
PJM Seeks One-Year Grid Studies for Data Center Loads
PJM wants FERC to halve grid-study timelines, expose data-center upgrade costs and require computational facilities of at least 50 MW to ride through grid disturbances.
- By
- Sophie Lindqvist
- Filed
- Channel
- AI & Compute
- Read
- 4 min read
By Sophie Lindqvist
PJM asked FERC on Monday, Nov. 16, to cut projected study timelines for new generation and large-load transmission service from two years to one, targeting data centers and other major electricity users.
PJM Interconnection, which runs the largest U.S. grid, said neither generation nor transmission is coming online fast enough to match demand growth driven mainly by data centers. It attributes higher electricity prices to the resulting supply shortfall. No independent queue or price data accompanies that assertion, so its scale remains unverified.
What does Fast Energization change?
Fast Energization would combine generation and load studies now handled separately, removing duplicate work. Quick initial screening rounds would give developers and customers usable information before they post readiness and financial requirements, which PJM says would apply earlier and more strictly. The stated aim is to limit the queue to projects ready to build.
The proposal retains the “first-ready, first-served” principle of PJM’s reformed interconnection process. That priority rule favors ready projects, but the one-year timeline remains a projection rather than a guaranteed outcome for every developer.
New transmission services would cover large loads with onsite generation and flexible loads able to curtail demand. PJM cites FERC’s finding that services should reflect some large loads’ willingness to limit grid use, including interim service while network upgrades are built. The filing tracks FERC directives in two dockets covering large loads and co-located loads, with generation beside a data center.
Can states recover data-center grid costs?
Under the plan, states would gain transparent access to transmission costs driven by data centers and could then bill large loads directly for those costs. PJM says the mechanism is “empowering states to bill Large Loads for those costs.” The proposal as described gives no upgrade-cost estimate, allocation formula or enforcement timetable, so the commercial benefit remains unquantified.
PJM says separate studies now obscure costs created by data-center development. Combined studies could expose those costs earlier, although the available details do not show how individual states would calculate or collect them. No state, consumer group or data-center developer had published a reaction to the Nov. 16 filings by Tuesday.
What must FERC decide, and when?
PJM requests a 60-day comment period and a ruling within 120 days, with its accompanying schedule placing the requested FERC order in mid-March of 2027. These dates are procedural targets, not an approval or implementation schedule.
- Nov. 16: PJM files Fast Energization with FERC.
- 60-day comment period: PJM asks states, developers and other stakeholders to review the tariff changes.
- Jan. 16, 2027: Existing large computational loads would begin a one-year compliance period under a separate filing.
- Mid-March 2027: PJM asks FERC to rule within 120 days of filing.
PJM said it would keep discussing the filing with stakeholders and could amend the proposal. The filing link was not yet on PJM’s filings page on Tuesday morning, limiting immediate review of the tariff language. FERC may also reject or reshape the request; PJM has no final decision in hand.
Why do data centers need ride-through rules?
In a separate Nov. 16 filing, PJM proposed voltage and frequency ride-through requirements for large computational facilities of at least 50 MW. In practical terms, those sites would need to remain connected through disturbances instead of transferring the load immediately to onsite generation.
On July 22, nearly 3.9 GW of data centers unexpectedly switched from utility power to onsite generators during a routine transmission disturbance, the largest such load transfer in PJM’s history. The event affected voltages, frequency and power flows in the Dominion zone and across the Eastern Interconnection.
New large computational loads would have to prove compliance as a condition of connecting. Existing loads would receive one year from the Jan. 16, 2027 effective date, or from commercial operation if that date falls later. The available details do not specify technical ride-through tolerances or test procedures.
PJM serves about 10 GW of this large computational load today and forecasts up to 30 GW more by 2030. The 30 GW figure is PJM’s forecast, not a confirmed addition or a guarantee that transmission and generation can serve it.
Who would pay for the grid response?
PJM’s separate Reliability Backstop Procurement plan would buy up to 15 years of new capacity against a 6,831 MW shortfall from its 2028/2029 capacity auction, with costs charged to large loads driving demand. An amended filing submitted Oct. 29 aims to start procurement on March 1, 2027.
Congressional action proceeds in parallel: the Senate’s permitting bill carries a rule requiring data centers to pay for their grid upgrades, while Exelon has asked FERC to eliminate PJM’s three-year-ahead capacity auctions. Neither the procurement amendment nor Exelon’s request changes the capacity market now. PJM has not published reactions from states, consumer advocates or data-center developers to Monday’s filings.
The next test is FERC’s approval or rejection. Approval could shorten some large-load studies toward one year, but PJM’s supply gap and price pressure would persist without faster delivery of generation and transmission.
Source: mlq.ai
More from Sophie Lindqvist
Show full bio
News editor covering money, funding and market shifts around Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,068 articles
Related articles
us-data-center-disputes-contrast-with-china-s-rapid-build-pace-2a609d6f
US Data Center Disputes Contrast With China's Rapid Build Pace
fugaku-supercomputer-breached-data-of-over-4-000-users-stolen
Fugaku Supercomputer Breached: Data of Over 4,000 Users Stolen
nato-consumes-up-to-300-billion-tokens-a-month-of-openai-cyber-models
NATO Consumes Up to 300 Billion Tokens a Month of OpenAI Cyber Models
sarvam-says-token-factory-serves-more-than-450-billion-tokens-a-day
Sarvam Says Token Factory Serves More Than 450 Billion Tokens a Day


