A correction first, because the record matters more than the narrative. DTE Energy’s 1.4-gigawatt data center deal was not revealed during mid-October 2026 results. It was announced on October 30, 2025, alongside the company’s third-quarter 2025 earnings, when DTE said it had finalized its first data center contract with a hyperscaler it did not name. DTE has not yet reported third-quarter 2026 results. What makes the deal worth revisiting now is everything the past year has filled in: the customer’s identity, the site, the contract terms as approved by Michigan regulators, construction under way, and a second hyperscaler deal that turns a single transaction into a template.
That is the right way to read this story. A year ago it was a headline — 1.4 gigawatts, unnamed customer, ramp over two to three years. Today it is a case study in how a regulated Midwest utility structures, prices, and defends a hyperscale load. DTE now reports 2.4 gigawatts of executed data center agreements and another 5 to 6 gigawatts in its development pipeline. The 1.4-gigawatt deal is the one that started it.
The deal as announced
The October 30, 2025 release was spare. DTE said it had finalized its first data center contract with a hyperscaler for 1.4 gigawatts of load, ramping up over the next two to three years. The new demand would be supported with existing capacity and new energy storage investments “which will be paid for by the data center.” DTE said the growth would create substantial affordability benefits for existing customers as it sells excess generation, and that contract terms would ensure the data center absorbs all new costs required to serve it.
On the earnings call that morning, president and CEO Joi Harris added the structure. DTE would serve the customer under its existing industrial tariff and combine it with an energy storage contract to support the incremental storage investment. The company planned nearly $2 billion of incremental energy storage investment, plus additional tolling agreements, to support the data center load. The key protections were a 19-year power supply contract with minimum monthly charges, and a 15-year energy storage contract through which the data center would fund its own storage needs. DTE said it would file for regulatory approval the next day, and that storage investments would begin ramping in 2026 to align with the projected load increase.
The pipeline figures Harris gave that day framed the stakes. Beyond the 1.4 gigawatts, DTE was in advanced discussions with additional hyperscalers for more than 3 gigawatts of new load, with a further 3 to 4 gigawatts behind that in an earlier-stage pipeline. Generation investments for those additional opportunities could arrive in the back end of the company’s five-year plan — incremental capital above what DTE was laying out that day.
What “existing capacity plus storage” actually means
The structure deserves a slower read, because it is the mechanism that makes the whole deal work — and the part most easily misread as a utility getting 1.4 gigawatts for free. DTE’s claim, in essence, is that it has generation capacity sitting underused for parts of the year, and that a large, steady new customer lets it sell more of what it already owns. Spreading the fixed costs of that existing fleet across more kilowatt-hours is the “affordability benefit” in DTE’s telling: the same plants, more sales, lower average cost per unit for everyone.
But existing capacity alone cannot cover a 1.4-gigawatt customer that expects power around the clock, including during the hours when DTE’s own system is tightest. That is where the storage comes in. Batteries charge when the system has surplus and discharge when the data center’s demand coincides with system peaks. Storage generates no new energy; it moves energy across hours, shaving the peaks that would otherwise force DTE to build or buy new firm generation. For a flat, always-on load, that time-shifting is the difference between “we have the energy” and “we have the capacity when it counts.”
The cost-allocation choice is the part that matters for other customers. DTE did not fold the storage into its general rate base and spread it across all ratepayers. It put the storage in a separate 15-year contract paid for by the data center. The customer that causes the incremental investment pays for it; existing customers, in theory, get the benefit of fuller use of the shared fleet without the bill for the new hardware.
There are limits to the logic, and they are worth stating plainly. Selling excess generation only helps existing customers if the generation is genuinely excess — if DTE’s fleet has headroom after serving its existing load through peaks and outages. If the data center’s arrival accelerates the day DTE needs new firm generation, the affordability math gets murkier. And a 15-year storage contract protects ratepayers only if the counterparty performs and the collateral holds. The structure is sound on paper; the test is the ramp, which is still ahead.
The name emerges: Oracle in Saline Township
Within months of the announcement, the unnamed hyperscaler had a name. The counterparty is Green Chile Ventures LLC, a subsidiary of Oracle, and the site is Saline Township in Washtenaw County, where a roughly $7 billion campus — about 2.2 million square feet on a 575-acre site, with additional acreage in nearby Bridgewater Township — is being developed alongside OpenAI and Related Digital. The 1.4-gigawatt figure is the contracted load DTE agreed to supply, not a statement about the campus’s computing capacity: contract demand is what the utility plans around.
On December 18, 2025, the Michigan Public Service Commission approved the Saline contracts unanimously, without a contested hearing, and attached conditions that hardened the customer-pays structure DTE had described. The primary supply agreement runs roughly 19 years, with an option for the customer to extend it by 20 years, against a standard five-year term for large new customers under DTE’s general large-load tariff. The minimum billing demand is 80 percent, versus 50 to 60 percent under the standard tariff — the data center pays for at least 80 percent of its contracted demand whether or not it uses it. And if the project fails to pay, DTE — not its other customers — absorbs the unrecovered costs. The storage agreement runs 15 years.
The conditions also pushed the framework forward. The commission required DTE to file a generally applicable data center tariff supported by six cost-of-service studies, to be reviewed in a contested case. The Saline approval was not just a green light for one project; it was the start of a standing regulatory framework for the pipeline behind it.
By the time DTE reported second-quarter 2026 earnings, construction on the Saline project had begun, with demand expected to ramp in the 2027–2028 timeframe, and DTE put the project’s capital requirements at up to $2 billion.
