In the global race for artificial intelligence supremacy, American tech giants like Microsoft, Google, Amazon, and Meta face an unexpected roadblock — electricity. As these companies pour billions into AI data centres and chip production, the availability of electric power has become the latest bottleneck threatening to slow their progress.
Microsoft CEO Satya Nadella recently acknowledged that power, not chips, is now the biggest constraint. “You may actually have a bunch of chips sitting in inventory that I can’t plug in,” he said during a podcast with OpenAI chief Sam Altman. With data centres consuming immense energy and water resources, companies are now racing to secure power before they can even deploy new AI infrastructure.
Read More: The Rise and (Eventual) End of Algorithms
According to industry data, the combined spending of major tech companies could exceed $400 billion in 2025 as they construct massive data centres — but building these facilities takes around two years, while establishing new power lines can take up to a decade. Dominion Energy, Virginia’s main utility provider, says demand from data centres alone has risen to 47 gigawatts — the output of nearly 47 nuclear reactors. Experts warn that by 2030, data centres could consume up to 12% of total U.S. electricity, double today’s levels.
The pressure has already pushed several U.S. utilities to delay coal plant closures and invest in natural gas — a quick but carbon-heavy energy source. Meanwhile, Amazon and Google are betting on nuclear power, with plans to deploy small modular reactors (SMRs) and even restart old nuclear facilities by 2029. Tech companies are also investing in solar energy, battery storage, and even orbital computing projects, hoping to harness solar energy from space.
Read More: The AI Cold War
As governments and corporations push the limits of technological advancement, the AI boom may soon depend less on silicon chips — and more on who controls the world’s power supply.