Here's an uncomfortable thought: the United States might lose the AI race not because its chips aren't good enough, but because it doesn't have enough electricity to run them. That's the warning coming from Tesla Inc. (TSLA) CEO Elon Musk and Futurum CEO Daniel Newman, who are pointing to some eye-opening data about China's energy expansion.

China's Energy Infrastructure Is Growing at Breakneck Speed

The numbers tell a striking story. Over the past year, China has been adding electrical generation capacity at a pace that makes U.S. efforts look almost quaint. The standout performer? Solar power, where China's additions dwarf American installations by several multiples. And it's not just solar—China is also ramping up wind, hydro, and nuclear capacity at the same time.

Musk's reaction to the data was succinct: he called China's solar expansion "incredible" and said the U.S. needs to match that pace to stay competitive. Which raises an interesting question about what "competitive" actually means in the AI era.

Chips Don't Matter If You Can't Power Them

Newman, sharing the same data, went deeper into what this means strategically. Sure, American companies like Nvidia Corp (NVDA) are years ahead in designing the advanced chips that power AI systems. But Newman argues that semiconductor superiority means nothing if you don't have the electricity to run the data centers those chips live in.

The problem, according to Newman, is decades of neglect. The U.S. hasn't meaningfully expanded its power generation or grid infrastructure in a very long time. Meanwhile, AI systems are becoming exponentially more energy-hungry. If electricity supply can't keep pace with chip performance, then chip leadership becomes somewhat academic.

Newman's prescription? An "all hands on deck" approach that uses traditional power sources in the near term while investing heavily in nuclear energy for the long haul.

Why Energy Is the Real Battleground

This isn't just tech industry speculation. Michael Burry, the investor famous for calling the 2008 financial crisis, has made similar observations about China's energy advantage in AI development. When the technology fundamentally depends on power-hungry processors, whoever can generate the most electricity has a structural edge.

The scale of what's coming is hard to grasp. Goldman Sachs forecasts that global electricity consumption from data centers, including AI workloads, will jump 175% by 2030. To put that in perspective, it's roughly equivalent to adding an entirely new country that ranks among the world's top 10 in electricity consumption.

In the U.S. specifically, power demand is projected to increase about 2.6% annually through 2030, driven largely by data center growth. That might not sound dramatic, but it's far above the average electricity demand growth seen over the past two decades. And the question is whether American infrastructure can actually deliver that increase.

The irony here is almost perfect. The U.S. semiconductor industry remains the global leader—Nvidia ranks in the 97th percentile for growth among its peers—but that leadership might not matter if the servers running those chips keep hitting power constraints.

So while everyone's been focused on chip design and semiconductor manufacturing, China has apparently been playing a different game entirely: building the power infrastructure that modern AI actually needs to function. It turns out the AI race might be less about who makes the best processors and more about who can keep the lights on.