TECH

A measure adopted in the United States highlights the impact of AI on the power grid

The expansion of artificial intelligence is often associated with technological advances, innovation, and increased productivity. However, behind the most sophisticated models lies an infrastructure that consumes vast amounts of energy. This reality became apparent during a recent heatwave in the United States, when authorities had to implement an emergency measure to prevent grid overloads, reigniting the debate over the environmental impacts of the new AI era.

Extreme temperatures across the eastern United States placed the country's largest power grid under severe strain. With millions of households running air conditioners simultaneously, electricity consumption neared historic highs.

Faced with this situation, the U.S. Department of Energy authorized an exceptional measure to maintain system stability.

The authorization allowed PJM Interconnection—the grid operator serving approximately 67 million people across parts of the Midwest and the U.S. East Coast—to tap into emergency generators owned by major energy consumers, such as data centers and industrial facilities.

These units typically remain idle, used only in the event of a power outage.

This time, however, they served as an additional power source to relieve the load on the main grid.

The authorization took effect as a temporary measure during the period of highest overload risk.

According to the operator, demand reached approximately 162,700 megawatts, very close to the region's all-time record.

Experts point out that the rapid growth of data centers is playing an increasingly significant role in this rise in energy consumption.

The expansion of artificial intelligence requires massive computing infrastructure capable of continuous operation, making these facilities some of the country's largest electricity consumers.

Data centers have become an essential—and problematic—part of the electrical infrastructure... At first glance, using the data centers' own generators seems like a highly logical solution. If these facilities can generate their own power during a supply outage, utilizing that capacity during critical moments helps relieve pressure on the public grid.

However, there is a significant aspect to this strategy.

Many of these emergency systems run on diesel or natural gas engines.

While highly effective at ensuring operational continuity, they also significantly increase pollutant emissions—precisely on days when air quality is often already compromised by intense heat.

In states like Virginia—considered a global hub for data centers—thousands of these generators are located near urban areas.

Recent studies indicate that the frequent operation of this equipment can increase emissions of nitrogen oxides, fine particulate matter, and other pollutants linked to the worsening of respiratory and cardiovascular diseases.

Researchers also link these emissions to a rise in asthma attacks and other health issues among populations living near these facilities.

Thus, the solution devised to prevent blackouts ends up creating another significant environmental challenge.

The growth of artificial intelligence also necessitates a new discussion about energy.

This situation highlights a transformation that often goes unnoticed.

Data centers have evolved beyond mere technical facilities for storing internet data.

Today, they form the physical foundation supporting cloud computing services, digital platforms, and—above all—large-scale artificial intelligence models.

As new AI applications emerge across virtually every economic sector, the energy required to power this infrastructure grows as well.

At the same time, factors such as more intense heatwaves, aging electrical grids, and increased electrification place further strain on distribution systems.

US authorities emphasize that the authorization granted was temporary and necessary to prevent supply interruptions during a critical period. Indeed, a major blackout amidst extreme temperatures could put hospitals, homes, public services, and millions of people at risk.

Several recent, aggressive measures highlight the U.S. power grid's vulnerability to surging AI and data center energy demands. The most immediate step was an emergency directive issued by the U.S. Department of Energy (DOE) authorizing grid operators to force large-scale AI data centers to disconnect from utility lines and run on backup power during brutal summer heatwaves.

The rapid integration of generative AI training models and commercial computing warehouses is driving residential and commercial power usage to record highs. Because of the severity of this crisis, several major, permanent, and long-term regulatory actions are currently rolling out:

Emergency backup order: During record-breaking heatwaves in the Eastern U.S., the DOE ordered dense clusters of AI data centers (particularly in Northern Virginia's "Data Center Alley") to switch to backup diesel or natural gas generators. This emergency measure successfully freed up critical megawatts to keep household air conditioners running.

FERC interconnection overhaul: The Federal Energy Regulatory Commission (FERC) issued sweeping orders forcing six regional grid operators to revamp their connection rules. This FERC Action provides a fast-track "priority lane" so energy-hungry data centers can plug into the electric grid more quickly, while establishing clear standards and timelines for transmission costs.

DOE power capacity commitment: The DOE previously committed $3 billion specifically for AI-centric smart grid programs to modernize grid operations and enable predictive analytics to better balance supply and demand.

mundophone