Summary: On August 28, Oklo filed an emergency FERC complaint, docketed EL26-101, asking PJM to restore its withdrawn 750-MW hybrid generation project to the running interconnection study cycle. PJM had withdrawn the application on August 3, citing modeling and stability issues that Oklo contests. Oklo says losing the cycle would add at least 14 months of deposits, site control, and carrying costs, but those impacts remain its allegations.
A nuclear startup, Oklo, carried an emergency complaint to the Federal Energy Regulatory Commission on August 28, docketed EL26-101. What it seeks is an order directing regional grid operator PJM to return the 750 MW generation project to the running interconnection study cycle, queue position intact. The deadline Oklo requested for FERC’s answer was September 4. The weight here is not scale but exposure — the bind at the heart of interconnection now that AI data centers set the pace: a queue that crawls, rules gone old, compute growing by the month.
A 750 MW Hybrid Plant, and an Application Pulled
Known as C01-1735, the project is assembled from 150 MW of advanced nuclear, 300 MW of fuel cells, and 300 MW of natural gas, and belongs to PJM’s first reformed-queue cycle among more than 700 applications. Synchronous machines cover the nuclear and gas portions; the fuel cells reach the grid via inverters — a shape of project the interconnection rules never anticipated. On August 3, PJM pulled the application. Customer and site are blacked out throughout; the one hint is a footnote about a “PJM / Dominion / Oklo Power” kickoff meeting, which sends observers toward Virginia’s data center corridor. Utility Dive draws a parallel between the project’s outline and the 1.2 GW power campus Oklo has planned in Pike County, Ohio. A nuclear supply deal with Meta, signed in January 2026, exists in Oklo’s record, but the filings keep the two transactions apart. American Buildout, meanwhile, catalogues Oklo’s federal record: selection twice under the DOE Reactor Pilot Program, advanced talks in a surplus plutonium program, and, for a microreactor at Eielson Air Force Base in Alaska, an Air Force notice of intent.
A Procedural Fight: Tardy Deficiency Notices, an Abrupt Withdrawal
As Oklo reconstructs it: everything required arrived April 27; May 15 brought PJM’s deficiency notice — six issue categories, 10 business days to answer; June 24 saw five fresh PJM comments land on the portal in one batch, 41 business days after that window closed, with no email alert and a declined scoping call, which Oklo says conflicts with the tariff duty to flag deficiencies without delay; then withdrawal on August 3, over model discrepancies, a dynamic-model report never supplied, and a stability simulation of the full plant that failed. Oklo contests each. The inverters’ “momentary cessation” problem had been fixed and passed unit testing, Oklo says, yet it failed an entire plant — while the tariff and the guidelines impose no facility-wide ride-through requirement. In Oklo’s words, the move is “the capstone to an unjust, unreasonable, and unduly discriminatory review process,” and by its arithmetic a lost cycle costs at least 14 more months of deposits, site control, and carrying costs. The filing calls the project one “of critical national importance,” needed to ease PJM’s regional supply crunch. Jeffrey Shields, a PJM spokesman, declined to discuss the individual case, stressing that close to 90% of the 811 new service requests cleared the latest cycle’s requirements.
Why Data Center Observers Are Watching
PJM’s first reformed cycle also opened on August 3: 715 projects, more than 200 GW in total, with 17.3 GW of it nuclear and reviews expected to take one to two years. Planning horizons at AI data centers are counted in years; a queue can stop a project before it moves. Oklo draws attention because, for the first time, the tension between grid queues and AI load growth has been laid squarely before FERC — reinstatement by September 21, the Phase 1 study getting underway on September 28, a timeline in weeks aimed at a process in years. Whoever the redacted offtaker turns out to be, hyperscalers underwrite nuclear campuses rather than wait their turn — a pattern familiar from Pike County to Virginia.
Where Hybrid Plants Push the Rules Past Their Limit
The outcome settles one question for every developer: is it shrewd packaging to bundle clean baseload and dispatchable capacity into a single project, or procedural self-destruction? Rule-makers owe explicit answers on several fronts — stability modeling where synchronous machines and inverter-based resources share a plant, ride-through requirements, review milestones. A plant split one-third nuclear, one-third fuel cells, and one-third gas is not what PJM’s own process was built to judge. For the utilities and data center developers scouting power for AI loads, September’s handling of this complaint will settle, more plainly than any white paper, whether grid queues can keep up with compute.