The second deal: Google in Van Buren Township
The pipeline behind Saline did not stay a pipeline for long. On October 1, 2026, the commission unanimously approved a second set of DTE data center contracts — this time for Google’s planned 1-gigawatt campus in Van Buren Township, known as Project Cannoli, a nearly 1.5-million-square-foot campus on about 282 acres near Haggerty Road and Interstate 94, being developed by Panattoni for Google.
The Google proceeding was the more searching of the two. Where Saline was approved without a contested hearing, Google’s contracts went through a contested case involving commission staff, the Michigan attorney general, and environmental and energy-sector groups. The outcome was a 20-year Primary Supply Agreement and a Clean Capacity Accelerator Agreement. Service is expected to begin in December 2027, reaching maximum contracted load of about 1 gigawatt by December 2028.
The terms rhyme with Saline’s, with additions that reflect the fuller review. Google’s minimum billing demand is 80 percent of contracted capacity, and the agreement includes an early termination payment equivalent to at least 15 years of minimum monthly charges, plus credit and collateral requirements. The clean capacity agreement requires DTE to develop, own, and operate up to 1,600 megawatts of renewable generation and 480 megawatts of battery storage, with Google paying the related fixed costs. DTE also described about 350 megawatts of demand response in the project’s resource mix, and roughly 700 megawatts of longer-term generation to be identified through its resource planning process. DTE estimated the arrangement could provide a $1.7 billion benefit over the 20-year term by having Google cover fixed costs of resources that can also support the broader system. The attorney general had sought a stronger 90 percent minimum billing requirement; the commission held at 80. And notably, the approval bars construction of new or expanded fossil-fuel plants for the project.
Why a regulated Midwest utility landing hyperscalers matters
Michigan is not Northern Virginia, and that is the point. The first wave of hyperscale load went where the fiber, the land, and the power already were. The second wave is going where utilities will build for it — and where regulators will write rules that let them. DTE’s pipeline, at 2.4 gigawatts executed and 5 to 6 gigawatts in development, puts a single Midwest utility in territory that would have seemed implausible five years ago. The company raised its five-year capital plan by 22 percent, to $36.5 billion from $30 billion, with data center growth the primary driver, and outlined an $11 billion distribution investment plan over five years aimed at cutting power outages by 30 percent and halving outage time by 2029.
The comparison that clarifies DTE’s approach is Black Hills’ Google deal in Cheyenne, Wyoming, announced the same month as the Google approval in Michigan. Black Hills structured its deal largely outside the regulated model — building generation through a non-regulated affiliate, managing a privately run microgrid of third-party resources, and earning contracted fees. DTE’s deals sit inside the regulated model: tariff service, commission-approved special contracts, utility-owned resources, and cost-of-service proceedings. Both approaches put the incremental cost on the customer. They differ in who owns the assets, who approves the terms, and where the risk lands if the customer leaves. Black Hills bets on contract; DTE bets on regulation. Michigan’s commission, by approving two deals with escalating scrutiny and writing a standing tariff requirement, is betting that regulation can move fast enough to keep up.
There is also the competitive geography. Utilities across the Midwest and Southeast are courting the same handful of hyperscalers — TVA with its capacity-charge tariff, Duke with new large-load rules, PJM territories wrestling with phantom load in their queues. DTE’s pitch is distinctive in one respect: near-term load served from existing capacity plus customer-funded storage, with new generation deferred to the back end of the plan. If the storage performs and the ramp holds, DTE gets load growth without an immediate building spree. If the pipeline converts faster than the storage can be built, the deferred generation question arrives early, and the commission’s next resource plan becomes the arena where it gets decided.
What to watch
The milestones, in order: the Saline storage build ramping through 2026 and into 2027; the Saline load ramp in 2027–2028; Google’s first service in December 2027 and full load by December 2028; DTE’s generally applicable data center tariff filing and its contested review; and the next agreement. As of its second-quarter 2026 report, DTE said it had about 2 gigawatts in advanced discussions and was targeting another agreement by the end of 2026, with a further 3 to 4 gigawatts in earlier pipeline stages. That target is an expectation, not a certainty — advanced discussions, land positions, zoning pathways, and community backing notwithstanding, deals at this scale close when they close.
Two nearer-term dates belong on the calendar. DTE is expected to report third-quarter 2026 results in early November, which will update the pipeline figures and test whether the year-end target still holds. And state environmental regulators have scheduled a hearing on a proposed permit for treated wastewater discharge from the Saline campus into the Saline River, a reminder that power contracts are only one of the approvals a campus of this size needs.
The bottom line
A year ago, DTE announced a 1.4-gigawatt deal with a hyperscaler it would not name, served from existing capacity plus storage the customer would pay for. A year later, the customer is Oracle, the site is Saline Township, the contracts carry 19-year terms, 80 percent minimum bills, and commission conditions that put unrecovered costs on DTE rather than other ratepayers — and a second, larger-reviewed Google deal has followed the same template. The structure DTE pioneered in that first announcement — customer pays for the incremental, long tenor, minimums, collateral, storage before new steel — has become Michigan’s de facto standard for hyperscale load.
Whether it reads, in 2028, as the deal that let Michigan absorb gigawatts without raising everyone else’s bills will depend on the things no contract can guarantee: that the storage gets built on schedule, that the load ramps as projected, that the affordability benefits survive contact with the next rate case, and that the pipeline converts at a pace the grid can serve. The template is written. The test is the ramp.