Inside Google’s Nuclear Deal With Constellation Energy: Why CEG Stock Jumped?

Google can design an AI model, build its own chips and fill a data centre with servers. It still needs someone to keep the electricity flowing. Its new agreement with Constellation Energy tackles that problem by backing upgrades to existing nuclear reactors. 

Constellation’s shares jumped 12.25% on October 6, but the bigger investor question is whether the profits ahead can justify the excitement already priced into the stock.

Let’s break down the deal, Constellation’s rally and what changes for Google. Then we’ll test the numbers to see what each company needs to deliver for investors.

What Is the Google-Constellation Nuclear Deal?

Announced on October 6, 2026, the collaboration combines new nuclear output, a separate agreement supporting existing generation, and a technology partnership. A power purchase agreement, or PPA, is a long-term contract for electricity. Here, Google’s commitment gives Constellation greater certainty about future revenue before it commits money to upgrades.

Sources: Constellation Energy and Google joint announcement, October 6, 2026; Reuters, October 6, 2026.

The distinction between the first two rows matters. The agreements cover 3,590 MW altogether, but only 890 MW, or approximately 24.8%, is identified as new nuclear capacity. Reuters reports that the separate 2,700 MW agreement is not tied to a specific generation source. Describing the entire arrangement as 3.6 GW of new nuclear power would overstate what was announced.

The $4.3 billion figure also needs careful handling. It is Constellation’s planned investment. Neither the electricity price nor the deal’s earnings contribution has been disclosed, as MarketWatch’s reporting also noted. Investors cannot turn that investment figure into a Google contract value or a guaranteed profit forecast.

This is a grid-based arrangement involving PJM, the regional electricity system spanning parts or all of 13 states and Washington, DC. It should not be pictured as a dedicated cable taking every extra nuclear electron straight to a Google server. The investment adds generation to the shared system while the commercial agreements support Google’s energy requirements.

Why Google Is Upgrading Existing Reactors

The new nuclear capacity will come from six operating sites in Illinois, Pennsylvania and New Jersey. Constellation expects the first uprate by 2028, while Google says the projects will deliver the full 890 MW before the end of 2032. The output is additional, but the reactors already exist.

A nuclear uprate increases the output of an existing plant through approved changes to equipment or operating capability. Think of improving an established factory’s production line rather than building another factory from scratch. The site, workforce and much of the infrastructure are already there; the investment aims to extract more useful output from them.

That distinction reduces some development risks without eliminating engineering work or approvals. The US Nuclear Regulatory Commission explains that changes to a reactor’s licensed thermal power require regulatory review. A corporate announcement is therefore a starting point for execution, rather than proof that every future megawatt is ready.

For Google, the practical attraction is dependable generation alongside its other energy sources. Solar and wind are useful parts of an electricity portfolio, but their output varies with conditions. Nuclear can supply power through the night and during periods when renewable production is low. It still has refuelling and maintenance outages, so dependable should not be confused with uninterrupted.

Google is also diversifying its nuclear approach. It says its agreements involving existing-reactor uprates and restarts have now enabled more than 1.5 GW of additional US nuclear capacity. Its wider portfolio includes the Duane Arnold restart with NextEra and advanced-reactor development with Kairos Power. These are different routes with different timelines, rather than one interchangeable nuclear solution.

There is a second, less glamorous reason this structure matters: public acceptance. PJM’s January 2026 proposals sought ways to connect large new electricity users while protecting reliability and affordability, including pathways for customers bringing new generation and accepting curtailment under certain arrangements.

Our interpretation is that contributing to supply can make a data-centre expansion easier to defend. Adding demand to an already stretched system invites objections; helping expand that system gives Google a stronger argument. The companies’ affordability commitments should nevertheless be read as commitments about this arrangement, rather than a guarantee that every household’s electricity bill will fall.

Why Constellation Stock Rose

The stock-market reaction was sharply uneven. Constellation finished the session at $300.40, up $32.78, while Alphabet’s Class A shares finished at $347.68, up $1.21.

Sources: Google Finance

The most plausible explanation is that the agreement affects the two investment stories differently. For Alphabet, power is an input into a much larger business. For Constellation, long-term demand for electricity directly strengthens the commercial case for its assets. That is an interpretation of the reaction, rather than proof that the agreement caused every part of either stock’s move.

For Constellation, three things become more credible. A major customer is committing for decades, existing sites have a route to additional output, and the company has another example of turning electricity demand into an investable project. Investors may also assign value to the possibility of repeating the approach elsewhere.

This is not its first recent example. On September 30, Constellation announced a 20-year Amazon agreement covering 690 MW at Calvert Cliffs, including a 190 MW uprate. The new output is expected between 2030 and 2032. Google’s agreement therefore follows another concrete expansion commitment rather than standing alone.

