PJM to Cut Power to Large US Data Centers From 2027 to Prevent Blackouts

The largest US grid operator said that from 2027 it will begin temporarily cutting power to large data centers to prevent mass blackouts. TechCrunch reported the news on July 28, 2026.

What the grid operator decided

The operator of the largest US power grid plans to introduce temporary power cuts for big data centers at moments when strain on the grid threatens to turn into rolling blackouts. The measure takes effect next year, in 2027.

This is not a permanent restriction but a forced load shed during peak hours: when consumption approaches the limit of generation, the operator will cut off large industrial consumers, not residential homes. Data centers fall under the curtailment first because they are the most power-hungry facilities on the grid.

The public decision to cut data centers specifically is a notable shift. Previously, rolling restrictions hit mainly factories and manufacturing, not the server halls around which the entire AI economy is now being built.

• Who: the largest US grid operator (PJM Interconnection)
• What: temporary power cuts for large data centers
• Why: to prevent mass blackouts during peak demand
• When: starting in 2027
• Source: TechCrunch, July 28, 2026

"The largest US grid operator said it will start cutting power to large data centers to prevent blackouts,"

TechCrunch reports.

Why are data centers straining the grid?

Data centers have become the largest new consumers of electricity because of the AI boom. Training and running large language models requires tens to hundreds of megawatts per facility, comparable to the consumption of a small city, and that load runs around the clock.

The largest US grid operator is PJM Interconnection, which coordinates electricity supply for roughly 65 million people in 13 states and the District of Columbia. Its territory includes Northern Virginia, the world's largest concentration of data centers, known as "Data Center Alley." This is where demand for capacity is growing faster than the operator can bring new generation and transmission lines online.

Timing is also part of the problem. A new power plant or major transmission line takes years to build, while a hyperscale data center can go up in one or two seasons. Because of this gap, demand from AI construction outpaces the physical capabilities of the grid, and it is easier for the operator to curtail a consumer than to instantly add generation.

What does this mean for the AI industry?

Power cuts directly hit the reliability of AI infrastructure, which is supposed to run without interruption. For cloud providers and AI startups, forced load shedding means the risk of downtime for training clusters and inference services on the hottest days, when the grid is most overloaded.

The operator's decision signals that the power system cannot keep pace with the speed of data center construction. Companies will have to build their own generation, energy storage, or flexible-consumption agreements into their projects so they do not depend on emergency cutoffs from the grid.

What this means

The power grid is becoming a bottleneck for AI development: compute capacity is now constrained not by chips but by kilowatts. The willingness of the largest US operator to cut off data centers shows that electricity shortage is not a hypothetical risk but an operational reality that the AI industry will have to factor into planning for 2027 and beyond.