Google's Nuclear Uprate Tariff Puts AI's Power Bill on Georgia's Rates
Georgia Power and Google have agreed on a way to expand nuclear output in Georgia without pouring concrete for a new reactor. Under a filing submitted to the Georgia Public Service Commission, the utility's customers would see about $900 million in projected benefits while Google underwrites uprates at the Vogtle and Hatch plants that add roughly 96 megawatts to the grid.
The agreement was announced September 21, 2026, and centers on a new nuclear uprate tariff. Google would subscribe to that tariff and receive Zero-Emission Credits in exchange. Georgia Power, a Southern Company subsidiary, says the structure is subject to PSC approval before anything takes effect.
The headline number is the $900 million. The mechanism behind it carries more weight for anyone watching how AI data centers get powered.
What the Nuclear Uprate Tariff Does
An uprate is not a new plant. It is incremental capacity drawn from reactors already running, produced by upgrading components and re-licensing them at higher output. Georgia Power's filing covers an extended power uprate (EPU) request for Units 1 and 2 at Plant Hatch, paired with the NU-1 tariff that sets out how Google pays for the resulting megawatts.
Vogtle is further along. An EPU for Units 1 and 2 at that site was approved in the 2025 Integrated Resource Plan, so the current filing extends an approach regulators have already accepted rather than opening a fresh case from zero.
| Element | Detail |
|---|---|
| Parties | Georgia Power (Southern Company) and Google |
| Sites | Plant Vogtle and Plant Hatch, Georgia |
| Capacity added | Approximately 96 MW |
| Projected customer benefit | About $900 million over the life of the units |
| Mechanism | Nuclear Uprate (NU-1) tariff subscription; Zero-Emission Credits |
| Regulatory status | Pending Georgia Public Service Commission approval |
| Uprate status | Hatch 1 and 2 EPU requested; Vogtle 1 and 2 EPU approved in the 2025 IRP |
The tariff is the novel piece. Traditionally, the cost of an uprate lands in the utility's rate base and is recovered from all customers, who receive the added capacity in return. Georgia Power's proposal inverts part of that arrangement: a single corporate buyer subscribes to the incremental output up front, and the projected customer benefit flows back to the general ratepayer base over the life of the units.
The distinction between an uprate and a new build matters for cost. Adding capacity at an existing site reuses switchyards, cooling systems and licensed operating staff, so the capital outlay is far smaller than a new reactor demands. It also avoids the land acquisition and new-site permitting that stretch greenfield nuclear timelines.
Google's role goes beyond a standard power purchase. The company is financing a capacity expansion and taking Zero-Emission Credits as part of the return, so its payment tracks an attribute of the generation as well as the electrons delivered.
Zero-Emission Credits are the second half of the compensation. They represent the environmental attribute of the generation rather than the electricity itself, and they give Google a claim tied to nuclear output. For a buyer that needs carbon-free energy around the clock, that attribute carries weight alongside the megawatt count, because it is what shows up in clean energy accounting.
Because the credits are separable from the electricity, the two buyers do not compete for the same product. Google holds the environmental attribute while Georgia Power's other customers continue to draw the physical output from the same reactors. The tariff is what makes that split explicit and billable, and it is the reason the deal had to run through a regulatory filing rather than a private contract.
Why a Tariff Instead of a Contract
Placing the arrangement inside a regulated tariff changes who bears what. A tariff is approved by the PSC, published, and applied to a defined class of customer. That gives the utility a recoverable revenue stream and gives regulators leverage over terms, while making the deal a matter of public record and public argument in a way a private contract would not.
For Google, the appeal is speed and predictability. New nuclear construction in the United States has a record of long schedules and cost escalation. Uprates at operating plants skip the riskiest phase of that process: site preparation, licensing of a new facility, and first-of-a-kind construction. The 96 megawatts arrive far sooner than a greenfield reactor could deliver them.
The trade-off is scale. Ninety-six megawatts is a meaningful addition to a state grid, but it is small next to the load a large AI data center campus draws. The deal reads best as one building block in a portfolio rather than a standalone answer to hyperscale demand.
The Trade-offs Hiding in the Fine Print
Three questions will decide whether this becomes a template or a one-off.
First, the tariff design. The filing does not settle, at least publicly, how long Google's subscription runs, how the payment escalates, or what happens if the company's power needs in Georgia change. A tariff that locks in a corporate payment stream for decades protects ratepayers. A tariff that lets the buyer exit early leaves them holding the cost.
Second, the cost allocation. Georgia Power projects roughly $900 million in customer benefits over the life of the units. That figure is a projection, and it rests on assumptions about output, operating costs, and the value assigned to the added capacity. The PSC will test those assumptions, and the outcome determines whether ratepayers are genuinely better off or simply less exposed.
The benefit is also spread thin over time. Nuclear plants operate on multi-decade timelines, so the $900 million accrues across the life of the units and depends on the reactors staying online and running at expected output for years. Google's subscription payments would begin far earlier than the customer benefit fully materializes.
Third, precedent. If the Georgia PSC approves the nuclear uprate tariff, other utilities with operating reactors gain a working template to bring to their own commissions. Regulators elsewhere would weigh the same question: should a single large customer be allowed to buy into the rate structure to accelerate capacity that everyone else also relies on?
The timing sharpens that question. Georgia is among the first states where this mechanism has reached a formal docket, which means the PSC's record will be the first detailed public account of how an uprate tariff is priced and allocated. Other commissions will read it closely.
The Verdict
Georgia Power gains a customer willing to fund capacity it would otherwise have to justify in a general rate case. Google gains carbon-free megawatts and credits without owning a reactor. Ratepayers gain a projected benefit they did not have to finance. Each side trades something real: the utility trades some control over its rate base, Google trades flexibility, and ratepayers trade a share of the upside if the projections prove conservative.
The structure is attractive because it sidesteps the two options that have dominated AI power procurement so far, which are building new generation or signing long-term contracts for it. Uprates sit between those poles. They use existing infrastructure, existing licenses and existing sites, and they let a corporate buyer absorb the incremental cost.
What remains unresolved is who carries the risk if the demand justifying the deal does not materialize. Google's data center buildout in Georgia is the implicit collateral. The PSC's review is the first real test of whether that is enough security for a regulated tariff.
Why this matters
The nuclear uprate tariff moves a slice of AI's energy bill out of the corporate capital budget and into a regulated utility structure, which means state commissioners rather than a single company's procurement team set the terms. For anyone tracking how data center demand gets paid for, Georgia is now the test case. If the model holds, utilities with operating reactors will copy it; if the commission trims the terms, hyperscaler appetite for this route gets measured against the alternatives. The next concrete milestone is the PSC's ruling on the filing.
Related Articles
- Washington's $1.9B bet on the Duane Arnold nuclear restart
- Google's Finland AI investment: €13B turns Nordic power into Europe's largest AI program
- Google adds $1.5B for Alabama data center AI push
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.