Oracle's Wisconsin nuclear payment buys political cover, not new power

Oracle will absorb roughly $300 million in rising power costs tied to Wisconsin's only nuclear plant, in a deal announced October 2 that adds no generating capacity to the grid and instead buys down a rate increase utility regulators were already set to approve. The structure, built around an existing power purchase agreement rather than new construction, offers a template other hyperscalers may use to secure scarce nuclear output while neutralizing the local political backlash that has followed data center buildouts elsewhere.

We Energies buys roughly 1 gigawatt from NextEra Energy's Point Beach plant under a contract whose per-megawatt-hour cost has nearly doubled over a decade, from $45.94 in 2016 to $75.51 this year, and is projected to hit $122.45 by 2033, according to figures in the utility's SEC filings. That single contract already accounts for about 20% of We Energies' proposed 2027 rate increase, adding an estimated $176 million in service costs and $59 million in fuel costs next year alone, even though Point Beach supplies only about a third of the utility's capacity.

A subscription, not a supply addition

Under the arrangement, Oracle is committing to cover 125 to 250 megawatts of the output We Energies already buys from Point Beach — roughly 12% to 25% of that allocation — to power its $15 billion Project Lighthouse data center campus in Port Washington, built with Vantage Data Centers and tied to OpenAI's compute needs. The mechanism differs from the new-generation deals hyperscalers have favored elsewhere: Oracle is not funding a reactor uprate or a new plant, it is paying to keep an existing, aging merchant asset's rising fuel costs off the bills of roughly a million retail customers who get none of the data center's output.

By absorbing approximately $300 million in rising energy costs, we are helping protect Wisconsin ratepayers, said Mahesh Thiagarajan, executive vice president of Oracle Cloud Infrastructure.

The rate-case math that forced the deal

The timing lines up with a contract extension: We Energies recently pushed its Point Beach purchase agreement out from 2030/2033 expirations to 2050 (Unit 1) and 2053 (Unit 2), locking in decades of a cost curve state consumer advocates had already flagged as a rate-case liability. Oracle's payment is projected to generate close to $300 million in customer savings from 2027 through 2033 — almost exactly offsetting the contract's near-term cost escalation, which suggests the figure was negotiated backward from the utility's rate filings rather than from any market price for the power itself.

It's definitely a step forward, said Tom Content, executive director of Wisconsin's Citizens Utility Board, adding, we would just want to know more about how significant the relief is.

Why supply-chain vendors should track the mechanism, not just the megawatts

For power-equipment and grid vendors, the deal matters less for its capacity figure than for what it signals about how utilities intend to monetize data center demand without adding firm generation. We Energies says it will backfill Oracle's committed share with excess capacity and cheaper market purchases, a claim Content says remains unverified in public filings. If replacement power costs more than projected, the arrangement could reset the economics other merchant nuclear operators use when pitching similar subscription deals near aging, cost-constrained plants rather than greenfield builds — a cheaper, faster path to locking up baseload power than the multi-year interconnection queues facing new gas turbines or reactor uprates.

This agreement will not impact customers' reliability, said Brendan Conway, a We Energies spokesman.

The arrangement still needs sign-off from the Public Service Commission of Wisconsin, and the political backdrop is unfavorable to easy approval: a Marquette Law School poll found 74% of surveyed Wisconsin voters believe data center costs outweigh the benefits, and a draft PSC report has projected statewide peak demand will rise 40% over five years, driven largely by data center clusters in Port Washington, Mount Pleasant and Beaver Dam. Oracle is betting that paying down a rate increase buys more regulatory goodwill than building new power ever could — a calculation other hyperscalers facing similar interconnection and public-opinion constraints are likely to test next.