The relevant investor question is whether the company can repeat these projects at attractive returns. A successful first project can demonstrate a process; several profitable projects can change the economics of a fleet. The market’s enthusiasm appears to extend beyond the output in this one announcement, but the value of that wider opportunity remains conditional on execution.

What the Deal Changes for Google and Alphabet Investors

Google’s main benefit is a better chance of converting infrastructure investment into usable computing capacity. Chips, buildings, electricity connections and customer demand all need to arrive together. An expensive server that cannot be powered is like a restaurant kitchen that has equipment and waiting customers but no working gas connection.

Alphabet’s latest financial results show why this matters. Cloud demand is expanding rapidly, while the cash required to build infrastructure is already substantial.

Source: Alphabet Q2 2026 earnings release and financial tables filed with the SEC. Figures rounded.

On its Q2 earnings call, Alphabet raised its 2026 capital-expenditure outlook to 195 billion-205 billion. Management also described continuing supply constraints and warned that depreciation and energy costs would put pressure on expenses. The nuclear agreement addresses part of that capacity problem; it does not remove the broader spending burden.

Our view is positive on the strategic decision. A company undertaking an infrastructure programme of this size should secure the inputs needed to make it productive. The stronger argument for Google is reduced risk of underused assets and delayed customer revenue, rather than an assumed windfall from cheap electricity.

The scale can be illustrated without inventing a contract price. At an assumed 90% capacity factor, the additional 890 MW would generate about 7.02 million megawatt-hours annually once fully operational. A $10-per-MWh difference in electricity cost across that amount equals roughly $70 million a year. That is a sensitivity calculation, not a disclosed saving, and it does not assume Google’s facilities consume all that output.

The more consequential benefit could be avoiding a bottleneck that prevents computing capacity from earning revenue. The amount is impossible to establish from the power announcement because the companies have not disclosed the related data-centre utilisation or revenue. Investors should assess whether future Cloud growth and cash generation improve as capacity becomes available.

There is also an extra commercial opportunity: Constellation is adopting Google’s technology. This gives Google Cloud another industrial deployment, but the technology contract’s value and profit are undisclosed. It would be premature to assign a material Alphabet earnings uplift to it.

For the wider business context, our explanation of how Google makes money shows how advertising, Cloud and other activities contribute. The nuclear deal supports that engine; it does not replace the need to judge its profitability.

What Constellation Gains Beyond More Megawatts

Constellation gains both an expansion opportunity and greater visibility over part of its existing business. The former depends on delivering new output profitably. The latter depends on how contract terms compare with the revenue and risk of operating without the agreement.

That second distinction is easy to miss. The 2,700 MW agreement supports assets that already generate electricity. Its full revenue cannot automatically be counted as additional revenue, because some existing revenue may be replaced. The economic benefit could instead be a better margin, lower volatility or a longer period of predictable cash generation.

An illustrative calculation shows why the undisclosed terms matter. If 2,700 MW operated at an assumed 90% capacity factor, it would represent approximately 21.29 million MWh annually. A $5-per-MWh improvement in realised margin against the alternative would equal roughly $106 million before tax. That is an math example, not an estimate of this contract’s output or incremental profit; the generation mix and delivery profile are unspecified.

The technology relationship offers another potential benefit. The announced focus includes planning, asset monitoring, outage management and infrastructure security. Our interpretation is that better planning or fewer avoidable outages could improve the returns from physical upgrades. Those are possible outcomes, however, rather than quantified savings already achieved.

The commercial test remains separate from the technology ambition. Google receives power and provides technology; Constellation supplies power and adopts technology. Each leg needs a sound economic purpose. Investors should not assume the cloud contract offsets Google’s electricity bill or that the two arrangements have equal values.

Constellation’s wider business also matters. Its Q2 materials reported a 55 GW fleet across several generation technologies following the Calpine acquisition, and approximately $19.6 billion of long-term debt at June 30. It is an established power operator with substantial assets and financing obligations. The nuclear announcement should therefore be assessed alongside fleet performance, integration and capital allocation.

For readers comparing different businesses in the sector, our guide to nuclear energy stocks and the AI power boom explains the distinction between operating generators, uranium suppliers and reactor developers. A Google agreement with Constellation does not establish equivalent commercial progress for all three groups.

The Original Investor Test: What Must the Upgrades Earn?

Large capacity figures describe a project’s scale. They do not establish its financial return. To make that distinction practical, we can construct a deliberately narrow test of the additional nuclear output against the announced investment floor.

Capacity is the potential rate of generation; energy is the amount actually generated over time. A capacity factor measures actual output relative to continuous operation at the stated capacity. Our 90% assumption allows for downtime and is an illustrative input, not a forecast for these upgrades.

