Megawatt for Megawatt

The Michigan Public Service Commission approved 1,332 MW of battery storage contracts on March 27. Three of the six projects fulfill DTE Electric’s integrated resource plan. The other three exist because Oracle, OpenAI, and Related Digital want to build a $7 billion, 1.4 GW data center on 575 acres of Saline Township farmland.

Oracle’s subsidiary, Green Chile Ventures LLC, must fund the development of 1,383 MW of battery storage over 15 years as a condition of service. DTE retains ownership and operational control. The load addition represents a 25 percent increase in DTE’s total electric demand, the equivalent of connecting a large American city to the grid overnight.

Storage matching as a condition of interconnection. Renewable portfolio standards require utilities to procure clean energy. Capacity markets require generators to offer reliability. Interconnection studies require applicants to fund grid upgrades. Michigan’s order does none of these. It conditions a customer’s right to draw power on that customer’s obligation to fund an equivalent volume of flexible storage capacity. Not a contribution to a general fund. Not an interconnection fee amortized across a rate base. A direct, megawatt-for-megawatt matching requirement.

The MPSC built the approval around mechanisms that prevent the arrangement from becoming a stranded cost. Contracts carry a 19-year minimum term. Green Chile must pay 80 percent of contracted billing demand regardless of actual usage, closing the scenario where a data center scales back and leaves ratepayers covering idle infrastructure. Termination penalties reach up to 10 years of minimum billing demand. DTE bears responsibility for any costs not recoverable from Oracle. Commissioner Katherine Peretick was direct: “If the affordability analysis turns out to be overly optimistic for any reason, DTE bears the responsibility of any extra costs.”

Curtailment priority and a 90-day deadline. Two additional conditions separate this from a standard large-customer contract. First, the data center must accept curtailment before other customers during electricity shortages. Oracle took a lower grid priority than existing ratepayers as the price of faster access. Second, and more consequentially: the MPSC ordered DTE to file a standard rate structure for major power users within 90 days, eliminating future one-off contracts.

That 90-day directive transforms a single deal into a rate class. DTE has disclosed plans to accommodate up to 8.4 GW of total data center demand, nearly doubling the utility’s current load. Michigan has seen publicly announced hyperscale proposals jump from zero to at least 15 since Governor Whitmer signed a 6 percent sales and use tax exemption for such facilities. Each will need a rate structure. Michigan is telling DTE to build one before the next application arrives.

MPSC Chair Dan Scripps framed the decision in national terms: “I would put the contracts that are in front of us today on par or better with any that have been approved in the country.”

PJM explores the same logic from the demand side. The same week, PJM formed the Connect and Manage Senior Task Force to develop a framework allowing loads above 50 MW to interconnect faster by accepting curtailment obligations during grid stress. The task force adapts ERCOT’s generation connect-and-manage model but applies it to demand, with transmission operators allocating curtailment based on contributions to real-time reliability issues. Constellation Energy, Exelon, Google, and the Data Center Coalition are all participating.

Michigan and PJM arrived at the same conclusion through different mechanisms. Michigan requires large loads to fund matching storage. PJM would require large loads to accept curtailment. Both create a category of customer that did not previously exist in utility regulation: the conditional load, whose grid access depends on accepting obligations that standard commercial customers do not bear.

FERC faces an April 2026 deadline to issue a rulemaking on large load co-location and data center interconnection, with 14 guiding principles already outlined. Whatever FERC produces will shape whether storage matching, curtailment acceptance, or some combination becomes the default framework for data center interconnection nationally.

1,332 MW against a contraction year. Wood Mackenzie projects the U.S. storage market will contract 11 percent in 2026 and 8 percent in 2027 after installing a record 18.9 GW in 2025 (a 52 percent increase over 2024). The contraction is concentrated in utility-scale projects facing FEOC compliance challenges and tariff disruption. Michigan’s 1,332 MW operates outside this dynamic entirely. The capacity is contractually obligated, ratepayer-protected, and backed by a customer whose load commitment precedes the storage commitment.

If Virginia, Ohio, and Texas (the states with the deepest data center queues) adopt versions of Michigan’s matching model, storage deployment driven by conditional load agreements could partially offset the utility-scale contraction that Wood Mackenzie forecasts. These are not speculative pipeline entries. A customer has already agreed to pay.

DTE claims the Oracle arrangement will save ratepayers $300 million annually. That figure will face scrutiny over 19 years and across 8.4 GW of potential data center load. The MPSC has built a structure where DTE absorbs the downside if the math does not hold. Whether other commissions will demand the same accountability from their own utilities is the question that follows Michigan’s order out of Lansing.


Sources

Michigan Approves 1,332 MW of BESS with 332 MW Supporting Oracle Data Centre (Energy-Storage.News)

MPSC Goes Big on Batteries (Michigan MPSC)

Michigan Regulators Approve Contract for Michigan’s First Hyperscale Data Center (Bridge Michigan)

PJM Forms Task Force to Explore Large Load Curtailment (RTO Insider)

US Annual Battery Storage Installations Will Not Reach 2025 Levels Again Until 2029 (Energy-Storage.News / Wood Mackenzie)

Federal Regulatory Outlook for Electric Storage, QFs, and Inverter-Based Resources (Morgan Lewis)