The 100 MW threshold defines entry into this class, placing both of Google’s proposed Minnesota facilities and Meta’s Rosemount facility within its scope

Decision Lens

Minnesota has structurally separated large data center customers from residential and commercial ratepayers under state law, requiring service agreements that cover full electric service costs, system upgrades, and stranded-cost risk. The first real test arrives when Xcel Energy files its agreement with Google for the Rochester-area facility in the coming weeks. The critical unresolved issue is not the framework itself—it passed—but whether Xcel’s “incremental cost test” accurately captures actual costs or leaves regulatory and financial gaps that advocates, the state AG, and data center operators will all contest differently. How that dispute resolves will define the replicable version of this model.

90-Second Brief

Today, minnesota enacted a law establishing a “very large” utility customer class, primarily data centers drawing 100 MW or more, whose service agreements must cover all grid costs they generate. Xcel Energy is preparing to file the first such agreement for a Google data center near Rochester, with Minnesota PUC approval of the governing tariff framework expected around mid-May 2026. Advocacy groups, the state attorney general, and the data center industry have each raised distinct objections to Xcel’s cost-forecasting methodology. The outcome will determine whether this becomes a workable national model or an extended regulatory dispute.

What’s Actually Happening

The Minnesota law, signed by Governor Tim Walz, created a regulatory mechanism that does not exist in most U.S. states: a discrete customer class with its own tariff structure, designed so that data center growth cannot cross-subsidize or be cross-subsidized by ordinary ratepayers. The 100 MW threshold defines entry into this class, placing both of Google’s proposed Minnesota facilities and Meta’s Rosemount facility within its scope.

Xcel’s tariff proposal operationalizes this through layered financial obligations: a minimum 15-year contract term, upfront financial collateral, a floor of 75% contracted capacity payment even if actual load is lower, mandatory direct connection to the transmission network, and an exit fee for early termination. Data centers willing to curtail during high-demand periods qualify for rate discounts reflecting avoided grid stress.

The mechanism under dispute is Xcel’s “incremental cost test”—the formula that determines whether data centers are paying their actual versus projected share. Minnesota’s Attorney General described the forecasting approach as “opaque,” raising concern that hypothetical cost projections could understate the real burden. Google, by contrast, wants the commission to bypass the test entirely in favor of standard rate cases with periodic true-ups. The Data Center Coalition supports cost-based rates but warns that overbroad forecasts could assign costs to data centers that belong to other customer classes.

Why It Matters for Global Heads of Data Center Energy?

This proceeding is a direct operational precedent. If the tariff structure survives its first stress test intact, operators in other states will face similar regimes. Oregon has already enacted comparable ratepayer protections. Virginia is actively considering cost-shift legislation. Georgia’s November regulatory elections signaled voter appetite for tighter oversight of utility-data center arrangements. The direction of regulatory travel is clear; what varies is how cost allocation is calculated.

For energy procurement leaders, the Xcel-Google structure introduces several contract dimensions not standard in most PPA or utility service negotiations: a 75% take-or-pay capacity floor, an exit fee provision, transmission-direct interconnection requirements, and a Clean Energy Accelerator Charge that bundles nearly 2 GW of carbon-free resource procurement—including what Xcel describes as the world’s largest battery project by GWh capacity announced to date—into the service agreement itself.

That bundled structure is operationally significant. The CFE commitment, the interconnection cost, and the battery storage build are not separate procurement exercises—they are a single contractual obligation. Portfolio managers evaluating similar arrangements in other regulated markets need to assess whether bundled structures of this scale can be negotiated, or whether the Minnesota model required unique political and grid conditions to assemble.

The Forward View

The Minnesota PUC vote on the tariff framework is expected around mid-May 2026. If approved, Xcel’s filed service agreement with Google will be the first real test of how the approved tariff is applied in practice. Advocacy groups have explicitly flagged the risk of a timing gap: if the tariff is not approved before the agreement is filed, a dispute over sequencing could delay or complicate approval.

Longer term, the debate over minimum contract duration remains open. The Citizens Utility Board has argued that contract terms should match the 30-year-plus asset lives of grid infrastructure—double Xcel’s proposed 15-year minimum. If regulators accept that argument in future proceedings, data center operators in Minnesota could face contract horizons that exceed typical hyperscaler planning cycles, with corresponding implications for exit flexibility and capital commitment modeling.

Beyond Minnesota, each state-level proceeding that reaches a tariff decision creates a reference point that regulators elsewhere will cite. The question is not whether cost-isolation frameworks spread—it is how prescriptive they become.

What We’re Uncertain About?

  • Whether Xcel’s incremental cost test will survive regulatory scrutiny. The AG’s office has called the forecasting methodology opaque, and Google has asked to replace it with standard rate cases and periodic true-ups. The PUC has not ruled. Resolution depends on what the commission decides in the mid-May tariff vote and, separately, whether the Xcel-Google agreement triggers a formal challenge to the test’s application.

  • Whether 15-year minimum contract terms will hold or be extended to 30 years. CUB Minnesota’s argument for extending minimum terms to match infrastructure asset lives has not been decided. If regulators accept the 30-year framing, the financial modeling for any hyperscaler entering this class changes materially.

  • How behind-the-meter generation will be treated under cost allocation. Geronimo Power raised a structurally distinct scenario: facilities with onsite solar or storage reduce their grid draw but also reduce utility revenue. How regulators value that trade-off—whether as a cost offset or a revenue gap—will determine whether behind-the-meter strategies remain viable under this framework.

  • Whether the Google-Xcel battery program will open to third-party and customer ownership. The battery deployment is currently utility-owned. Advocacy groups want access for residential and commercial customers. If regulatory pressure forces a shared-ownership model in later phases, it could affect the economics Xcel represented to Google when the deal was structured.

One Question to Bring to Your Team

If Minnesota’s cost-isolation framework—with its take-or-pay floor, exit fee, and 15-to-30-year contract horizon—becomes the regulatory default in states where you hold or plan interconnection positions, which of your current or pipeline agreements would face material renegotiation risk, and do your financial models account for that exposure?

Sources

  • Minnesotareformer — With new law in effect, data centers shouldn’t mean higher electric bills for the rest of us (Link)

Tags