FERC Blocked Mandatory 'Bring Your Own Power' for Data Centers. Google Signed the Voluntary Version Anyway, at 10x the Size.

Nine days ago, federal regulators told PJM, the nation's largest grid operator, that it could not force new data centers to supply their own grid capacity or risk being cut off first during emergencies. On October 22, Google signed up for a voluntary version of exactly that idea — and made it ten times bigger than the version it signed in June.

The grid operator that can't say yes

PJM's capacity costs have risen more than eightfold in recent years, driven largely by data center demand that the grid operator expects to add another 30 to 34 gigawatts of load by the early 2030s, on top of a current system peak of roughly 162.6 gigawatts. In August, PJM asked the Federal Energy Regulatory Commission to let it connect new 'large loads' — its term for big data centers — ahead of new generation, in exchange for curtailing those loads first when the system is stressed, on compensation terms PJM itself would set. FERC's October 9 order rejected that core proposal along with a related requirement that new loads 'bring their own new capacity.' Regulators kept only a 50-megawatt definition of a large load and a registry tracking where those loads sit and how fast they ramp up. The commission's objection was jurisdictional: it found PJM's plan effectively dictated retail rates and curtailment terms for customers, a power that belongs to the states, not a grid operator.

Same idea, no mandate

What FERC blocked as a utility-run mandate is, commercially, close to what Google has been buying on its own since June, when it hired demand-response firm Voltus to assemble up to 100 megawatts of batteries, smart thermostats and other small devices into a single 'Bring Your Own Capacity' package inside PJM. On October 22, the companies said they were scaling that up by more than tenfold, to what Voltus CEO Dana Guernsey called 'gigawatt scale' — though the release's own numbers (an additional 1,070 MW on top of the original 100 MW) add up to roughly 1.17 GW, not the clean 1 GW both companies used in their headline. Guernsey told Latitude Media the first 100 MW was 'always the first bite'; she told Data Center Knowledge it was oversubscribed. The new tranche runs on contracts through 2040, far longer than PJM's own capacity auction, which locks in supply only three years out, and longer than most demand-response deals. Behind-the-meter commercial and industrial batteries will make up 'a chunk' of the total, Guernsey said, with some residential batteries supplied through Sunrun; neither company has disclosed pricing or the exact asset mix.

What's actually new here

Google frames the arrangement as adding capacity without construction: no new gas turbines, no multi-year interconnection queue, no combustion on site. That's a real contrast with how other data center developers have been chasing power lately — leasing diesel and gas generators to skip the queue entirely, or betting on new grid hardware to speed connections. But Voltus and Google haven't said which specific Google data centers the new megawatts are meant to offset, and neither the June pilot nor the October expansion comes with public performance data showing the capacity holds up during an actual grid emergency — the real test for any demand-response program. Jacques Beaudry-Losique of grid-tech firm GridSparq raised a related concern when the pilot launched: roughly 37 gigawatts of backup generators already sit behind data centers and businesses nationally, and if a 'bring your own capacity' model ends up leaning on those generators rather than batteries, local air quality — not just grid reliability — is at stake. He also asked a question regulators still haven't answered: when a virtual power plant's distributed assets need new wires or substation upgrades to connect, does Google pay, or does the cost land on everyone else's electricity bill.

The pitch to everyone else

Former DOE loan-programs chief Jigar Shah has made the opposite argument — that a hyperscaler funding accredited capacity should be treated 'more like a grid asset than a grid burden.' Voltus, which says it already manages more than 9 GW of demand-response and virtual-power-plant capacity nationwide and wants to reach 20 GW by 2030, is betting the Google deal becomes a template other hyperscalers copy. Its own framing of the stakes is blunt: the real test, the company has said, is proving distributed capacity can work at '10-, 15-, 20-gigawatt levels' within the next three to five years — an order of magnitude beyond anything a virtual power plant has delivered so far, and still short of the new demand data centers alone are adding to PJM.