.

Guidance style. Set in March, affirmed in May after a $0.18 beat, raised $0.50 in August after a $0.19 beat, and raised $0.65 at the midpoint in November after a $1.16 beat. The company passes through about the amount it can name and leaves the rest in the fourth quarter. That pattern, together with what management describes as a fifth consecutive year above the original midpoint, argues that the top of the range is a more likely outcome than the midpoint.

What comes next. The year-end call will update 2027 and 2028 guidance, refresh the growth outlook through 2030, add this year's contracts to the contracted-clean modeling disclosures and, on or before that call, update the capital-allocation plan. That is the most important disclosure event for this equity since the Calpine close.

Analyst Q&A Highlights

Seven questioners, no pushback, and a Q&A that spent most of its time on how the contracts are built rather than on what they earn. Nobody asked about the implied fourth quarter or the 2030 production tax credit reset.

Whether the $20 to $50 Premium Still Holds for Peak+ and New Deals

The dominant line of questioning was whether the premium-to-floor range from the March sensitivity table still describes the economics, now that the product set includes uprates, Peak+ bundles and attribute sales. Management held the range on every version of the question and said the new product is a way to stay inside it rather than a reason to widen it.

Q: "I wanted to ask a little more on the Peak+ that you announced today and just how we should be thinking about the economics of that opportunity. And then how much of that is going to be a template for going forward, maybe ex the uprate opportunities?"
— Multiple analysts incl. Michael Sullivan, Wolfe Research; Constantine Lednev, Wells Fargo

A: "For our existing portfolio, it's safe to assume that the sensitivities we gave you, and we've highlighted on Page 13 of the $20 to $50 a megawatt-hour, those still exist, right? So this is essentially our next lever to ensure that we're able to contract within that sensitivity range we've already provided. And so this product will show up similar to other clean contracted megawatts when we update the disclosures in the year-end call."
— Shane Smith, EVP and Chief Financial Officer

Assessment: The answer is consistent with August's "the transactions all fit that profile," and it is the third consecutive quarter in which the only price information is a range whose top is 2.5 times its bottom. Holding the range while the volume quintuples is mildly reassuring, because it means the larger deals did not require a price concession. It still leaves the low end as the modelling default.

How Much of This Year's Outperformance Is Structural

The most consequential answer of the call came in response to a question about whether the conditions behind the second guidance raise would persist. Management drew a line between base earnings, where the new contracts will land, and enhanced earnings, which carried 2026, and declined to extend the latter.

Q: "And just maybe quickly touching on the strong financial outcomes this year. Maybe how much of the conditions driving that second guidance raise do you see becoming structural?"
— Constantine Lednev, Wells Fargo

A: "The long-term agreements are going to come through the base earnings profile, right? So you're also -- in the long run, as you're moving those into more on a contracted area, you're probably going to get a smaller contribution out of the enhanced. [...] Our commercial team did a great job through the volatility. And also, we had the benefit of higher generation performance, like I highlighted, in the summer months. And so those 2 both contributed to this year, but I wouldn't say that is transferable to all years."
— Shane Smith, EVP and Chief Financial Officer

Assessment: An unusually candid answer, and the reason our 2027 estimate goes down while our 2026 estimate goes up. It also describes the mix shift that defines the next three years: as the fleet is contracted, earnings migrate from the volatile enhanced bucket to the visible base bucket, and total growth runs below base growth while the migration happens.

Whether the Delayed Backstop Auction Changes the Bilateral Opportunity

The question tested whether the slip in PJM's reliability backstop procurement to March 2027 had slowed bilateral contracting. Management said the opposite: the delay removed a deadline that had compressed customer conversations, and the direction of the rules is now clear enough for customers to transact.

Q: "And I was wondering if you could also maybe give us an update on how you're thinking about the bilateral contracting opportunity in PJM. I guess we've seen the time line shift here in a couple of the rules and regulations are still kind of in flux. But where do you stand there? And has the RBP process shifted that opportunity at all for you?"
— David Arcaro, Morgan Stanley

A: "So I actually -- I shouldn't say this, but I actually like a little bit of the regulatory delay because it's allowed us to have better conversations with customers and share with them what we're able to do in this space. [...] So look, I think regulatory uncertainty is always a bit of an overhang, but I think it's been a little bit overbaked in the market's reaction. I think, quite frankly, we probably surprised a lot of people with the amount of contracting we've been able to do in spite of the fact that we don't have final rules yet."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: The claim is backed by the quarter: 4,560 megawatts signed with no final rule in place. It also confirms what we said in August, that the backstop auction would be a poor read on Constellation's own contracting. The auction now reports on May 10, 2027, three months after the year-end disclosure that matters more.

