Since its meteoric emergence, artificial intelligence (AI) has disrupted many sectors. However, its energy impact, often underestimated, worries experts. In the United States, this technological revolution is putting a strain on an electricity network already weakened by aging infrastructure.
Growing pressure on the network
The exponential development of artificial intelligence requires massive data centers to store and process gigantic volumes of data. According to Bloomberg, in March 2024 the United States had around 5,381 data centers, or 17 times more than in France. These installations are real energy ogres. For example, a 5,000 m² data center consumes as much electricity as a city of 11,500 inhabitants.
Disturbances, called voltage distortions, intensify near these centers. These energy fluctuations affect nearly 3.7 million American homes, according to Whisker Labs, an organization dedicated to monitoring the power grid. The consequences can be dramatic: overloading of devices, frequent breakdowns, even domestic fires.
Why do these disruptions persist?
America’s energy supply is struggling to keep up. Aging power plants, combined with growing demand (electric vehicles, connected objects), create an explosive cocktail. Aman Joshi, Chief Commercial Officer, Bloom Energy Corp., said: “ No network is designed to handle such intense load fluctuations, especially when they come from multiple data centers simultaneously ».
Additionally, the fluctuating needs of data centers make regulation difficult. Consumption peaks, sawtooth, unbalance the network and increase the risk of blackout. In some areas, like California, targeted outages are becoming the norm to avoid overloading.
Towards an imminent energy crisis?
Faced with this critical situation, several solutions are being considered:
- Massive investments in modern and resilient infrastructure.
- Energy optimization data centers using technologies such as natural cooling or renewable energies.
- Decentralization of the network with microgrids capable of supporting localized loads.
However, these measures require time and resources. The urgency is there: if nothing is done, the United States could face an unprecedented energy crisis, slowing down the development of artificial intelligence and its applications.
| Element | Numbers |
|---|---|
| Number of data centers in the United States | 5 381 (2024) |
| Average consumption of a data center (5,000 m²) | Equivalent to a town of 11,500 inhabitants |
| Homes affected by voltage distortions | 3,7 millions |
| Main risk | Blackouts and house fires |
A reaction? Leave a comment
Did you like this article? Subscribe to our free newsletter for engaging articles, exclusive content and the latest news.
