FERC June 18, 2026: AI Cloud Is Now a Power-Queue Business. France Has No Queue.
On June 18, 2026, the Federal Energy Regulatory Commission issued orders to all six regional transmission operators under its jurisdiction — PJM, MISO, SPP, CAISO, ISO-NE, and NYISO — requiring them to justify or reform how they handle large-load interconnection for data centers and manufacturing facilities. Operators have 60 days to respond. The four pillars: Protect Consumers, Enhance Transparency, Safeguard Reliability, Foster Innovation.
The framing from FERC Chair Laura Swett was unambiguous: large technology companies are expected to pay for their own grid upgrades. Socialising those costs onto household ratepayers is not acceptable. A faster queue is coming — but into the same constrained grid.
Scope: PJM · MISO · SPP · CAISO · ISO-NE · NYISO — two-thirds of the US grid
Deadline: 60 days to justify current tariffs or propose reforms
Generation report: 30 days to report on generation adequacy for existing and new large loads
Four pillars:
1. Protect Consumers — prevent cost-shifting via mandatory contracts · smarter studies · Grid Enhancing Technologies
2. Enhance Transparency — transmission cost disclosure · load forecasting · utility investment reporting
3. Safeguard Reliability — new large load impact studies · accelerate interconnection · NERC standards
4. Foster Innovation — accelerate co-located and proximate loads · flexible load · studies in 60–90 days
Key new service type: Interim Network Integration Transmission Service — load and generator get non-firm service immediately, convert to firm when grid upgrades complete. Generator can serve load while awaiting grid connection.
THE NEW AI INFRASTRUCTURE STACK — FERC JUST MADE IT OFFICIAL
The old mental model: more chips + more capital = more AI capacity. The new model has four layers that FERC's order now formalises as a regulatory reality.
Layer 1 — Power access: sites where large power blocks can actually become usable capacity, not just contracted capacity. Layer 2 — Interconnection timing: Berkeley Lab data shows 2,060+ GW of generation and storage seeking US grid connection at end of 2025 — only 13% of capacity requesting interconnection from 2000 to 2019 reached commercial operation by end of 2024. Layer 3 — Cost allocation: FERC is explicitly pushing cost transparency and cost-shifting prevention — large loads pay their own upgrade costs. Layer 4 — Operational flexibility: data centers may need to prove when they can shift, curtail, self-supply, or coordinate with grid conditions.
CoreWeave operates 49 AI data centers across North America and Europe.
End 2023: 10 data centers · ~70 MW active
End 2024: 32 data centers · ~360 MW active
End 2025: 43 data centers · 850+ MW active
Current: 1 GW+ active · 3.5 GW contracted
The gap between 1 GW active and 3.5 GW contracted is not a financial problem. It is a grid problem.
Between those two states: interconnection studies · transmission upgrades · tariffs · local politics · generation adequacy · cooling · procurement · construction timing · customer commitments.
FERC's order addresses the process. It does not create the generation capacity.
WHAT FERC'S ORDER ACTUALLY DOES — AND WHAT IT DOES NOT
The critical distinction: a faster queue is useful. A faster queue into a constrained grid is still a constrained grid. FERC's order creates process clarity — faster studies, cost transparency, co-location rules, interim service options. It does not create new generation capacity. It does not add substations. It does not build transmission lines. Those take years regardless of regulatory direction.
The new Interim Network Integration Transmission Service is genuinely useful: it allows a load and a co-located generator to get non-firm service immediately, then convert to firm service when grid upgrades complete. This is the regulatory formalisation of the "bring your own generation" model that hyperscalers have been implementing informally with on-site gas turbines, fuel cells, and solar. FERC is not blessing it reluctantly — it is accelerating it by design.
The Electrically Proximate Load provision goes further: studying a generator and nearby load together, within two electrical buses, reduces the required infrastructure and timeline for both. This is the co-location model that PJM has been litigating for two years, now formalised as a supported path across all six operators.
THE FRANCE COMPARISON — NO QUEUE, NO COST UNCERTAINTY
FERC's four-pillar framework is a blueprint for what French grid governance already provides — but arrived at through regulatory crisis rather than design.
Protect Consumers — Cost allocation clarity
US post-FERC: mandatory contracts · smarter studies · cost-shifting prevention · still being defined
France RTE: connection costs published · split between RTE and developer by type of works · no retroactive reallocation
Enhance Transparency — Cost and timing visibility
US post-FERC: 60-day response deadline · load forecasting improvement · utility disclosure
France RTE: published schema directeur · connection offer with defined cost range · 12–24 month deterministic timeline
Safeguard Reliability — Large load impact studies
US post-FERC: new large load impact study framework · NERC standards development
France RTE: HTB feasibility study (3–6 months) · detailed study (4–8 months) · sequenced, defined durations
Foster Innovation — Speed to power
US post-FERC: interim non-firm service · co-location · proximate load · 60–90 day study target
France RTE: fast-track programme · 5 sites · 250 MW in 2 years confirmed · brownfield reactivation 3–9 months
Key difference: FERC is building toward what RTE already has. France's framework is not aspirational — it is operational today.
The most important number from FERC's order for France context: 60–90 days for studies as a target. RTE's feasibility study runs 3–6 months, detailed study 4–8 months — longer in absolute terms, but with a critical difference: they are sequential, defined, and the outcome is a binding connection offer with published cost ranges. There is no equivalent in the US system of a "connection offer" with defined cost and timeline that a developer can rely on for investment committee approval.
WHAT THIS MEANS FOR AI INFRASTRUCTURE STRATEGY
FERC's order signals that the US market is entering a phase of regulatory clarification — which is necessary but not sufficient. Clarity on process does not eliminate the physical constraint. The 2,060 GW queue is not a queue of paperwork. It is a queue of physical infrastructure that needs to be built, permitted, and energised.
For developers and infrastructure funds with deployment targets in 2026–2028, the strategic implication is direct: the FERC order extends the timeline for US capacity to become usable, not shortens it. Regulatory reform takes 12–24 months to translate into operational change. Physical infrastructure takes 3–7 years. The window for France's advantage — 4,800 MW fast-track, brownfield HTB at 18 months, EU transformers at 20–32 months — does not narrow because FERC issued an order. It narrows because the fast-track sites get allocated.
The right read of FERC June 18 is not "US grid is getting fixed." It is "US grid complexity has been officially acknowledged at the highest regulatory level." That acknowledgement is valuable for investors who needed regulatory certainty before committing to European alternatives. The signal is now clear: the US queue is real, structural, and will take years to resolve. France's 18-month brownfield timeline is not a temporary anomaly. It is a structural advantage that FERC's order just made more legible by comparison.
FRANCE HAS NO POWER QUEUE. YOUR NEXT DEPLOYMENT DOESN'T HAVE TO EITHER.
GridReadiness provides grid connection advisory, transformer procurement, and brownfield site identification for AI data center deployment in France. 30-minute framing call. No commitment.
→ Related: FERC ruling preview — our prior analysis · $55M Brookfield FERC filing · France site selection guide
Sources: FERC orders June 18, 2026 (ferc.gov) · FERC Commissioner Rosner infographics · DCmag June 19, 2026 · "FERC Makes AI Cloud A Power-Queue Business" analysis · CoreWeave 2025 Form 10-K · Berkeley Lab Interconnection Queue data 2025 · RTE 2026 · GridReadiness field intelligence.