How Much Unused Capacity the Grid Still Has

A question on latent grid capacity drew the CEO's estimate of how long existing plants can absorb new load before new generation is needed. It supports the Peak+ logic and complicates the scarcity argument at the same time.

Q: "But how much of that capacity do you think still exists today, even if we factor in kind of curtailments, things like that, that we could better utilize our current grid besides building a bunch of new generation?"
— James West, Melius Research

A: "And our view is depending on how fast the data centers get built, we could have probably another 6 or 7 years' worth of stranded capacity, or there is presently stranded capacity that could be brought to bear and allow data centers to come online. [...] Now at some point, that will run out, but it could be closer to 2035."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: This is the most two-sided statement on the call. If the grid has six or seven years of stranded capacity, the case for paying a premium for existing firm megawatts rests on reliability, emissions and price certainty rather than on physical shortage, which is a narrower argument than the one that drove this stock in 2024 and 2025. It is also exactly the argument for Peak+, which sells the energy and buys the peak. Management is betting on the second reading; the first is the bear case in its own words.

The Election and Data-Center Development

Asked how the election and labor's role in it would affect the build-out, management argued that both parties now want guardrails rather than a halt, and that labor's stake in construction work makes a halt unlikely.

Q: "And then as you think about the recent elections, 1 of the biggest impacts, I think, is bringing labor to the table and making sure labor is appropriately compensated when we think about data centers and the AI build-out. How does that factor into your thinking?"
— James West, Melius Research

A: "I think the market scare that people got last week was not lost on Democrats. They do not want to be the reason that the economy takes a nosedive by halting data center development, nor do they want to halt the jobs that come with data center development, which, to your point, feeds an important constituency of the party. But they want to make sure these -- that there are adequate guardrails in place."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: A political forecast rather than a disclosure, and a reasonable one. The more useful part of the answer came later, when the CEO said the company has "reach across all of PJM" and does not need a data center next to a plant. That is the real hedge against local siting fights: the signed agreements sell megawatts into the grid, not into a specific building.

When the Co-Location Rules Land

The co-location timeline moved little from August, and the general counsel gave it a quarter.

Q: "First, I guess, just on PJM and the co-location proceeding and the show cause orders. Can you just talk about the process there and the direction that you guys are expecting?"
— Rinny Singh, Bank of America

A: "As you said, we're expecting the compliance filings come in from the RTOs in the next couple of weeks and then FERC action on those filings sometime in the next year, probably in the first quarter, we'd expect."
— David Dardis, EVP and General Counsel

Assessment: In August the expected window was the first to second quarter of 2027; it is now the first quarter. The premium co-located repricing of the fleet remains a 2027 event, and the quarter's contracting showed it is no longer the only path to a premium.

How the Company Prices Components That Miss Its Hurdle

A question about hurdle rates on incremental capacity produced the clearest statement yet of how the company prices bundles. The 10% unlevered threshold applies to the whole transaction, not to each piece.

Q: "And then just on -- as we look at incremental generation opportunities, how should -- how do you guys think about the hurdle rate and the kind of additionality need when contracting for long-term customers?"
— Rinny Singh, Bank of America

A: "So when we think about the hurdle rate, we have -- we've said from the beginning of the business, 10% unlevered returns. When I think about what that incremental capacity needs to come in priced at, it's all part of the broader deal, right? We think about it as part of the bigger deal, including the sale of our existing megawatts. And that has to be attractive and meet our hurdle rates, but not every individual component is going to need to be."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: Sensible economics and a disclosure risk. Bundling lets the company subsidize storage or demand response with the premium on existing nuclear, which is how Peak+ can undercut a new combined-cycle plant. It also means the return on any capital the company deploys inside a bundle cannot be judged in isolation, and investors will see only the blended result. That raises the value of the year-end disclosure on what the contracted megawatts earn.

