PJM Plans Temporary Power Cuts for Big Data Centers as Grid Struggles to Keep Up

The operator of the largest power grid in the United States will begin allowing temporary electricity cuts to very large data centers in June 2027 if demand spikes threaten reliability. PJM Interconnection is taking the step after a capacity auction failed to secure enough new generation, underscoring how rapidly expanding AI and cloud infrastructure is straining the grid.

The policy could mark a turning point for the data center boom. Facilities that consume 50 megawatts or more will be subject to curtailment during shortages, and the companies behind them will be paid for participating, much like long-running industrial demand-response programs. The move reflects a broader scramble by utilities, regulators, and hyperscalers to secure power as electricity use from data centers is forecast to surge over the next decade.

Why PJM is preparing to cut data centers first

PJM says the change is necessary because its latest auction did not attract enough new generating capacity to match growing load, especially from massive digital infrastructure projects. The grid operator covers a sprawling region from Virginia to Illinois and serves about 67 million customers, making stability in its territory a national concern.

What PJM is proposing is not a blanket shutdown of data centers. Instead, it is a targeted demand-response mechanism aimed at the biggest electricity users that can reduce load quickly when conditions tighten. The idea is to avoid blackouts by shaving demand at the moments the system is most vulnerable.

The timing also matters. PJM says the curtailment program will not begin until June 2027, giving operators time to plan, build backup systems, or negotiate power arrangements. In the meantime, the grid operator is running another auction in hopes of adding more supply before the new rules take effect.

How the curtailment program will work

The program will resemble other demand-response arrangements that have existed for decades in the power sector. Large industrial users have long been asked to temporarily reduce consumption when grid conditions are tight, and they typically receive compensation in return.

Advance notice is a core feature of those programs. Depending on how quickly conditions evolve, participants can be warned anywhere from 30 minutes to several days before a curtailment event. That lead time helps operators move workloads, switch to backup power, or otherwise reduce stress on the grid.

PJM has not said every large data center will be cut at the same time, but it has made clear that the new rules will apply only to facilities with a peak demand of at least 50 megawatts. That threshold captures many modern hyperscale sites while excluding smaller enterprise facilities.

What this means for the data center industry

For developers, the policy sends a signal that grid access can no longer be assumed, even in regions eager for investment. Large facilities may increasingly need to secure their own power sources if they want guaranteed uptime during demand emergencies.

That could accelerate two trends already visible in the market: on-site generation and private energy deals. Some operators may pursue dedicated power plants, battery systems, or long-term arrangements with utilities and independent generators. Others may simply expand their use of backup engines to bridge short interruptions.

The cost implications are significant. Backup systems are expensive to install and maintain, and running them is usually much pricier than drawing electricity from the grid. In some cases, though, the risk of a service interruption is even costlier, which is why many operators are willing to pay for resilience.

Why backup generators are back in the spotlight

Diesel generators remain common at many data centers because fuel can be stored on-site and suppliers are widely available. But those systems are increasingly controversial because of their air pollution, especially in densely populated regions.

Federal rules allow diesel backup generators to operate for limited hours in demand-response events, as well as in emergencies and for maintenance. The source material notes that these units can be used for up to 50 hours per year for demand response, or up to 100 hours annually when other permitted uses are included.

That flexibility helps operators keep services online, but it also exposes communities near data centers to more exhaust during periods when the generators are running. The issue is becoming more visible as the scale of the industry grows and more facilities cluster in the same regions.

Industry and grid officials increasingly frame the problem as one of balancing reliability with the pace of data center expansion, rather than choosing between growth and restraint.

How fast is electricity demand from data centers growing?

Very quickly. PJM and other grid operators are facing a dramatic rise in electricity demand tied to cloud computing, AI workloads, and the physical buildout required to support them. According to the source material, data centers are expected to use four times as much electricity by 2035 as they do today.

That forecast helps explain why grid planners are sounding alarms. Data center construction has moved faster than power plants and transmission projects can usually be brought online, creating a mismatch between demand growth and supply expansion.

Unlike many other loads, data centers often concentrate large and continuous demand in a relatively small area. A single campus can consume as much electricity as a small city, making local grid bottlenecks more likely even when broader regional generation is adequate.

What prompted the decision now?

The immediate trigger was PJM’s failed capacity auction, which fell short of securing enough new generation to cover future demand. Capacity markets are designed to ensure enough resources are available when electricity use rises, but they depend on enough sellers coming forward with new power projects or existing assets.

When that mechanism does not keep pace with demand, grid operators often turn to alternatives: demand response, emergency reserves, new transmission, or fast-track generation procurement. PJM’s decision shows that the system is leaning harder on the demand side because supply is not arriving quickly enough.

The operator is also running another auction, suggesting it still hopes to solve the problem through additional generation rather than through repeated curtailments. But the mere fact that temporary cuts are now on the table reflects how severe the imbalance has become.

