Constellation Energy’s latest Google agreements turn data-center demand into unusually long contractual visibility. A 15-year arrangement covers 2,700 megawatts of existing power, while a separate 20-year power purchase agreement supports 890 megawatts of new nuclear capacity through uprates at 11 reactors.
The new capacity is not immediate. Constellation expects the first uprate in 2028 and plans to invest more than $4.3 billion. That pairing—long contracts against long-lived assets—is the core attraction. It can reduce merchant power-price exposure and give the company confidence to invest in reactor improvements that would be difficult to justify on spot prices alone.
The economics are stronger than a simple “AI power” label
Nuclear plants carry high fixed costs but relatively low incremental fuel costs. Adding output through an uprate can therefore be economically attractive when licensing and engineering work succeeds: more megawatt-hours are spread across an existing site, workforce and operating platform. A contracted buyer lowers the risk that the added capacity arrives into a weak market.
Constellation also said the agreement includes demand response and load shaping. Those provisions matter because a data center’s headline capacity is not identical to constant grid demand. Flexible load can reduce stress during scarce hours and improve the value of the broader contract.
Capital execution remains the counterweight
The $4.3 billion investment is real before the 890 megawatts are fully operating. Dividing the stated investment by capacity implies about $4.8 million per incremental megawatt. That BTI calculation is a rough project-intensity measure, not a forecast of return, and it excludes the time value of money and any operating changes.
Licensing, outages, engineering scope and inflation can change both the cost and schedule. A delayed uprate postpones contracted output while capital is already committed. The company’s second-quarter adjusted earnings of $2.55 per share show a profitable base business, but the stock’s risk profile increasingly includes large-project execution in addition to commodity prices.
Why the deal can still lower portfolio risk
Long-term agreements with a large counterparty can make future cash flows more predictable. They do not eliminate operational risk, counterparty concentration or regulatory exposure. Investors also need to know pricing escalators, cost-sharing and termination protections, which were not fully disclosed in the announcement.
The appropriate conclusion is neither that nuclear demand removes risk nor that capital spending overwhelms the benefit. The Google contracts improve revenue duration and support incremental nuclear output. The decisive milestones are license approvals, project budgets and the first 2028 uprate; those will show whether contractual visibility becomes attractive realized returns.
There is an additional portfolio question: the new contracts may reduce commodity exposure while increasing counterparty and regulatory concentration. Google is financially strong, but the value of a 20-year agreement depends on enforceable terms across changing market rules, tax treatment and nuclear regulation. Investors should not equate long duration with zero risk.
The employment figures—4,400 existing jobs supported and 7,200 construction jobs—show the political constituency behind the projects. That can help permitting, yet it also raises execution visibility and public scrutiny. Cost overruns at regulated or politically salient energy projects can attract intervention.
Constellation’s 2,700-megawatt supply agreement and 890-megawatt uprate plan should be modeled separately. The first monetizes existing output; the second requires capital and years of work. Combining them into one capacity headline would overstate how much new generation is available now. Progress reports on each reactor uprate will be more informative than aggregate AI-demand commentary.