Court Killed the Federal Coal Lifeline — and It Just Reset the AI-Grid Stack

On September 11, 2026, a unanimous three-judge panel of the U.S. Court of Appeals for the D.C. Circuit threw out the Department of Energy's Section 202(c) order that had kept Consumers Energy's J.H. Campbell coal plant running in West Olive, Michigan since May 2025. The court held that the Federal Power Act's emergency provision is a narrow, last-resort tool, and that the agency had invoked it without demonstrating a genuine emergency.

The decision is the first loss in court for the Trump administration's broader coal-rescue campaign, which had used the same provision to delay retirements at plants including the Centralia plant in Washington and the F.B. Culley and R.M. Schahfer generating stations in Indiana. Per Consumers Energy's own financial disclosures, keeping Campbell online past its planned May 31, 2025 retirement had already cost ratepayers roughly $259 million.

The Section 202(c) tool was the only federal mechanism that could force a generator to keep operating against its owner's wishes. Every major U.S. utility with a coal retirement on its books had been quietly pricing in the option that, if a regional grid crunch got bad enough, Washington could extend a unit's life by fiat — absorbing the cost through ratepayer or DOE channels rather than utility earnings.

The D.C. Circuit ruling makes that option materially harder to invoke. Subsequent orders will face a binding precedent that 'emergency' means emergency, and that preventing a planned, regulator-approved retirement is not one. For coal-plant owners who were already losing money on marginal units, the ruling removes the federal backstop that propped up the late-life optionality thesis.

Michigan sits in MISO, not PJM, so the ruling's most direct price impact is on the MISO Planning Resource Auction. The 2026/2027 PRA cleared summer capacity at $424.30/MW-day across Zones 1–7 — including Michigan — down sharply from the $666.50/MW-day record in 2025/2026. New solar and delayed retirements pushed available capacity up roughly 4% to 142.6 GW, easing the supply tightness that drove last year's spike.

But the structural tension is unresolved. PJM's 2028/2029 Base Residual Auction cleared at the FERC-approved $325/MW-day cap for the entire footprint, with coal's UCAP share down 2,941 MW on retirements and conversions. The J.H. Campbell ruling hardens the binding cap on capacity prices rather than loosening them — utilities and large-load customers will pay the cap, not a free-market clearing price, because policymakers no longer have a credible emergency lever.

The court ruling lands in the middle of an unprecedented data-center build-out in Michigan. DTE Energy lifted its five-year capital plan 22% to $36.5B for 2026–2030, anchored to a hyperscale pipeline of more than 8 GW — headlined by the 1.4 GW Oracle/OpenAI Saline project cleared by the Michigan Public Service Commission and a separate 1 GW Google campus near Detroit. Consumers Energy, meanwhile, is contracting up to 1 GW of new load for a hyperscaler and is part of a separate $7B project announcement tied to tech giants.

Sister utility Consumers Energy had been counting on the J.H. Campbell extension as a bridge while new generation came online; the ruling forces the retirement on schedule and pushes an extra ~1,420 MW of capacity into MISO's already-tight summer market. With MISO summer capacity at $424.30/MW-day and total data-center load in the state projected to grow from 0.2 GW in 2025 to 1.2–2.3 GW by 2030, the order shifts the cost of the AI-grid build squarely onto ratepayers via capacity charges and DTE/CMS capex.

Upstream coal producers are the clearest losers. Peabody Energy has been leaning into U.S. thermal coal as AI-driven demand re-prices Powder River Basin volumes, but every coal plant that closes on schedule is a permanent demand loss. Alliance Resource Partners, the second-largest eastern U.S. producer, similarly gains from data-center load growth but loses when retirements accelerate without offset. Coal-bridge optionality — the idea that AI demand would indefinitely extend coal-plant lives — just lost its strongest policy premise.

The winners are the nuclear-adjacent, behind-the-meter, and greenfield-data-center plays. Constellation Energy, with 32.4 GW of mostly nuclear capacity and a $715M Rhode Island State Energy Center acquisition, sells 24/7 carbon-free power to data-center customers at premium contract terms. Talen Energy is already running a co-located data-center campus at Susquehanna and expanded its Amazon PPA to 1,920 MWe in June 2025. Vistra, with 22.3 GW in PJM, signed Meta to 20-year PPAs supporting three nuclear plants and is in active talks to expand power for data centers through nuclear and gas options.

The cleanest read-through is to the NextEra Energy/Brookfield Asset Management $100 billion Paducah, Kentucky data-center campus, announced in July 2026 with 4–6 GW of dedicated new generation and 1.8 GW of utility capacity targeted by 2032. The deal's economics depend on building new clean generation behind a federal-site anchor; the J.H. Campbell ruling makes that 'build new' model the de facto template for every AI data-center power story that follows.

Over a one-to-three-year horizon, the ruling accelerates three structural shifts that were already underway. First, late-life coal valuation compresses further: investors can no longer underwrite a Section 202(c) rescue into a coal-plant DCF, which tightens the bid-ask between owners and potential data-center offtakers hoping to convert sites. Second, MISO and PJM capacity prices stay capped rather than market-clearing — utilities and large-load customers carry the bill, but the upside for merchant generators narrows. Third, the AI-grid capex story at DTE Energy and Consumers Energy becomes the supply response that coal was never going to be — every dollar of the $36.5B DTE plan and the CMS Energy hyperscaler capex now needs to clear without a federal coal backstop.

The deeper test is whether the AI demand growth rate keeps absorbing the cost. Michigan data-center load is projected at 1.2–2.3 GW by 2030, and national data-center demand is forecast to triple from 2023 levels. If those forecasts hold, the NextEra Energy/Brookfield Paducah model, the Talen Energy-Amazon co-location template, and the Constellation Energy nuclear-PPA architecture all clear on their economics. If demand softens, the DTE Energy $36.5B capex program is the name most exposed to rate-case pushback — the Michigan AG has already intervened on data-center tariffs.

The D.C. Circuit ruling is the inflection that re-rates the U.S. AI-power stack. Until September 11, every late-life coal plant had an implicit federal put option priced into its equity — the belief that a sufficiently tight grid would trigger Section 202(c). That option just expired. The cost was always paid in ratepayer money and stranded retirement plans; J.H. Campbell alone ran up $259M before the court intervened.

What replaces it is a cleaner two-track market. Track one: utilities like DTE Energy and Consumers Energy must build new generation to serve the 8+ GW of contracted Michigan data-center load, and they will recover the cost through MISO capacity charges ($424.30/MW-day summer) and rate cases. Track two: independent power producers — NextEra Energy, Brookfield Asset Management, Constellation Energy, Vistra, Talen Energy — capture the premium for new, behind-the-meter, or nuclear generation paired directly with hyperscale customers. Upstream coal producers (Peabody Energy, Alliance Resource Partners) lose a permanent demand layer every time a forced retirement becomes a clean one.

Fact versus inference: the ruling itself, its date, the Section 202(c) theory, and the $259M figure are all verified primary-source facts. The interpretation — that the ruling removes a priced-in tail-risk hedge for late-life coal — is inference supported by how the option was being modeled in utility IRPs and coal-equity research, not a directly cited claim. Speculation, not inference: that DOE will appeal successfully, or that the Saline and Google projects will clear Michigan regulatory scrutiny on schedule. Those are binary, dated catalysts to watch.