Three gigawatts vanished from the PJM grid in minutes on July 22. Not a generation failure — a demand collapse. Northern Virginia's data centers, sensing a transmission fault, switched to backup power simultaneously, and the resulting voltage disturbance rippled from Washington to Chicago. It took ten minutes to fully stabilize. Ten minutes. On a grid that normally corrects disturbances in milliseconds.
That event is a preview of what happens when you strip baseload out of a grid faster than you replace it. Coal retirements have been accelerating across PJM's territory for years, and the capacity that's left — the firm, dispatchable megawatts that hold frequency and voltage steady when everything else goes sideways — is increasingly thin. The data center boom didn't create this problem. It exposed it.
The Baseload Gap Is Already Here
PJM serves 67 million Americans across 13 states and the District of Columbia, including the world's largest concentration of data centers. It is also, right now, a grid in genuine distress. Federal regulators convened a technical conference on July 23 to discuss PJM's "power shortfalls and surging electricity prices" — language that would have been unthinkable for America's flagship grid five years ago. FERC Commissioner David LaCerte put it plainly: the board needs to "get a backbone" rather than defer to stakeholders whose conflicting interests slow down the primary work of keeping the grid reliable and affordable.
The structural problem is this: coal retirements remove firm capacity — generation that runs when called, regardless of weather. Wind and solar add nameplate capacity, but nameplate capacity and firm capacity are not the same thing. A solar farm rated at 500 MW delivers that output for a few hours on a clear afternoon. It delivers nothing at 9 PM when a heat dome parks over the Mid-Atlantic. The EIA's April 2026 data shows average retail electricity revenues up 6% nationally year-over-year, with Ohio up 22% and Maryland up 17.1% — states squarely inside PJM's footprint. That's not a coincidence. That's a capacity crunch showing up in consumer bills.
PJM Is Now Threatening to Cut Data Centers Off
The situation has deteriorated fast enough that PJM sent a letter to stakeholders on July 27 warning that data centers failing to secure adequate generation may face involuntary outages during high-demand periods as soon as mid-2027. An emergency power auction is being launched specifically to address the AI demand surge. The biggest grid in the country is telling its largest and fastest-growing customers: we might have to cut you off.
This is what a baseload exit looks like in real time. It's not a theoretical future problem. It's PJM announcing emergency measures while FERC simultaneously debates whether to restructure or potentially break up the grid operator entirely. The Bloomberg reporting on the possible PJM breakup frames it as a Trump administration frustration with the organization's inability to ensure adequate supply at reasonable prices. That's the political surface. The underlying physics is simpler: you can't retire firm capacity without replacing it with firm capacity.
Nuclear Is the Only Firm Capacity That Scales
Here's where the civilizational argument becomes a market argument. The IEA's Electricity 2026 forecast projects global electricity demand growing at an average annual rate of 3.6% through 2030, with advanced economies — including the U.S. — accelerating after 15 years of stagnation. That demand growth is structural: AI compute, EV charging, industrial electrification, heat pumps. It doesn't go away when a coal plant retires. It compounds.
Wind and solar are essential — nobody serious is arguing otherwise — but they cannot solve the firm capacity problem that coal retirements create. Storage helps at the margin. What actually fills the hole is nuclear: dispatchable, carbon-free, high-capacity-factor generation that runs at 90%+ utilization regardless of whether the sun is shining or the wind is blowing. The EIA's nuclear industry data makes the capacity factor advantage clear — no other generation source comes close on a per-unit basis.
The PJM crisis is accelerating this logic from aspiration to necessity. When the grid operator is threatening to involuntarily disconnect data centers, the conversation about nuclear timelines stops being abstract. Every month a new reactor takes to clear permitting, financing, or supply chain is a month PJM runs closer to the edge. The emergency auction buys time. Nuclear buys the future.
Watch for two things in the next 90 days: whether FERC's PJM reform process produces binding capacity requirements that explicitly value firm generation over nameplate capacity, and whether the emergency auction pricing creates a signal strong enough to accelerate advanced nuclear project financing in the Mid-Atlantic. If PJM's capacity prices spike — and they will — the economics of new nuclear in that footprint change overnight.
The grid is telling us what it needs. The question is whether we build it fast enough to matter.