How Much of the Available Fleet Is Really Marketable

The final question asked how much of the 113 million MWh the deck shows as available in 2032 can realistically be sold after existing obligations and delivery constraints. The answer restated the strategic objective in its broadest form.

Q: "Just of that 113 million megawatt-hours that you guys kind of cited as available, I guess, what -- how much of that is realistically marketable today after we kind of think about existing obligations, outage reserves and potential delivery constraints?"
— Nicholas Amicucci, Evercore ISI

A: "Well, largely, all of it. [...] But the objective is to take the entirety of the fleet and move it to long-term contracting. How long that will take? We'll see. But that's what we're aiming for."
— Joseph Dominguez, Chairman, President and Chief Executive Officer

Assessment: A fully contracted nuclear fleet would be a different equity: lower volatility, longer duration, and earnings that track contract escalators and the production tax credit index rather than forward curves. That is worth a higher multiple, and it is worth less upside in a tight market. We do not underwrite the end state, but the direction makes the year-end disclosures on contracted volumes and prices the central valuation input.

What They're NOT Saying

  1. That the 2030 production tax credit increase announced in August has been reversed. The August call added $0.30 to 2030 base earnings on a strike moved to $50.88. The November appendix moves it back to $49.88 under the same 2% case, attributes the change to a re-baselined 2023 index, and shows negative $0.30 against the August outlook. On the call the floor was described as "going up over time."
  2. What the guidance implies for the fourth quarter. The range implies $2.15 to $2.45, flat on a year ago at the midpoint although third-quarter generation rose 74%, and $0.59 to $0.89 below the Street's $3.04. Neither the prepared remarks nor Q&A reconciled the two, and the CFO's only calibration guidance concerned 2027.
  3. What the contracts are priced at. Third consecutive quarter. Volume, duration, counterparty ratings and uprate cost were all disclosed; price was not, beyond fitting inside a $20 to $50 per MWh premium range that is 2.5 times wider at the top than the bottom.
  4. How the 2,700-megawatt Google agreement differs from the physical PPAs. The largest piece of the quarter's 4,560 megawatts is described in the release and 10-Q as a 15-year "financial energy load following agreement." The call did not distinguish it from the unit-contingent agreements or say how it will appear in the contracted-clean disclosures.
  5. That the free-cash-flow ranges were not restated. The $8.4 billion for 2026 to 2027 and $11.5 to $13.0 billion for 2028 to 2029 appeared in March and August and do not appear in the November deck, nor does the enhanced power-price sensitivity that sat beside them. The CFO said the growth outlook will be refreshed at year-end, so this may be timing. On the call he referred to the March free-cash-flow outlook once, to note that it excluded further premium deals and the $5 billion buyback, and did not restate either range.
  6. That the legacy nuclear regions grew 2.5% in a quarter when their matched hub prices rose about 25%. PJM West, the Mid-Atlantic hub, rose 49.6%. The production tax credit line fell $131 million, the state-programs line fell $60 million and Midwest revenues net of fuel declined. The release credited "favorable market and portfolio conditions," the call credited "higher capacity prices in PJM," and neither discussed the legacy-region result.
  7. What drove $175 million of higher net transmission cost in Midwest. The 10-Q attributes the increase in Midwest purchased power to "higher net transmission costs" with no further detail. It is the single largest cost driver in the legacy fleet this quarter.
  8. That Calpine's earnings still cannot be determined. Third quarter running, same reason as August. The roughly $2 per share of accretion in the guide has no earnings marker, only segment margins.
  9. What Rhode Island State Energy Center adds. The $715 million acquisition was called immediately accretive and credit neutral with returns above the double-digit unlevered threshold. No earnings contribution or closing date was given.

