Constellation Energy Stock Analysis 2026: CEG Share Price, Earnings and Outlook

Constellation Energy closed at USD 300.40 on Tuesday 6 October 2026, up 12.25% from a USD 267.62 prior close, after announcing a 20-year power purchase agreement with Google covering 890MW of new nuclear capacity on the PJM grid. The capacity does not come from a new reactor. It comes from uprates — new turbines, steam generators and digital controls — at 11 existing nuclear units across Illinois, Pennsylvania and New Jersey, and it will take more than USD 4.3bn of investment, with first power from the initial upgraded facility expected in 2028.

For a Constellation Energy stock analysis, that distinction is the analysis. The company is not building nuclear; it is monetising the scarcity of nuclear it already owns, spending roughly USD 4.8m per megawatt to squeeze more output from a fleet that is already running, and locking the result away on a contract that runs to 2048. Against full-year 2026 guidance for adjusted operating earnings of USD 11.50 to USD 12.50 a share, raised by USD 0.50 at the second-quarter report on 6 August, the close puts the stock on about 25 times the midpoint of that guidance. Where that leaves it: contracted revenue visibility now extends two decades, but the megawatts underpinning it arrive between 2028 and the early 2030s, so the 12.25% repricing has paid in advance for growth that is contracted and not yet built.

What actually happened

The Google arrangement has two distinct parts, and conflating them overstates what is new. The headline 20-year agreement covers the 890MW of incremental capacity from the uprate programme. Separately, a 15-year supply agreement covers 2,700MW drawn from Constellation's existing PJM fleet, not tied to any particular generating source. Together the commitments reach 3,590MW, described as the largest such arrangement in the United States. Constellation was also selected for Google Cloud and Gemini Enterprise under a five-year technology alliance; Bloomberg reported a value above USD 1bn, and the price was not disclosed.

That split is the subject of a legitimate criticism worth putting plainly. Of the 3,590MW, only the 890MW adds generation to the PJM grid. The 2,700MW contract reallocates existing output between buyers — valuable to Constellation, because it converts merchant exposure into fifteen years of contracted cash flow, but not additional electricity. Headlines citing the full 3,590MW as new nuclear power are wrong, and anyone sizing the growth should work from 890MW.

Sources differ on how to describe the footprint — 11 reactors in some accounts, six plants across three states in others — which are consistent descriptions of the same fleet. Completion dates vary too, with first power in 2028 and the programme running towards the end of 2032. The work supports about 4,400 existing jobs and roughly 7,200 construction jobs during the build.

This was also not Constellation's only hyperscaler deal of the week. On 30 September the company announced a 20-year agreement with Amazon covering 690MW at Calvert Cliffs in Maryland, of which a 190MW uprate forms part — not an addition to the 690MW, a distinction several accounts got wrong. That uprate is expected online between 2030 and 2032 on top of the plant's existing 1,790MW, enables more than USD 3bn of Maryland investment, and gives Constellation the revenue certainty to relicense the plant for a further 20 years. A related retail supply agreement covers Amazon across the 13-state PJM market. Two 20-year hyperscaler contracts in a week is the pattern, not the Google deal alone.

Why the market reacted the way it did

A 12.25% move in a company of Constellation's size is not a reaction to volume. It is a reaction to duration and to pricing power.

Constellation's fleet is in a position most generators would envy: it is carbon-free, dispatchable, sited where data-centre load is growing fastest, and effectively impossible to replicate. New nuclear takes a decade and a construction risk premium nobody wants to carry. Uprates take a few years, use existing licences, sites and transmission, and the counterparties bidding for the output are the best-capitalised buyers in the economy. The market repriced Constellation on 6 October because the Google contract showed this scarcity converting into twenty-year contracted cash flow at scale, and because the Amazon deal six days earlier had already shown it was not a one-off.

The second-quarter print set that up. Adjusted operating earnings came in at USD 2.55 a share against a USD 2.34 consensus, and management lifted full-year guidance to USD 11.50 to USD 12.50 from USD 11.00 to USD 12.00, crediting stronger commercial results, portfolio optimisation and buybacks, partly offset by higher operations and maintenance costs tied to performance pay. The guidance assumes about 357m average diluted shares, with the low end reflecting buybacks already completed and the high end the remaining buyback capacity plus growth investment.

One note on the day's figure, because the reporting was inconsistent. Intraday accounts put the move above 14% and at 15%, and the session range ran from USD 291.11 to USD 309.80 on 13.5m shares. The close is the number that matters: USD 300.40, a 12.25% gain.

The Openbook read

Momentum is strongly positive and the quality of it is better than a single-day gain suggests. The move was not a sentiment spike on a sector theme; it was a repricing on two signed 20-year contracts within a week, following a quarter that beat and a guidance raise. The risk attached to the score is the familiar one for any stock that has rerated on a scarcity narrative: momentum built on multiple expansion rather than delivered earnings unwinds faster than momentum built on cash flow, and the earnings here arrive from 2028.

