From spark to fire: data centers are harming the energy supply of US homes, and with AI data centers everything will only get worse

There is new evidence that AI data centers are negatively impacting the supply of electricity to ordinary homes. This increases the risk of fires and damage to household appliances, Bloomberg reports, citing research data from Whisker Labs and DC Byte.

Research has shown that the rise of AI data centers is placing greater strain on the U.S. power grid and impacting the quality of power delivered to millions of consumers, especially in regions with a high concentration of data centers like Virginia. Large data centers lead to the appearance of harmonic oscillations in the electrical network, periodic distortions of frequency and voltage. For ordinary consumers, this means an increased risk of random sparking, which can even lead to a fire. According to Whisker Labs, the increasingly frequent occurrence of such harmonics is an early sign of impending trouble. Research has shown that more than three-quarters of reported problems in US power grids occur within 80 km of large data centers. Moreover, new facilities are often adjacent to large cities.

Image source: Dina Lydia/unsplash.com

A Bloomberg analysis based on Whisker Labs data collected from 1 million sensors in homes showed a strong link between proximity to a data center and the presence of dangerous harmonic vibrations. More than half of the households with the worst quality of supplied energy are located within 30 km of significant data center campuses. More than 3.7 million people live in the most disadvantaged areas in this regard.

The worst situation is for residents of Northern Virginia, which has the largest concentration of data centers in the world, which together consume almost 3 GW. While the national average is only 1.7% of sensors detecting power outages at least once a month, in Loudoun County this figure is four times or more higher, and in Prince William County The peak of the problem is already detected by 12.9% of sensors.

The problems are expected to increase as large data centers come online. It is especially important, according to scientists, to take into account the impact of AI in data centers, since they are the ones that create the greatest load. New data centers can be built in just a year or two, but the development of power grids does not keep pace with their construction. Energy demand has been fairly stable in recent years, but over the next five years it is expected to grow by almost 16%, which without meaningful investment in infrastructure will only make things worse for ordinary users.

Image Source: Bloomberg/Whisker Labs/DC Byte

The US Department of Energy intends to invest $2 billion in 38 energy projects in 42 states to improve grid reliability. But much more funds are required. Rystad Energy warns that $3.1 trillion needs to be invested in energy infrastructure by 2030 to support the integration of renewables and the grid connections of large consumers. In addition, the quality of household supplies may be affected by the mass introduction of electric vehicles and even such a good cause as the expansion of solar and other renewable energy.

In December 2024, it was reported that in 2028, US data centers could account for 12% of the entire country’s energy consumption. Meanwhile, the United States is preparing a plan to build gigawatt-scale AI data centers and power plants on federal lands, and 269 GW of new nuclear reactors will be available to older US nuclear and coal power plants.