The nuclear revival has finally produced a deal that gets the sequence right: use existing reactors to deliver clean power now, then build new nuclear where the grid and customers can support it. Google’s agreement with Constellation is real progress, but it is not the SMR breakthrough the market is pretending it is.

The arrangement supports 890 megawatts of additional nuclear capacity across Constellation’s PJM fleet and includes a long-term power purchase agreement. That distinction matters. The capacity comes from upgrades and efficiency improvements at operating plants, not from waiting a decade for a new reactor to clear licensing, construction, and commissioning. Google gets firm, carbon-free electricity for data centers. Constellation gets a customer willing to pay for reliability and nuclear attributes. The grid gets more output from assets that already exist.

This is the most economically rational form of nuclear expansion available in the United States. A reactor that is already licensed, connected, staffed, fueled, and producing electricity does not need to prove nuclear power from scratch. It needs capital for uprates, modernization, and maintenance, plus customers prepared to sign contracts long enough to justify the investment. That is a much lower-risk proposition than funding a first-of-a-kind advanced reactor.

The market’s reaction exposed the industry’s central problem. Shares of NuScale, Oklo, and NANO Nuclear rose sharply even though Google did not buy power from any of them. Those companies benefited from the narrative, not the transaction. NuScale has the strongest regulatory position among them, with an NRC-certified design, but certification is not construction. Oklo remains a developer without an operating commercial reactor. NANO Nuclear is earlier still. Nuclear investors need to stop confusing a hyperscaler’s demand for electrons with demand for a particular reactor vendor.

Google’s deal is important because it puts a value on nuclear’s most underappreciated product: dependable capacity in a constrained power market.

The 890 megawatts associated with the Constellation agreement is roughly equivalent to the output of a large nuclear unit operating at a high capacity factor. The difference is that Constellation can pursue that capacity through uprates and efficiency work rather than a greenfield project. The time value of that distinction is enormous. A modest gain at an existing plant can deliver revenue years before a new reactor produces its first kilowatt-hour.

The Department of Energy has also announced a conditional loan commitment of up to $4.2 billion for Vistra’s upgrades and uprates at Beaver Valley, Davis-Besse, and Perry. That is a direct signal that the federal government sees existing nuclear capacity as infrastructure worth preserving and expanding, not merely as a bridge to advanced reactors. The economics are straightforward. If a plant can add output, extend its useful life, or reduce operating costs without taking on full new-build risk, the resulting electricity can compete far more easily than power from a first-of-a-kind project.

This is where nuclear’s levelized-cost debate often becomes intellectually dishonest. Comparing an existing reactor’s incremental megawatt-hour with a new solar farm, gas plant, or SMR is not an apples-to-apples exercise. The existing reactor has already absorbed much of its capital cost. But comparing a new SMR with an operating plant is equally misleading in the opposite direction. The SMR must pay for licensing, site preparation, manufacturing qualification, financing during construction, commissioning, and first-of-a-kind engineering. Its claimed overnight cost is not its delivered cost.

Google’s decision suggests that the buyer is optimizing for delivered energy, not reactor novelty. Data centers do not need an elegant reactor architecture. They need power that is available when the computing load arrives, with credible carbon accounting and low exposure to gas-price volatility. A long-term PPA with an established nuclear operator meets those requirements immediately.

That does not make SMRs irrelevant. It clarifies their hurdle. An SMR company must offer something an operating nuclear fleet cannot, such as siting flexibility, lower financing risk, industrial heat, remote-grid service, or a standardized build schedule. A smaller reactor that merely promises nuclear power in a different package has not yet earned a premium.

The market should also distinguish between capacity and energy. An agreement to add 890 megawatts does not mean 890 megawatts of new reactor construction. It means additional available output from the PJM fleet, likely through uprates, equipment modernization, and efficiency improvements. That is still valuable, and in a capacity-constrained grid it may be more valuable than a press release announcing a future reactor. But it is not evidence that SMR construction costs have fallen, licensing has accelerated, or first-of-a-kind risk has disappeared.

First, it accelerates the commercial life of the existing US fleet. Constellation and Vistra now have a stronger case for capital spending, license extensions, uprates, and major component replacements. Other operators will ask whether hyperscalers, manufacturers, and large industrial customers will sign similar contracts. The answer will determine how much dormant nuclear capacity becomes economically recoverable.

Second, it strengthens the case for nuclear in PJM. The region faces rising demand from data centers, electrification, and industrial load while transmission construction remains slow. Nuclear plants already located near major load centers have an advantage that no new technology can manufacture quickly: grid connection. A reactor on the system is not just an energy asset. It is a permitted, synchronized source of firm power.

Third, it helps NuScale more than its stock reaction suggests, but for a narrower reason. NuScale’s certified design gives utilities a regulatory starting point that competitors lack. The company still needs a financeable customer, a bankable construction plan, and evidence that its cost structure survives contact with a real project. Google’s deal does not provide those things. It does, however, validate the existence of a premium customer segment willing to pay for firm clean power.

Holtec’s submission of a Preliminary Safety Analysis Report for two SMRs at the Palisades Energy Center is a more meaningful signal for advanced nuclear than the stock-market rally. It is still a regulatory milestone, not a poured foundation. The distinction matters. A completed construction permit application demonstrates that a developer is doing serious engineering. It does not demonstrate that the project has solved financing, supply chain, workforce, or construction risk.

The likely losers are pre-revenue developers that use corporate nuclear announcements as substitutes for customer contracts. Google did not select Oklo, NuScale, or NANO Nuclear to supply this power. It selected an established operator with operating assets. That is the message investors should hear beneath the celebration.

The nuclear industry is heading toward a two-track revival. Existing plants will capture the first wave of demand because they can deliver reliable clean electricity on commercial timescales. New reactors will follow where they can show repeatability, not just technical promise. The winners will be companies that understand the difference between a certified design and a bankable project, between a memorandum and a PPA, and between raising money for nuclear and actually producing it.

Google’s deal is therefore real progress, but not for the reason the market claims. It proves that large customers will pay for nuclear power. The next question is whether reactor developers can convert that willingness to pay into steel, concrete, and operating megawatts.