Market Reaction

  • Pre-print setup: Closed at $249.95 on Friday, November 6, down 29.2% year to date against an S&P 500 up 7.8%, down 30.3% over twelve months and down 16.1% over thirty days from $298.07 on October 9. The 52-week closing range entering the print was $236.50 to $378.60, so the shares sat 34.0% below the high. The prior week included a 6.7% fall on November 4, the day after the election, when Vistra fell 13.8%, Talen 8.5% and NRG 13.5% while the broad market rose, and a further 2.5% on November 5 when the S&P 500 ETF fell 4.4%.
  • Premarket: Indicated up about 5% ahead of the 9:00 AM call.
  • Reaction session (before-the-open reporter, same day): Opened at $260.40, a gap of 4.2%. In the first five minutes, with the call still running, the shares fell through the prior close to the session low of $247.00, down 1.2%. From there they rose for most of the session to a high of $277.66 at 2:15 PM, up 11.1%, and closed at $276.09, up 10.5% or $26.14, near the high.
  • Volume: 7.3 million shares against a thirty-day average of 4.1 million, 1.8 times normal.
  • Relative: The S&P 500 rose 2.3% and the utilities sector ETF rose 1.5%. Among independent power peers, NRG rose 10.3%, Vistra 7.3% and Talen 7.2%.
  • Since our August note: The shares closed at $276.75 on August 28. They fell 4.0% on September 30, the day of the Amazon announcement, rose 12.2% on 13.5 million shares on October 6, the day of the Google announcement (Vistra rose 10.8% that day), then gave back 16.8% between October 6 and the pre-print close. The reaction close of $276.09 is 0.2% below the August close.

The first close at the high in three prints. In May the stock gained 4.6% intraday and closed down 1.3%. In August it gapped up 4.8% and closed down 1.5%. This time it opened up, broke the prior close within five minutes while the call was still running, and then rose for the rest of the day. That pattern is the opposite of the earlier two: a shareholder base that sold strength twice bought weakness this time. The session's shape is the clue: the early low came while management was still presenting, and the rest of the day was a steady climb with no afternoon fade.

Roughly 70% of the move was the group. NRG matched the 10.5% and Vistra and Talen each rose about 7% on a day the market rose 2.3%. The independent power names had fallen hard the previous Wednesday on post-election fears for data-center development, and Monday's tape was partly a reversal of that. The Constellation-specific component, roughly three points above Vistra and Talen, is the market's verdict on the quarter. It is a positive verdict, and a smaller one than the headline.

The market is still paying August's price. After two contracting announcements, a second guidance raise and a fivefold increase in contracted volume, the shares closed within 0.2% of where they stood when we upgraded. The Google day showed what the market will pay for a large contract on the day it is announced: 12.2%. The month that followed showed how quickly that is given back when the sector trades on policy. The stock is not being rewarded for duration yet, and that is the opportunity the rating rests on.

Street Perspective

Debate: Does the Election Change the Data-Center Demand Math?

Bull view: The bull case being made on the Street is that the post-election selloff priced a halt that neither party wants. Data-center development was already advancing in states with Democratic control, labor wants the construction work, and the guardrails both parties discuss are the large-load and bring-your-own-capacity rules the industry was already preparing for. Constellation signed 4,560 megawatts in a quarter without final rules, with counterparties 91% rated AA- or better, which is the strongest evidence that demand does not wait for politics.

Bear view: The bear camp contends that the risk is not federal but local. State siting rules, moratorium proposals and community opposition can slow the load that the next tranche of contracts needs, and a regulator focused on household bills can tighten the capacity collar as easily as it extended it. On this view the group's one-day fall after the election was a correct repricing of a political risk that had been ignored, not an overreaction.

Our take: The bulls are right about the signed book and the bears are right about the pace of the next one. The agreements already signed sell megawatts into PJM rather than into specific buildings, so they are insulated from local siting fights. What the election changes is the probability that the 113 million MWh still available are contracted at the same pace, and we would not extrapolate 4,560 megawatts a quarter.

Debate: Is 2026 the High-Water Mark for Enhanced Earnings?

Bull view: A growing view on the Street is that volatility is now structural. Load growth, a larger retail book from Calpine, and a fleet that ran at 99% through a record summer produced the outperformance, and those conditions persist. The company says 2026 will be its fifth straight year above its original guidance midpoint, forward power prices for 2027 rose between June and September, and a larger buyback is on the way. On this reading 2027 consensus reads of $13.31 to $13.41 are conservative.

Bear view: The skeptics point to management's own words. The CFO said this year's enhanced performance is not "transferable to all years," pointed modellers back to a March 2027 outlook that grosses up to roughly $12.10 to $13.25 including the base earnings locked in since March, and left the 2027 enhanced gross margin at $2,150 to $2,550 million against $3,200 to $3,300 million for 2026. On that arithmetic 2027 is flat to slightly up, not 5% to 6% up.

