Google’s Power Agreements Put Contracts Behind New Capacity

Google and Constellation Energy announced two long-duration power agreements on October 6 that turn electricity demand into something more consequential than a procurement decision: a financing foundation for new infrastructure.

The headline number is 3,590 megawatts, but it combines two different commitments. A 20-year power purchase agreement supports 890 megawatts of new nuclear capacity from upgrades at 11 existing units. A separate 15-year supply agreement covers another 2,700 megawatts of power in the PJM market. Constellation says the broader arrangement supports more than $4.3 billion of investment.

The judgment

• Implication: A creditworthy customer can finance infrastructure without owning it by giving the supplier durable revenue certainty. The quality of the contract—not just the size of the power commitment—determines how much investment the demand can support.
• What would change the conclusion: The case weakens if upgrade costs rise materially, delivery slips beyond the expected schedule, contract protections shift too much volume or price risk back to Constellation, or the added capacity proves smaller than expected.
• Management action: Separate total contracted supply from incremental capacity, then map contract duration, termination rights, escalation clauses and customer credit against the asset’s investment and operating life.

The 3.59-gigawatt headline contains two different economics

Reuters reported that Google contracted for 3,590 megawatts from Constellation. About one quarter—890 megawatts—comes from upgrades intended to increase output at nuclear plants in Illinois, Pennsylvania and New Jersey. Constellation expects the first upgraded plant to begin delivering electricity in 2028.

The remaining 2,700 megawatts sit in a 15-year supply agreement that is not tied to a single generation source. Reuters described that contract as long-term revenue certainty for Constellation’s operating fleet.

Those commitments should not be treated as interchangeable. The 890-megawatt agreement underwrites incremental capacity and carries execution risk: equipment must be installed, approvals must be secured and output must be delivered. The 2,700-megawatt agreement monetizes existing portfolio capability and supports cash-flow visibility. One helps finance expansion. The other stabilizes the base.

A power contract can function like project finance

The company announcement says the 20-year agreement will enable investment in equipment and technology at 11 nuclear units. The commercial logic resembles other contracted infrastructure: a long-lived asset becomes more financeable when a strong counterparty commits to purchase its output for a period that covers a meaningful portion of the investment horizon.

That does not mean the customer eliminates the supplier’s risk. Constellation still faces construction, operating, regulatory and performance exposure. Nor does it mean Google has acquired a power plant. The contract instead allocates enough demand and price certainty to make capital investment more credible.

This distinction matters for CFOs evaluating any capacity agreement. A lease, take-or-pay contract, minimum purchase commitment or long-term service agreement can move risk without moving legal ownership. Finance teams should therefore evaluate the economic obligation, not just the accounting label.

Duration is valuable only when the risks are matched

A 20-year commitment can support major investment because it reduces uncertainty about future sales. It can also create a mismatch if the customer’s demand forecast, technology strategy or location needs change faster than the contract.

The board-level questions are straightforward:

• What portion of the payment is fixed, indexed or exposed to market prices?
• Who bears cost overruns and schedule delays?
• What performance level must the supplier deliver?
• Can the buyer reduce volume, relocate demand or terminate early?
• What happens if the customer’s credit quality changes?
• Does the contracted term match the useful life and financing period of the investment?

These questions apply well beyond power. The earlier Numbers & Judgment analysis of Project Jupiter’s power delay showed how a missing operating dependency can become a financing problem. The analysis of Nscale’s contract book made the related point that customer commitments become revenue only when financing, equipment, construction and operations come together.

The capacity bridge belongs in the capital model

Management should not present a single “megawatts secured” number without a bridge. At minimum, the model should separate current supply, contracted existing supply, incremental capacity under construction, capacity awaiting regulatory approval and uncontracted forecast demand.

The same bridge should connect capital spending to cash flow. For each project or contract, finance should track committed capital, expected in-service date, contracted volume, expected margin, downside cost, counterparty concentration and termination exposure. A deal can improve visibility and still concentrate risk.

The free Forecast Scenario Planner can structure the base case alongside slower delivery, higher cost and lower utilization outcomes. The point is not to predict one precise result. It is to identify which assumptions determine whether the investment still earns its return.

What to watch next

The most important evidence will not be the share-price reaction or the aggregate power headline. It will be the timing and cost of the 890 megawatts of incremental capacity, the contract protections behind the $4.3 billion investment plan and the extent to which new supply keeps pace with data-center demand on the PJM grid.

Numbers & Judgment’s analysis is that long-term demand can be a financing asset. But it becomes one only when the contract is bankable, the project is executable and the risk allocation survives a downside case.

Sources

• Constellation Energy, “Google and Constellation Announce Landmark Agreement to Bring 890 MW of New Nuclear Capacity to PJM Grid,” October 6, 2026
• Reuters, “Google enters massive 3.6-GW power deal with Constellation Energy,” October 6, 2026

Reported facts about the agreements, investment amount, capacity and expected timing are attributed above. The conclusions about contract finance, risk allocation and management controls are Numbers & Judgment analysis.