PJM отключит питание крупным дата-центрам США с 2027 года ради защиты от блэкаутов

The largest power grid operator in the US said it will begin temporarily cutting power to major data centers starting in 2027 to prevent mass blackouts. This was reported on July 28, 2026 by TechCrunch.

What the grid operator decided

The operator of the largest power grid in the US plans to introduce temporary power cuts for large data centers at moments when grid load threatens to trigger rolling blackouts. The measure will take effect starting next year — that is, from 2027.

This is not about a permanent restriction, but about forced load shedding during peak hours: when consumption approaches the limit of generation, the operator will cut off large industrial consumers specifically, not residential homes. Data centers are first in line for cuts as the grid's most energy-intensive facilities.

The public decision to specifically cut off data centers is a notable shift. Previously, rolling restrictions primarily affected factories and manufacturing plants, not the server halls around which the entire artificial intelligence economy is built today.

*Who: the largest power grid operator in the US (PJM Interconnection)

*What: temporary power cuts for large data centers

*Why: to prevent mass blackouts during peak loads

*When: starting in 2027

*Source: TechCrunch, July 28, 2026

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

TechCrunch reports.

Why are data centers overloading the grid?

Data centers have become the largest new consumers of electricity because of the artificial intelligence boom. Training and operating large language models requires tens and hundreds of megawatts per facility — comparable to the consumption of a small city, with such load held around the clock.

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

There's also a timing problem. A new power plant or trunk transmission line takes years to build, while a hyperscale data center can be built in one or two seasons. Because of this gap, demand from AI construction outpaces the grid's physical capacity — and it's easier for the operator to restrict a consumer than to instantly ramp up generation.

What does this mean for the AI industry?

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

The operator's decision is a signal that the power system is not keeping pace with the rate of data center construction. Companies will need to build their own generation, energy storage, or flexible consumption agreements into their projects to avoid depending on emergency grid shutdowns.

What it means

The power grid is becoming a bottleneck for the development of artificial intelligence: computing capacity is now constrained not by chips, but by kilowatts. The readiness of the largest US operator to cut power to data centers shows that the electricity shortage is not a hypothetical risk, but an operational reality that the AI industry will need to account for when planning for 2027 and beyond.