The annual energy calculation is:

890 MW × 8,760 hours × 90% = 7,016,760 MWh.

Now assume $4.3 billion of investment, flat annual cash flows for 20 years, an illustrative 8% annual discount rate, and no remaining value after year 20. Treat all spending as occurring upfront and full output as starting immediately. These assumptions create a standardised financial test; they do not reproduce the project’s phased construction schedule or its contract terms.

A discount rate adjusts future cash for the return required and the wait to receive it. At 8%, twenty annual payments of $1 have a combined present value of approximately $9.82. The question is how much cash the project needs to generate for those payments to justify the initial spending.

Source: Original calculations using the disclosed 890 MW and $4.3 billion investment floor. All cash margins and financial assumptions are illustrative. Net cash means after operating costs, taxes and ongoing maintenance, before financing.

Under this narrow model, annual cash flow of roughly $438 million is needed to recover the investment and meet the assumed return. That works out to approximately $62.42 of net cash per MWh. It is not the required electricity tariff: the tariff would need to cover costs as well, and the actual economics may include other benefits.

The test deliberately excludes additional capacity payments, tax support, benefits to existing output and value beyond the assumed period. It also uses the minimum investment figure, even though the announcement says more than $4.3 billion. Delays, phased spending, ramp-up and cost increases would change the answer; benefits to the wider plants or continued operation beyond the contract could change it in the other direction.

The useful lesson is that a long contract can make a project financeable without automatically making it exceptionally profitable. Duration supplies visibility. The price, costs and delivery determine the return. Investors should ask for those details before treating the reactor programme as a known earnings windfall.

For Google, the equivalent test is different: does the cost of securing power produce enough incremental computing profit and reduce enough disruption risk? The same agreement can be sensible for both companies even though each measures its value differently.

Has Constellation’s Share Price Already Captured the Benefit?

Constellation’s Q2 release raised its full-year 2026 adjusted operating earnings guidance to 11.50-12.50 per share. Using the $12 midpoint gives a straightforward comparison of the valuation before and after the announcement. This uses management’s adjusted measure, rather than GAAP earnings.

Sources: Constellation Q2 2026 earnings guidance; Stock Analysis historical closes; original calculations. This is a current-year guidance multiple, not a consensus forward P/E.

The unchanged earnings denominator matters. The stock’s immediate rise reflects a higher valuation relative to that earnings base. It does not demonstrate a 12.25% increase in this year’s profit, and the announced upgrade timeline gives no basis for expecting the full new output immediately.

Our assessment is that the agreement strengthens Constellation’s long-term business case while raising the evidence needed to justify its share price. An investor paying a higher multiple is placing more weight on future projects, delivery and contract economics. That can be reasonable, but the burden of proof has increased.

A simple future scenario shows the risk. If adjusted EPS eventually reached $15, a 25% increase from the $12 reference point, a valuation of 20 times earnings would produce a $300 share price. At 25 times, it would produce $375. Neither figure is a forecast or target; the example shows that strong earnings growth can coexist with a disappointing share-price outcome when the valuation multiple falls.

For Constellation Energy stock, the sharper question is therefore how much earnings growth the current valuation requires. Contract announcements are useful evidence, but returns on construction spending and cash generation are stronger evidence.

Google needs a different valuation check. Its Q2 earnings release reported $9.11 of EPS and approximately $98 billion of other income, primarily unrealised equity gains. Annualising that headline EPS would confuse investment revaluations with recurring operating earnings.

When assessing Google stock, focus on sustainable operating profit, infrastructure costs and cash generated per share. A sensible electricity agreement improves an input to those economics. It cannot, by itself, establish an attractive valuation for the entire company.

The Risks That Could Change the Outcome

The most important risks are the ones that break the connection between an announced agreement and returns for shareholders. They differ across the two companies, even though both depend on the same physical delivery.

Source: Analytical risk framework based on the announced structure, NRC uprate requirements and PJM’s large-load proposals.

There is also a concentration risk for portfolios. Combining a technology company with a power generator can look diversified because they sit in different sectors. If the investment case for both depends heavily on the same AI infrastructure expansion, their underlying risks may be more closely connected than the sector labels suggest.

Our view is that the agreement is strategically sound for Google and commercially important for Constellation. Google strengthens an essential input to growth; Constellation gains another route to monetising existing infrastructure and expanding output. The more direct deal beneficiary is Constellation, while Google’s potential benefit comes through the productivity of its wider computing business.

The decisive evidence from here is concrete: upgrade delivery against the schedule, returns against the capital spent, disclosed contribution from the contracts, and Alphabet’s conversion of additional capacity into recurring profit and cash. The opportunity is credible. The price paid for exposure still needs its own analysis.