Georgia Power's Google Nuclear-Uprate Deal: Hyperscalers Pay for Utility Generation Under a New Tariff Class
Georgia's large-load contracting framework — already the test bed for the OpenAI 3.2 GW contract and the $15/month residential-benefit pledge — just produced a second, structurally different model: a hyperscaler directly funding utility generation upgrades through a new tariff class, with the utility claiming the deal lowers other customers' costs.
The deal (announced Sept 21, 2026, filed at the Georgia PSC the same day). Google will "support uprates on Georgia Power's owned portion of the nuclear units at Plants Vogtle and Hatch that will allow the facilities to add approximately 96 megawatts of new capacity to the electric grid," and its subscription "is expected to enable approximately $900 million in projected benefits to customers over the life of the units" (Georgia Power release). The filings — Dockets 44280 and 56002 — include a new Nuclear Uprate (NU-1) tariff and a request to approve an extended power uprate (EPU) for Hatch 1 & 2 (the Vogtle 1 & 2 EPU was already approved in the 2025 IRP). Georgia Power owns 45.7% of the reactors; the NRC expects Hatch uprate applications in Q2 2027 and Vogtle in Q4 2028.
The structure. Google subscribes to the NU-1 tariff and receives the Zero-Emission Credits (ZECs) — the carbon-free attributes of the uprated output — while the release states the structure "helps protect non-participating customers from incremental costs related to the uprate work1." Georgia Power's Aaron Mitchell framed it as "the latest example of how large-load growth can benefit all electric customers"; Google's Lucia Tian promised "hundreds of millions of dollars in direct customer benefits." This is the "bring your own capacity" principle (cf. PJM's BYONC, PJM's Market Monitor Puts a Number on It: Data Centers = 46% of the Last Four Capacity Auctions' Cost) transplanted into a vertically integrated state: the hyperscaler capitalizes the utility's asset expansion, buys the clean attributes, and the regulator is asked to approve a tariff that makes the math accretive to everyone else.
Why it matters now. The deal was filed three days after Gov. Spanberger's Virginia EO and two days before Maryland's — the same week the governor-level backlash crested — and lands in a state where data-center power costs dominated the September Savannah PSC meeting and two PSC seats are on the November 3 ballot (watch queued). Approval would hand Georgia Power and parent Southern (SO) a replicable template for hyperscaler-funded nuclear uprates; rejection — or conditions rewriting the cost-allocation — would signal the political wind has shifted even in a pro-growth state. Either outcome is a clean read on whether "large-load pays" has become the price of admission everywhere, not just in PJM.
-
An instance of The grid costs of powering AI cannot be socialized onto residential ratepayers. — The hyperscaler directly capitalizes utility generation expansion under a new tariff class explicitly structured to hold non-participating customers harmless — the developer-pays principle transplanted into a vertically integrated state. ↩︎