Home / Latest News / Policy & Regulation

Forced Outages Are Up. Capacity Is Short by 6.8 GW. These Are Not Two Problems.

Aug 05, 2026

In late June and again on July 31, two institutions filed papers that describe the same situation from different angles. NERC released its 2026 State of Reliability report, which formally documented large data centers as a source of grid destabilization for the first time. Then PJM filed an emergency backstop capacity auction at FERC, because its capacity market came up 6.8 gigawatts short, with data center demand responsible for the bulk of the gap. One report describes what is already happening to grid operations. The other describes the market scrambling to prevent what is coming. Together they outline a grid that is under pressure in two directions at once.

What NERC Found

NERC’s finding on the operational side is precise. Conventional generator forced-outage rates rose to 9.2 percent in 2025, above the historical norm of 7 to 8 percent. Coal units recorded a 39.8 TWh increase in unavailable energy. Combined-cycle gas units added another 19.1 TWh. The thermal fleet that grid operators lean on when they need to respond to a contingency is delivering less dispatchable capacity, exactly as demand growth accelerates.

At the same time, NERC counted two instances in 2025 where data center load dropped by more than 1,000 megawatts in a single event. That scale of demand departure is equivalent to losing a large generating unit, but it comes from the load side, where no ride-through obligations or performance standards currently apply. NERC is proposing a new “Computational Load Entity” classification that would bring large computational loads under mandatory reliability standards. That process will take years. The loads are being energized now.

The NERC framing is useful: frequency does not care which direction an imbalance comes from. A grid with less dispatchable reserve and more demand-side volatility has less margin for each contingency. Both trends are present at once.

What PJM Filed

PJM’s July 31 filing describes the capacity side of the same story. Its last base capacity auction, for the 2028/29 delivery year, failed to secure enough capacity to meet its 20 percent reserve margin target. That reserve margin is calibrated to ensure no more than one unexpected major outage per decade. The shortfall was 6.8 gigawatts. Data center demand growth is the primary driver. PJM projects that data centers and other large loads in its footprint could grow by 70 gigawatts by 2038.

The backstop auction PJM filed will run September 30 through October 21. It is expected to pay up to $20 billion for new capacity, with a price cap of $555 per megawatt per day, up from $325 in the prior auction. Separately, PJM plans to file rules that would allow it to curtail large loads without their own generation supply when the grid approaches emergency conditions. The Natural Resources Defense Council put it plainly in response to the filing: “Requiring future data centers to bring their own power might be the single best thing they could have done to stabilize the grid.”

The Same Problem

The NERC and PJM reports are describing the same structural condition at different layers. The NERC data shows the dispatchable reserve is thinner: a 9.2 percent forced-outage rate against a 7 to 8 percent norm means fewer megawatts behind every contingency response. The PJM data shows the capacity market cannot fill that gap at the rate demand is arriving. A grid that is simultaneously operating with less reserve and experiencing larger demand-side events, while running 6.8 gigawatts short of its own reliability targets, is not experiencing a temporary stress. It is being asked to do something it was not built to do.

The regulatory responses will take time. NERC’s Computational Load Entity standards process is measured in years. The PJM backstop auction requires FERC approval. The curtailment rules PJM plans to file on August 7 will govern data centers that come online after June 2027. None of this changes the situation for projects being developed today.

Outside Both Failure Modes

Next Century-Power builds power infrastructure that is structurally exempt from what both reports described. Our facilities are islanded, single-tenant microgrids. They are not PJM loads, so they do not participate in the capacity market that is now filing emergency papers at FERC. They do not contribute to the demand-side volatility NERC documented, because a facility that is not grid-connected cannot generate a grid disturbance when it adjusts load. There is no curtailment order that reaches a facility powered by NCP during a PJM grid emergency.

We start at 1 GW and above, gas turbine generation in islanded configurations with four-nines uptime targets and on-site carbon capture. Phase 2 is a 1.2 GW pod of four 300 MW small modular reactors, providing stable, carbon-captured generation at the contract horizons that large infrastructure investment requires. The path from first energization through decades of operation runs through a single power partnership, not through a capacity market that is currently 6.8 gigawatts short of its own reserve target.

The NERC report is a rearview mirror: it measures 2025. The loads driving its sharpest warnings are being connected in 2026 and beyond. The gap between what the grid was designed to absorb and what is being asked of it will not close through the mechanisms described in either report on any timeline that matches the pace of data center development. That gap is the reason a different infrastructure model exists.

Categories

Recent Posts

Newsletter

Get weekly insights delivered to your inbox.