The intense competition over artificial intelligence is reshaping the US power grid. With data center electricity demand expected to nearly triple by 2030, technology companies are investing heavily in new power plants and even building their own private electricity networks.
AI has a serious energy problem. As AI companies struggle to secure electricity for their continually expanding data centers, the technology is reshaping the US power grid.
According to the US Energy Information Administration (EIA), nationwide electricity demand, which had been stagnant for more than a decade, has grown by an annual average of 1.7% since 2020. The EIA attributes this growth primarily to the rapid expansion of data centers during that period.
Analysis by S&P Global suggests that this trend is expected to accelerate further. The organization estimated that these facilities’ demand on the grid would increase by 22% by the end of 2025 and nearly triple by 2030.
Data centers have long consumed large amounts of electricity, but the scale and speed of AI deployment have brought their size and consumption to unprecedented levels. Utilities directly affected by this shift have been forced to revise their long-term plans to respond to rapidly growing demand.
Utilities expand investment plans
Dominion Energy, which supplies electricity to Virginia, the world’s largest data center market, announced that it had signed contracts to supply about 48.5 GW of power to data centers by the end of last year. In response, Dominion Energy increased its capital spending plan for the next 5 years by about 30% to $64.7 billion.
CenterPoint Energy, another major utility serving the Houston area, increased its 10-year capital investment plan to $65.5 billion in response to rapidly rising demand. The company now expects peak electricity demand to increase by 50% by 2029, 2 years earlier than previously expected.
The pace of change is expected to substantially alter the US energy mix. In an outlook released in March, the EIA forecast that natural gas generation could increase by 7.3% between 2025 and 2027 if data center demand remained higher than expected. It also projected that this scenario would slow the decline in coal generation, which has been falling steadily in recent decades.
Perhaps the most notable change, however, is that technology companies are now directly funding new facilities. Nuclear power is experiencing a substantial revival as AI providers and data center operators invest in developing new reactors and sign long-term contracts with existing power plants. These developments could increase nuclear generation capacity by 63% by 2050.
Meta funds electricity infrastructure
Meta recently made an unusual private investment commitment to substantially expand Louisiana’s power grid to supply its new $27 billion Hyperion data center. Scheduled to begin operating in 2028, the facility is eventually expected to consume more than 7 GW of electricity, enough to power millions of homes.
Taking account of its impact on the grid, Meta agreed to pay for 7 new natural gas power plants to be built by utility Entergy, in addition to 3 plants already approved. It will also fund 240 miles of new transmission lines connecting southern Louisiana with northern Louisiana and Arkansas, as well as 3 new battery storage facilities.
The agreement appears to be a response to growing public dissatisfaction over data centers’ impact on energy prices. People are also concerned about how rapidly rising demand will affect long-term grid reliability.
PJM Interconnection, the largest US grid operator, warned in February that the United States could face a power shortfall of up to 60 GW over the coming decades, and that insufficient electricity capacity could cause widespread blackouts as early as 2027.
One potential solution is to limit data centers’ workloads, and therefore their energy use, when the grid is under strain. Major power companies, including AES, Constellation, NextEra Energy and Vistra, are reportedly developing these so-called “flexible AI factories.”
However, the idea remains experimental, and it is uncertain whether major technology companies will voluntarily agree to regularly take their data centers offline. IT consultant Hunez told Reuters that companies could lose about $9,000 per minute when a data center goes offline.
Dedicated power networks and their effects
Given the complexity of meeting this new demand, pressure is growing on data center operators to address their own electricity needs. Even President Trump, who has generally taken a favorable stance toward the AI boom, urged technology companies to build their own power plants for data centers in the State of the Union address in February.
This is already happening. According to energy consultancy Cleanview, 46 data centers with a combined capacity of 56 GW are planning dedicated electricity infrastructure. This trend is prompting the emergence of a new energy system that operates alongside public electricity infrastructure.
This could have knock-on effects for everyone. For a start, most projects rely on natural gas generators because managing variable renewable energy output is difficult, potentially leading to a sharp increase in carbon emissions.
Because the most efficient turbines are difficult to obtain at short notice, generating facilities are using more polluting generators. Technology companies are also now competing with utilities to secure equipment. This could cause costs to rise sharply and ultimately be passed on to consumers.
The article concludes that it is becoming increasingly clear that the AI boom will fundamentally reshape the US energy system. It argues that companies are moving so quickly to build new facilities that there is little room to proceed carefully and sustainably.
Original Korean article: ainet.link
English brief by Korea Wind Intelligence, koreawindintel.com