Our take: The bears have the 2027 arithmetic and we have moved our estimate toward it, to $13.00. The bulls are right that the company has a habit of beating its own conservative tools. The more important point is that enhanced earnings shrinking is the plan: as the contracts start in 2028, earnings move from the enhanced bucket to the base bucket, and the question is what the base earns, which is a year-end disclosure.

Debate: Are the Contracts Priced Well Enough to Justify Their Capital?

Bull view: The bull argument is that counterparties do not sign 15- and 20-year agreements at uneconomic prices for a seller with this balance sheet, that the company has held its $20 to $50 per MWh premium range through a fivefold increase in volume, and that the uprates come with zero fuel cost, contracted offtake and a federal loan at Treasuries plus 37.5 basis points. Contracted clean output at 45% by 2032 is worth a higher multiple whatever the exact price.

Bear view: The bear camp notes that price remains undisclosed for a third quarter, that the largest single agreement is a financial load-following contract rather than a physical PPA, that the uprates cost about $3,690 per kilowatt against the company's own $3,000 per kilowatt gas benchmark, and that management said individual bundle components need not meet its hurdle. With $3.0 billion of new growth capital arriving between 2028 and 2032, investors are being asked to fund returns they cannot see.

Our take: The economics are plausibly good, and the evidence for them is indirect: counterparty quality, duration, and a held price range. That is enough to raise conviction and not enough to raise the multiple. The year-end refresh, which will add this year's contracts to the contracted-clean disclosures, is the test, and it is the one event that could move our valuation range rather than our estimates.

Model Update & Valuation

Two estimate changes, in opposite directions, and one correction. The 2026 change follows the guidance. The 2027 change follows management's own pointer to the March outlook.

ItemPrior (Q2 2026)RevisedReason
2026E adjusted operating EPS$12.15$12.75Guidance $12.50 to $12.80. Nine months already at 81.8% of the midpoint, and a fourth quarter at the midpoint would be flat on a year ago although third-quarter generation rose 74%. We carry $0.10 above the midpoint, inside a range now only $0.30 wide.
2027E adjusted operating EPS$13.25$13.00The CFO called the March 2027 outlook "still a good starting point" and said 2026's enhanced outperformance is not transferable to all years. The March base of $7.60 to $7.70 plus $0.25 locked in since March grosses up to $12.10 to $13.25. We sit near the top for forward prices and buybacks. Growth of 2.0% on 2026E.
Legacy four-region revenues net of fuelMid single digitsLow single digits+2.5% in Q3 and +3.1% for nine months against matched hub energy prices up about 25% on a nuclear-output weighting (PJM West, the Mid-Atlantic hub, up 49.6%). Production tax credit and carbon mitigation credit offsets absorb most of the price until the Illinois program ends in May 2027.
Calpine contribution~$2.00 per share annualized~$2.00 per share, summer-weightedSegment revenues net of fuel of $1,927M in Q3 against $1,126M and $1,028M in Q1 and Q2. Earnings still "impracticable to determine."
Diluted share count~357M average for 2026~356M average for 2026Company assumption. 3.79M shares retired in Q3 at $271.62; 351.6M outstanding at October 31.
Net debt~$24.0B~$22.8BShort-term borrowings down $1.2B. Brazos Valley's $860M closed after quarter-end; the $5.0B PJM sale is guided to close by year-end.
June 2028 capacity step$0.51 to $0.65 per share~ZeroCorrected. Against the 2027/2028 price of $333, the 2028/2029 price of $325 is a small step down; net of fossil volumes, about flat.
2030 production tax credit strike (2% case)$50.88 (+$0.30 to 2030 base EPS)$49.88Company re-baselined the 2023 index; the August uplift is reversed.
2029 contribution from 2026 contractsNot modelled$0.25 to $0.62 per share, illustrativeAbout 6M MWh of 2029 contracted volume above the base table at a $20 to $50 per MWh premium. Most volume arrives 2030 to 2032.
Uprate growth capital$1.3B in the 2026–2027 planPlus ~$3.0B in 2028–20321,165 MW at about $4.3B, or about $3,690 per kW. DOE conditional loan of up to $5B at Treasuries plus 37.5 bp.

