The artificial intelligence gold rush has companies burning through billions in pursuit of superintelligence. But according to Eric Schmidt, the former Google CEO who guided the company through its most transformative decade, Silicon Valley is racing toward a cliff it refuses to see.
The problem isn’t computational power or talent acquisition. It’s something far more pedestrian: America’s aging electrical grid simply cannot handle what’s coming next.
When the Lights Go Out on AI Dreams
Schmidt made this stark assessment during a Thursday conversation on the Moonshots podcast, where he spoke with hosts Peter Diamandis and Dave Blundin about the infrastructure realities behind AI’s grand ambitions.
“AI’s natural limit is electricity, not chips,” Schmidt said, cutting through the industry’s semiconductor obsession with characteristic bluntness.
The numbers tell a sobering story. The United States needs an additional 92 gigawatts of electrical capacity to support the AI revolution currently underway. Each gigawatt represents roughly one nuclear power plant’s worth of generation. Yet in the past three decades, America has built exactly two nuclear facilities, and none are currently under construction.
This creates a mathematical impossibility that few in Silicon Valley want to acknowledge. Every major technology company is simultaneously racing toward the same goal: artificial superintelligence that surpasses human cognitive ability across virtually every domain. Meta’s Mark Zuckerberg and OpenAI’s Sam Altman are locked in an expensive talent war, pouring resources into research teams and computational infrastructure.
But Schmidt, who led Google from 2001 to 2011 and witnessed firsthand how quickly digital infrastructure can become a bottleneck, sees a different constraint emerging.
“Superintelligence is intelligence beyond the sum of all humans,” he wrote in a LinkedIn post Thursday. “It is reasonable to predict that we are going to have specialized AI savants in every field within five years.”
The Hidden Cost of Silicon Valley’s Biggest Bet
The energy crisis isn’t theoretical anymore. Microsoft has already felt the squeeze, committing to a 20-year power purchase agreement with Constellation Energy to resurrect Three Mile Island’s shuttered nuclear reactor. The Pennsylvania plant closed in 2019, but Microsoft needs it operational by 2028 to fuel its AI operations.
The resource drain shows up in unexpected places. Microsoft’s latest environmental report revealed a 34% surge in water consumption between 2021 and 2022, reaching 1.7 billion gallons. Industry experts attribute this increase directly to cooling requirements for AI data centers.
Global projections paint an even more dramatic picture. Researchers estimate that AI workloads could consume between 4.2 and 6.6 billion cubic meters of water by 2027. That volume could fill up to 2.6 million Olympic-sized swimming pools, or supply Canada’s entire population for more than a year.
OpenAI’s Sam Altman acknowledged this challenge last year, calling an energy breakthrough “essential for AI’s future.” His response has been characteristically ambitious: personal investments in Helion, a nuclear fusion startup targeting a 2028 pilot plant.
The Policy Collision Nobody Saw Coming
The tension between AI ambitions and infrastructure reality is already creating political friction. In May, Microsoft and AMD jointly lobbied U.S. senators to accelerate energy permitting processes, warning that AI’s power demands could overwhelm the electrical grid.
Environmental groups see this differently. Greenpeace argues that current AI growth trajectories could derail both national and international climate commitments, creating a policy collision between technological advancement and environmental protection.
Schmidt’s warning carries particular weight because of his unique vantage point. During his Google tenure, he watched the company navigate similar infrastructure challenges as internet usage exploded. The difference now is scale and urgency.
“We don’t know what AI will deliver, and we certainly don’t know what superintelligence will bring, but we know that it is coming fast,” Schmidt said. “We need to plan ahead to ensure we have the energy needed to meet the many opportunities and challenges that AI puts before us.”
The Irony of Progress
There’s a striking irony in Schmidt’s assessment. The technology sector that prides itself on disruption and innovation finds itself constrained by the most basic infrastructure challenge: generating enough electricity to power its ambitions.
This could fundamentally reshape the AI race. Companies with superior energy partnerships might gain decisive advantages over those with merely superior algorithms. The winners in artificial superintelligence might be determined not by who builds the smartest models, but by who secures the most reliable power sources.
The implications extend beyond Silicon Valley. If Schmidt’s analysis proves correct, the global race for AI supremacy might be won or lost based on electrical grid capacity rather than research breakthroughs. Nations with abundant, clean energy infrastructure could find themselves with unexpected strategic advantages.
Will energy constraints force a reckoning with AI’s superintelligence timeline, or will necessity drive the breakthrough solutions the industry desperately needs?
