What if the limiting factor for the AI boom isn’t bits, but electrons?

Big Tech companies like OpenAI, xAI, Meta, Google, Microsoft, and Amazon are racing each other to spend tens of billions of dollars on massive AI data centers. But no matter how many Nvidia GPUs you acquire through complex partnerships, there is one factor that may limit the industry’s AI dreams: energy.

A new analysis by the Financial Times found that the tech industry is currently moving forward with plans to build out a staggering 44 gigawatts’ worth of computing infrastructure. The problem is that there’s only about 25 gigawatts of power coming online in the next three years.

That 19-gigawatt gap is going to be a problem. It’s no question that one of the most difficult elements of the data center build-out has been how exactly to power them.

Even with the full support of the Trump administration, it could be hard for all the tech companies to get the power they want, something that Nvidia CEO Jensen Huang as well as Microsoft CEO Satya Nadella have both noted is the main obstacle to rapid AI expansion.

The energy part of the equation is one of the more politically perilous questions as well. As we saw in Memphis when Elon Musk’s Colossus data center came online, one of the more polarizing elements of the build-out was all those gas turbines and the pollution that came from them.

The Takeaway

A key issue for the government has been a trilemma centered around the AI boom’s power needs, one that’s been the bugbear of George Pollack, the senior US policy analyst at Signum Global Advisors. Essentially, the Trump administration can achieve two of three conflicting goals: (1) preside over the AI boom, (2) boost fossil fuels and stifle renewables, and (3) avoid household angst over energy prices.

The current course of action — stymie renewables and gas the AI boom — will begin to fall under the rather brutal laws of supply and demand, invariably resulting in higher energy costs for ordinary citizens as the massive utility companies can only produce so much power.