Valuation. At the November 9 close of $276.09 and 351.6 million shares outstanding, the market capitalization is about $97.1 billion. Adding net debt of about $22.8 billion gives an enterprise value near $119.9 billion, before the $20.4 billion nuclear decommissioning trust and the $12.9 billion of asset retirement obligations, both of which we leave out. The trust exceeds the obligation by about $7.5 billion. The shares trade at 21.8 times the $12.65 guidance midpoint, 21.7 times our 2026E and 21.2 times our 2027E. The dividend of $0.4265 a quarter yields about 0.6%.

Same-year comparisons. On 2026 guidance the multiple fell to 21.8 times from 23.1 times in August, because the midpoint rose and the price did not. On our 2027 estimate it was 20.9 times in August and is 21.2 times now, because the estimate fell and the price did not. The pre-print close of $249.95 was 19.2 times our new 2027E. Measured against the March 2029 base-earnings midpoint of $11.65, which rests on contracts, the production tax credit floor and minimum fossil results, the shares trade at 23.7 times, or 23.2 times including the $0.25 of base earnings locked in since March.

Fair value framework. We hold the multiple range at 22 to 25 times our 2027 estimate, the same range as at initiation and in August. On $13.00 that is $286 to $325, with a midpoint near $306, about 10.7% above the November 9 close and about 11.3% including the dividend. The unrounded midpoint of $305.50 sits $5.88 below August's $311.38 because the estimate fell $0.25 at a 23.5 times midpoint multiple. We considered and rejected raising the multiple for the contracting program: a fleet moving toward 45% contracted by 2032 with AA- counterparties deserves a lower discount rate, but we cannot yet see what the contracts earn, and the year-end refresh is a dated event that will tell us. Until then, the contracting raises our conviction and not our price.

Thesis Scorecard Post-Earnings

The scorecard grades the ten pillars of the standing thesis, unchanged in wording, against what this quarter's print, call and filings revealed.

Thesis pointStatusWhat Q3 2026 showed
Bull 1: Scarce firm clean baseload. About 55 GW post-Calpine, the largest US nuclear fleet plus about 23 GW of dispatchable gas.On track, strengthened97.2% capacity factor, the second-highest quarter in the company's history, and 99.2% across June to August with no unplanned outage days. The 10-Q's decommissioning note now assumes later retirement dates for Calvert Cliffs, FitzPatrick, Limerick and Nine Mile Point, a $1,277M reduction in the obligation. Byron Unit 2's turbine upgrade finished ahead of schedule and under budget. Gas, oil and pumped-storage forced outage factor improved to 5.9% from 6.2%.
Bull 2: Guidance does not require a new contract.On trackRaised a second time, to $12.50 to $12.80, with no contribution from this year's agreements. The raise came from enhanced earnings that management says will not carry into every year, so the pillar holds for 2026 and says little about 2027.
Bull 3: Free cash flow inflection. $8.4B before growth across 2026 and 2027, $11.5B to $13.0B across 2028 and 2029.On track, not restatedQ3 operating cash flow of $3,426M against $940M of capital expenditure; nine-month operating cash flow up 45.1%. The pass-through payable is down to about $155M. The deck did not restate either range, and $3.0B of new uprate growth capital arrives from 2028, outside the before-growth measure.
Bull 4: Calpine widens the aperture.On trackSegment revenues net of fuel of $1,927M at a 56.9% margin in its peak season, 37.4% of the company. Brazos Valley closed at $860M with a gain of about $250M expected in Q4. Third consecutive quarter of "impracticable to determine Calpine's earnings."
Bull 5: The contracting channel is open and the deals are unconditional.On track, strengthenedAbout 4,560 MW in the quarter, five times August's tranche; contracted clean baseload about 45% by 2032 from about 30%; 91% of nuclear counterparties AA- or better. Pricing undisclosed, and the largest component is a 15-year financial load-following agreement.
Bear 1: The nuclear fleet is not levered to spot power the way the narrative assumes.Contained → EmergingLegacy four-region revenues net of fuel +2.5% against matched hub energy prices up about 25% on a nuclear-output weighting (PJM West +49.6% for Mid-Atlantic, ComEd +20.4% for Midwest); nine months +3.1%. Production tax credit revenue down $131M, state-programs line down $60M, Illinois carbon mitigation credit at negative $26.54 per MWh, Midwest down 1.2%. Mid-Atlantic +15.1% shows the transmission works where nothing claws it back, and the Illinois credit ends in May 2027.
Bear 2: PJM capacity upside is administratively collared.Materializing, offset withdrawnUnchanged on the collar. We correct the offset: against 2027/2028 at $333 per MW-day, the 2028/2029 price of $325 is flat to slightly lower, not the $0.51 to $0.65 per share step we described in August.
Bear 3: The catalyst is regulatory and outside the company's control.EmergingDependence on the regulator fell: 4,560 MW signed with no final PJM rules. The backstop auction slipped to March 2027 with results May 10; IRAS was largely rejected; a co-location order is expected in Q1 2027. Political risk rose with the election, and the group fell sharply on November 4. Net unchanged.
Bear 4: Integration and balance-sheet risk.ContainedNet debt down $1.2B to $22.8B and short-term borrowings down $1.2B. Brazos closed and was released from the CCFC collateral; the CDHI revolver was repaid and terminated in October. The $5.0B PJM sale is still pending against a year-end target. Collateral on a downgrade below investment grade rose to about $3.6B.
Bear 5: The headline growth rate is a base-earnings rate, not a total.Emerging → MaterializingManagement's own words: enhanced earnings will be "a smaller contribution" as contracts migrate into base, 2026's outperformance is not "transferable to all years," and the March 2027 outlook is "still a good starting point." On the company's tools 2027 total earnings are flat to modestly higher than 2026. The 2030 production tax credit uplift was reversed.