Growth improves materially, and the mechanism is unusually clean. Uprates are the lowest-risk capacity addition available in nuclear: no new site, no new licence, no new transmission corridor, and a construction scope that is largely turbine and steam-generator replacement. Three qualifications belong in the score. The 890MW sits against a fleet measured in tens of gigawatts, so this is incremental rather than transformational on volume; the value is in the contract term, not the megawatts. Delivery runs from 2028 to around 2032, so nothing lands inside the current guidance period. And USD 4.3bn for 890MW is near USD 4.8m per megawatt — cheap against new build, but a substantial commitment to a programme whose engineering gains are not uniform across reactor types.

Profitability holds up well. A USD 2.55 adjusted operating earnings print against USD 2.34 expected, and a guidance raise driven by commercial results rather than one-offs, is a healthy read. The label is worth stating explicitly: USD 11.50 to USD 12.50 is adjusted operating earnings per share, Constellation's own non-GAAP measure, and it should not be set against another generator's GAAP earnings. The honest caveat inside the raise is that O&M costs rose on performance-related pay, which is a good problem but a real cost.

Solvency is where a USD 4.3bn commitment deserves scrutiny, and it is a smaller question here than the headline implies. Constellation is funding uprates incrementally, out of the cash flows of an operating fleet, with the output contracted for twenty years before most of the spend lands — close to the ideal financing profile for a capital programme. Add the USD 3bn-plus Maryland commitment under the Amazon deal and the aggregate is substantial; against that, the company runs buybacks large enough to feature in guidance mechanics, which tells you management does not consider the balance sheet constrained. The score stays solid, with the caveat that two large programmes now run concurrently through the early 2030s.

Reward/Risk is the factor that has genuinely tightened, and not in the holder's favour. At roughly 25 times the midpoint of 2026 adjusted operating earnings guidance, after a 12.25% single-session gain, the market has capitalised two decades of contracted revenue that begins flowing in 2028. The reward case is that this is the first clear template for repricing an entire nuclear fleet, and further contracts on the same model would extend the contracted book. The risk case is specific: uprate engineering gains vary by reactor design and the pressurised water units have already captured some of their available margin, so 890MW is an engineering target rather than a delivered number; the schedule stretches to 2032; and the multiple now assumes the template repeats. If the next contract is slower to arrive than the six days between Amazon and Google implied, the gap between a contracted story and a delivered one is what gets repriced.

The read-across

The clearest peer signal came the day before the Google announcement. On 5 October the US Department of Energy, through its Office of Energy Dominance Financing, issued a conditional loan commitment of up to USD 4.2bn to Vistra for uprates and modernisation at Beaver Valley in Pennsylvania and Davis-Besse and Perry in Ohio, with an option covering future uprates at Comanche Peak in Texas. The Department projects 433MW of new capacity and the preservation of nearly 4GW of existing baseload output, extending operations 20 years beyond current licences. Vistra shares added about 4%. The commitment is conditional — technical, legal, environmental and financial requirements remain — so it is not funded yet.

Taken together, these are the same trade executed three ways. Vistra had already signed 20-year contracts with Meta covering 2,176MW of operating output plus the 433MW of uprates, deliveries starting by 2031 and completing by the end of 2034; earlier 10-Q disclosure put its January 2026 Meta agreements at 2,609MW from its PJM nuclear plants. Constellation has done it with Google and Amazon, funded from its own balance sheet; Vistra is doing it with Meta and, conditionally, federal credit support. The uprate is now the standard instrument for turning existing nuclear into contracted, data-centre-facing revenue, and the PJM nuclear fleet is being repriced on that basis.

Two implications follow for anyone screening the sector on the Openbook screener. First, the differentiator between independent power producers is no longer installed capacity but the proportion of output under long-dated contract and the credit quality behind it — a metric that barely mattered three years ago. Second, the buyers matter in their own right: Alphabet has now committed to 3,590MW in PJM alone, and procurement at that scale is a capital expenditure line with its own disclosure consequences.

What to watch next

The near-term checkpoint is the third-quarter report, which on Constellation's recent pattern falls in early November — no date is confirmed, and last year's came on 7 November. Two things matter: whether the USD 11.50 to USD 12.50 adjusted operating earnings guidance is narrowed or lifted again, given management flagged a floor USD 0.20 above the old low end and upside of more than USD 0.75; and whether the Google and Amazon commitments are quantified in the capital plan rather than described in press releases.

On delivery, 2028 is the first hard test: first power from the initial upgraded facility under the Google programme. The Calvert Cliffs uprate follows between 2030 and 2032, alongside relicensing, and the Google programme runs to around the end of 2032. Until then the evidence to watch is regulatory: uprate applications and approvals are public, and they are the earliest reliable indication of whether 890MW is achievable on schedule.

Finally, watch the next contract. What the market priced on 6 October assumes the Amazon-then-Google sequence continues. A third 20-year hyperscaler agreement would confirm the template; a quiet six months would test the multiple.