FERC has turned the AI power argument into a clock.

The commission said it launched an aggressive, targeted action to speed large-load integration, ordering six grid operators to report within 30 days on available generation capacity and to justify or reform their tariffs within 60 days [1][2]. The fact sheet frames the move as grid efficiency, reliability, and competitiveness [2]. Commissioner David LaCerte's concurrence in the PJM docket makes the narrower implementation point: large-load growth has become a reliability and cost-allocation problem that needs a record, not slogans [3].

That record is the update after the paper's June 18 article on FERC forcing data centers into a state power fight. Friday's order scales the question nationally. It asks grid operators to show where generation exists, how fast large loads can connect, and what tariffs do when a data center arrives like a small city.

The deadline is the news because it disciplines the rhetoric. AI companies can promise infrastructure. Governors can welcome investment. Utilities can warn about reliability. Ratepayers can fear subsidy. None of those positions answers the operational question: what capacity is actually available, who pays for upgrades, and which tariff puts the answer in writing?

FERC's own release casts the action as speed [1]. That is true but incomplete. Speed without cost assignment is a political transfer. The paper's June 18 Maryland story said AI power had become a refund docket, because customers need a forum before costs settle on bills. The new order does not settle that fight. It creates a calendar on which the fight can be documented.

The technology angle is therefore not that data centers need electricity. Everyone already knows that. It is that model growth, cloud contracts, and GPU financing are now pressing into old grid institutions. This FERC action shows the physical side of the boom [1][2]. A backlog is not real capacity. A power request is not a megawatt. A megawatt is not a tariff.

The X/MSM split is familiar. X sees AI ratepayer anger, suspicion of hyperscaler subsidies, and a broad feeling that households are being drafted into someone else's server farm. No verified status URL survived, so the article should not invent one. Mainstream regulatory copy sees interconnection reform and reliability. The paper's useful sentence sits between them: FERC has created a filing clock that can show whether speed becomes a cost shift.

LaCerte's concurrence is important because concurrences are where process anxiety often becomes readable. His PJM concurrence locates the question in docket EL26-67 [3]. The public does not need to love docket numbers. It needs to understand that docket numbers are where AI's abstractions become bills, queues, and legal duties.

The next 30 days should produce names and numbers. Which grid operator says it has available generation? Which says it needs new resources? Which proposes new tariff language? Which state regulator objects? Which data-center customer accepts direct cost responsibility? Those answers will matter more than any ceremonial claim that America is supercharging the grid.

AI's energy appetite has been a metaphor for months. FERC just made it a schedule.

-- DARA OSEI, London