Why the issue extends beyond PJM

PJM is the most visible example, but it is not the only grid operator confronting a data center wave. Across the country, utilities are revising load forecasts, delaying connections, or requiring more proof that new campuses will not destabilize local systems.

The challenge is especially acute in regions with attractive power markets, ample land, and favorable permitting conditions. Those same advantages have made Virginia and other nearby states magnets for data center development, which in turn puts pressure on substations, transmission lines, and generation reserves.

As the AI buildout expands, the question is no longer whether these facilities need power. It is whether the grid can deliver that power reliably without forcing other customers to pay the price through higher rates, worse service, or both.

Why some operators may build their own power

Because reliability is central to data center economics, many companies are likely to respond by adding self-supply options. On-site generation can give operators more control, and for critical workloads it may be the only practical way to guarantee continuous service during grid stress.

That path is not cheap. Building private power infrastructure requires land, permits, fuel contracts, maintenance crews, and capital spending that can run into the tens or hundreds of millions of dollars. Still, for facilities that cannot tolerate downtime, the cost may be justified.

Some operators may also seek hybrid models that combine grid power, battery storage, diesel or gas backup, and direct contracts with power producers. The result could be a more fragmented energy system in which the largest data centers increasingly resemble self-contained utility customers.

Environmental concerns are rising too

The likely move toward more backup generation has already drawn scrutiny from environmental advocates and local communities. Diesel generators, while dependable, emit pollutants that can worsen air quality and health outcomes, particularly where multiple facilities cluster close together.

This week’s backlash against Vantage Data Centers highlights how contentious the issue has become. The company was criticized for its apparent coordination with Virginia regulators in challenging a report that linked diesel generator use at a 96-megawatt Northern Virginia data center to tens of millions of dollars in possible annual health harms for nearby residents.

That dispute illustrates a broader tension: the digital economy depends on abundant, reliable power, but the easiest backup solutions can carry local environmental costs that are increasingly hard to ignore.

Critics say the expanding use of diesel backup power risks shifting the burden of data center growth onto nearby communities, especially in already burdened areas.

What are the wider market consequences?

PJM has already faced criticism for the way it has managed the influx of new load and the timing of new generation. Independent market watchers have argued that data centers are now a major driver of higher wholesale electricity prices, which have nearly doubled over the last year in PJM’s market.

That price pressure has implications well beyond big tech. Higher wholesale costs can filter through to utilities, businesses, and households, especially if system planners need to spend more to maintain reserves or build faster infrastructure.

In that sense, the curtailment decision is not just about blackouts. It is also about the economics of the grid: who pays for reliability, who gets priority when supply is tight, and how fast the market can absorb a historic surge in electricity demand.

Timeline: how the PJM situation escalated

1. Rapid data center expansion: New AI and cloud facilities add large, continuous load across PJM territory.
2. Capacity concerns grow: Grid planners warn that supply additions are lagging behind demand growth.
3. Recent capacity auction falls short: PJM does not secure enough new generating capacity.
4. Policy shift announced: PJM says it will allow temporary cuts to very large data centers and other major users.
5. Program launch set for June 2027: Operators get time to prepare before curtailments begin.

Who will be affected first?

The first facilities in line are the largest ones: data centers with at least 50 megawatts of capacity. That threshold is high enough to catch the most energy-hungry campuses, including many built for AI training and large-scale cloud services.

Other large users could also be included under the program, although the source emphasizes data centers because they are the fastest-growing and most politically sensitive load on the system. Traditional industrial users have lived with demand-response rules for years, but digital infrastructure is now entering that same conversation at a much larger scale.

For operators, the lesson is clear: future site selection may depend as much on energy strategy as on fiber connectivity, tax incentives, or land costs. Access to power is becoming one of the defining constraints of the AI era.

What happens next for PJM and the grid?

PJM is likely to spend the next year and a half trying to soften the need for forced curtailment through additional auctions, market adjustments, and negotiations with large customers. But even if more generation is added, the broader trend will remain: electricity demand from data centers is rising too quickly for planners to treat it as a niche issue.

That means governments, utilities, and developers may need to revise assumptions about where AI infrastructure can be built and how it will be powered. The current policy could become a template for other regions if data center growth continues to outstrip supply.

For now, the most immediate message from PJM is straightforward. The grid’s biggest users are no longer just customers at the edge of the system — they are central to whether the system can keep running at all.

Key facts at a glance

• PJM Interconnection is the largest U.S. grid operator by geography and customer count.
• It plans to allow temporary power cuts to data centers of 50 megawatts or more.
• The curtailment program is scheduled to begin in June 2027.
• Participants would be compensated, following the standard demand-response model.
• Backup diesel generation may become more common, despite cost and pollution concerns.

As the data center buildout continues, PJM’s move may prove to be an early sign of how the U.S. will manage the power demands of artificial intelligence and cloud computing: not by eliminating growth, but by rationing it when the grid is under stress.