SAN FRANCISCO — The U.S. electrical grid faces a critical challenge from the rising power demands of artificial intelligence data centers. Major technology companies are making long-term investments to secure their own energy infrastructure, a direct response to grid limitations.
Microsoft signed a 20-year agreement to restart the Three Mile Island nuclear plant, which has been offline since 2019. Amazon acquired a data center campus for $650 million to co-locate with the Susquehanna nuclear station in Pennsylvania. Google entered agreements with Kairos Power for small modular reactors, and Meta issued a request for proposals seeking up to 4 gigawatts of new nuclear capacity.
These moves reflect a growing concern about the reliability of existing power infrastructure. The American grid was designed for historical electricity demand growth of one to two percent annually.
AI processing significantly increases energy consumption. A single ChatGPT query uses approximately 10 times the energy of a standard Google search. Training large language models requires power equivalent to that of small cities.
Industry forecasts highlight the scale of this demand. McKinsey projects AI data center capital expenditure will reach $5.2 trillion between now and 2030. Goldman Sachs Research predicts global data center power demand will surge up to 165 percent by 2030 compared to 2023 levels.
Grid strain is already visible. Berkeley Lab found that more than 70 percent of grid interconnection requests in the United States are ultimately withdrawn because the existing grid cannot accommodate the proposed new loads.
Kevin O'Leary, an infrastructure investor, said 50 percent of the data centers currently planned across the United States will never be built. He attributes this to the grid's inability to deliver sufficient power, not a lack of capital or demand.
The hyperscalers have recognized this constraint. Their multi-billion-dollar, multi-year commitments to dedicated power sources represent a strategic move to establish competitive moats and ensure future operational capacity.
An alternative view holds that the grid will ultimately adapt — that the data center building boom, backed by the wealthiest companies and powerful governments, will find ways to secure necessary electricity. Some analysts also warn of an AI investment bubble, which could lead to inflated electricity demand forecasts.
If these demand forecasts prove inaccurate, ratepayers could be left funding unnecessary power plants and grid upgrades. Meta CEO Mark Zuckerberg said on an earnings call that his company's strategy is to "aggressively front-load building" infrastructure, signaling commitment despite potential market volatility.
The long-term capital allocation decisions by these tech giants mark a fundamental shift in how large-scale compute power will be provisioned. The market has not yet fully priced in the infrastructure buildout and its associated costs and complexities.