Overall: the thesis is stronger in duration and weaker in near-term growth. The long-dated half of the case improved substantially: contracted output, counterparty quality, a priced uprate program and a federal loan. The near-term half deteriorated: the legacy fleet barely converted a large price move, 2027 now looks close to flat on a 2026 inflated by non-recurring enhanced earnings, and two favorable items from August were reversed or corrected. Two status tags moved this quarter, both against us. The pillar that improved most, the contracting channel, was already on track and has no higher tag to move to.

Action: maintain Outperform with a fair value range of $286 to $325. We graded the triggers we published in August as follows.

  • Downgrade trigger, "a slip in the LS Power closings past year-end that leaves the buyback bridge outstanding": not fired. Brazos Valley closed in October, the PJM sale remains guided to year-end, and short-term borrowings fell $1.2 billion. The test arrives in seven weeks.
  • Downgrade trigger, "evidence that the commercial margin outperformance driving the guidance raise is cyclical rather than structural": partly fired. The quarter repeated the outperformance rather than giving it back, which is the event we wrote the trigger to catch. But the CFO's statement that it is not "transferable to all years" is the cyclical reading in management's own words. We act on it through the estimate: 2027E falls $0.25 and the fair-value midpoint falls $5.88. We do not downgrade because the reduced fair value still sits 10.7% above the price.
  • Downgrade trigger, "any tightening of PJM's capacity price collar below $325": not fired.
  • Conviction trigger, "a disclosed price for the signed nuclear agreements": not fired, for the third quarter.
  • Conviction trigger, "a third quarter that lets Calpine's earnings become determinable": not fired. We now treat that disclosure as unlikely to come.
  • Conviction trigger, "a second contracting tranche at materially larger scale": fired, at about 4,560 megawatts against 920. Conviction rises to 7 from 6.

We add two triggers, both outside our own estimate. We would downgrade on a year-end 2027 guidance midpoint below $12.40, which would cut our fair-value midpoint to about $291, roughly 5% above the current price. We would consider a higher multiple range if the year-end refresh, with this year's contracts in the contracted-clean disclosures, lifts the 2029 base-earnings midpoint by $0.50 or more from the March $11.65, because that would turn the contracting program from a conviction argument into a valuation one.

Independence Disclosure As of the publication date, the author holds no position in CEG and has no plans to initiate any position in CEG within the next 72 hours. Aardvark Labs Capital Research maintains a firm-wide policy of not trading any security we cover. No compensation has been received from Constellation Energy Corporation or any affiliated party for this research.