Google’s agreement to support uprates at Georgia Power’s Vogtle and Hatch plants is more important than another speculative SMR announcement because it puts a real buyer behind existing nuclear capacity. The technology already works, the sites already have operating licenses, and the grid connection already exists. That is what nuclear’s revival looks like when the customer has an actual load to serve, not a press release to sell.

The roughly 96 MWe of additional capacity is modest against Google’s data-center ambitions, but the signal is large. Google is not merely signing a power purchase agreement for future clean electricity. It is helping fund engineering work that can extract more output from reactors already in service. That means better utilization of sunk capital, faster delivery than a greenfield plant, and less exposure to the construction risk that has damaged nuclear economics for a generation.

The important distinction is between **buying nuclear energy** and **making nuclear construction financeable**. Google’s partnership does the first, and perhaps begins to do the second. It does not prove that advanced reactors are commercially ready, and it does not erase the cost and schedule problems associated with new nuclear. But it demonstrates that large technology companies increasingly value firm, carbon-free electricity enough to participate in the asset itself. That is a much harder commitment to fake than an SMR memorandum of understanding.

The 96 MWe figure deserves more attention than it is likely to receive. Uprates are among the few nuclear investments capable of producing new electricity without starting from an empty site. Operators can add capacity through turbine upgrades, generator replacements, improved instrumentation and controls, fuel-cycle changes, and other modifications that increase thermal efficiency or allow the plant to operate closer to its design envelope. The exact package matters, and Georgia Power has not disclosed the full scope or cost, so nobody should treat the figure as guaranteed capacity until the work is completed and the regulator accepts the changes.

Still, the economics are structurally attractive. A nuclear uprate does not need to pay for a reactor island, a new cooling system, a new transmission corridor, a new site, and an entirely new licensing case. Much of the expensive infrastructure is already earning money. That changes the capital problem from “finance a multibillion-dollar megaproject” to “finance a controlled modification to a productive industrial asset.” It also shortens the period during which capital is tied up without revenue.

The value of that power is not captured by comparing the uprate with the cheapest midday solar bid. Nuclear plants produce high-value electricity across the day and through seasons, while hyperscale data centers impose a load profile that is both large and persistent. A marginal megawatt from an existing reactor can displace gas generation, reduce the need for storage or firming, and lower the amount of new transmission required to serve a growing load center. For Google, the relevant comparison is not simply the levelized cost of solar or wind. It is the cost of obtaining dependable, around-the-clock, low-carbon electricity without waiting years for a new grid buildout.

That does not make existing nuclear automatically cheap. Vogtle’s new units demonstrated the opposite, with severe schedule and cost overruns. But the financial lesson is precisely why uprates and refurbishments matter. The nuclear fleet contains valuable infrastructure whose replacement cost is far above its operating cost. Closing an economically viable plant and rebuilding the same firm capacity with intermittent generation, storage, and transmission can be more expensive even when the headline price of renewable energy is lower.

Google’s participation also exposes a weakness in the current clean-energy contracting model. Corporate buyers have spent years purchasing certificates and intermittent generation while relying on the grid, often gas-backed, when demand rises. That model can increase nominal renewable procurement without guaranteeing physical decarbonization. Supporting nuclear uprates moves closer to the thing data centers actually need, reliable power delivered when the servers are running.

The catch is that this remains a partnership, not a disclosed investment case. We do not yet have the cost of the uprates, the ownership structure, the precise allocation of output, or a public schedule for completion. The 96 MWe is therefore a credible development target, not delivered electricity. The progress is real, but the economics still have to survive engineering, licensing, procurement, and execution.

First, it strengthens the case for the existing American fleet. Companies such as Constellation, Southern, Duke Energy, and Entergy have valuable assets that can potentially support uprates, license extensions, power-contract restructuring, or direct arrangements with large industrial customers. The fastest nuclear capacity in the United States is not waiting for an advanced reactor design. It is already operating.

Second, it gives utilities a more credible customer for nuclear investment. Data-center demand has been treated as both a threat and an opportunity. It is a threat when utilities respond with speculative gas plants and unpriced transmission expansion. It is an opportunity when a creditworthy customer helps underwrite firm clean power and accepts that nuclear projects require long-term commitments. Google’s role at Vogtle and Hatch will be watched closely by Microsoft, Amazon, Meta, and other companies facing the same electricity problem.

Third, the agreement raises the standard for SMR developers. Kairos Power, TerraPower, NuScale, GE Vernova and Hitachi, Rolls-Royce SMR, and other vendors are competing for attention in a market increasingly populated by buyers who understand the difference between a reactor concept and delivered electricity. Strategic investment, EPC commitments, fuel qualification, site work, and a path through licensing matter more than another deployment target.

That is why the same news cycle contains a useful warning. ONE Nuclear Energy is moving to Nasdaq after completing a SPAC transaction, while its Louisiana portfolio includes an SMR campus, gas generation, batteries, and a data-center campus. Capital formation is not construction. A public listing can provide money and visibility, but it does not establish a qualified supply chain, an approved design, a financed customer contract, or a date for commercial operation. Investors should treat the listing as a financing event, not a nuclear milestone.

The near-term winners will be companies that can monetize the distinction between nuclear ambition and nuclear delivery. Utilities with operating reactors have the advantage because they possess sites, workforce, grid access, operating data, and regulatory relationships. Vendors with credible construction partners have the next advantage. Developers that combine nuclear with gas and batteries may find customers faster, but they must explain whether nuclear is the core business or the most attractive label in a broader power-development portfolio.

The lesson from Google’s agreement is straightforward: nuclear’s revival is moving from public opinion to load procurement. The decisive customers are no longer asking whether nuclear is politically fashionable. They are asking which megawatts can run continuously, meet carbon targets, and arrive before their electricity demand does. Existing reactors answer that question now. New reactors will win only when they can answer it with the same level of commercial specificity.