Wisconsin's 40% Load Spike: Three Data Centers, Redacted Contracts, and the Ratepayer Bill

The Wisconsin Public Service Commission released a draft strategic energy assessment projecting peak electricity demand will climb from 14.2 gigawatts in 2026 to more than 20 gigawatts by 2032, a 40% increase ^[1][2]. That alone would demand significant infrastructure. But the real story lives in the footnote: 72% of that spike, or roughly 4.17 gigawatts, is attributable to three hyperscale data center projects in Mount Pleasant (Microsoft, 2.6 GW), Port Washington (Oracle and OpenAI, 1.3 GW), and Beaver Dam (Meta, $1 billion facility) ^[3][5]. We Energies will serve Microsoft and Oracle; Alliant Energy will serve Meta.

What the PSC assessment does not disclose: the terms of the special contracts binding these three customers to the utilities, the cost allocation for new generation and transmission, whether ratepayers bear the risk of underperformance, and the exit fees if the load fails to materialize. These are commercial negotiations filed under confidentiality claims. The cascade is familiar. A utility accepts a hyperscaler's demand projection, uses it to justify massive capex (new gas plants, transmission upgrades), files that capex in a rate case citing "reliability" and "grid stability," and spreads the cost across all ratepayers while the tech company's contract often contains weak demand ratchets, short terms relative to asset life, or collateral insufficient to cover stranded investment. The forecast becomes self-fulfilling: the utility gets its capex approved, and residential and small-business customers fund infrastructure built to serve a single customer class whose actual draw remains unknown.

The questions unanswered by the PSC report: Are these 4.17 gigawatts contracted firm capacity, or announced capacity the developers hope to build? Has the PSC netted out duplicate interconnection requests shopped across utilities? What is the historical realization rate of the utilities' own prior load forecasts? The Wisconsin Policy Forum notes that since 2005, Wisconsin's total energy sales have fallen 9% and peak demand has dipped 2.6%, meaning the state's utilities have systematically overestimated growth for two decades ^[7]. Yet this 40% projection, concentrated in three customers, is being treated as gospel for planning purposes.

The utilities' response is an "all of the above" generation strategy: natural gas, solar, wind, and battery storage. Translation: natural gas is the fallback, and if the data-center load does not arrive on schedule or at full capacity, ratepayers will carry the stranded gas-plant debt. We Energies operates the Elm Road plant, recently transitioned from coal to natural gas ^[3], a sunk cost that anchors the utility's preference for firm, dispatchable load. Hyperscalers, by contrast, can operate flexibly; they tolerate curtailment. The utility could offer a tiered interconnection: bring-your-own-generation (BYOC) or flexible grid access (curtailable at 0.25% to 1% of annual hours) in exchange for faster, cheaper interconnection and lower cost to ratepayers. This model, documented in academic research, allows data centers to energize sooner while sparing the system ~$764 million per gigawatt in supply-side capex. It is not being studied here.

The window is closing. Wisconsin's utilities will file rate cases or seek tariff riders citing this PSC projection. Consumer advocates must demand, by name, a large-load tariff with hard protections: minimum-demand ratchets at 85% of contracted capacity, collateral of ~$1.5 million per megawatt, 10-14 year minimum terms matched to asset life, 100% cost responsibility for dedicated network upgrades, and cost isolation so data-center infrastructure does not bleed into residential rates. Virginia's GS-5 tariff and Ohio's large-load rider offer templates. Without such structure, the burden lands on ratepayers, and the socialized cost of tech infrastructure growth becomes the scandal that nobody admits is a scandal.