TLDR: Former Google CEO Eric Schmidt has issued a stark warning that the rapid advancement of artificial intelligence, particularly towards superintelligence, is facing a critical bottleneck: electricity, not computing chips. He estimates the U.S. alone will require an additional 92 gigawatts of power to support AI’s growth, equivalent to building 92 new nuclear power plants, a monumental challenge given current infrastructure development rates. Major tech firms are already scrambling to secure energy sources, including reviving old nuclear facilities, as environmental concerns over AI’s resource consumption also mount.
Former Google CEO Eric Schmidt has delivered a sobering assessment on the future trajectory of artificial intelligence, asserting that the true limiting factor for AI’s exponential growth is not the availability of advanced chips, but rather the sheer demand for electricity. Speaking on the ‘Moonshots’ podcast and in a LinkedIn post, Schmidt, who also chairs the Special Competitive Studies Project, a pro-AI think tank, highlighted an impending energy crisis that could significantly impede the AI revolution.
According to Schmidt, the United States alone is projected to require an additional 92 gigawatts of power to sustain its burgeoning AI infrastructure. To put this into perspective, one gigawatt is roughly equivalent to the output of a single nuclear power station. This means the U.S. would need to construct approximately 92 new nuclear plants, a daunting task considering only two such facilities have been built in the country over the past three decades. This stark reality underscores a significant infrastructure challenge as Silicon Valley races towards achieving artificial superintelligence—AI that surpasses human cognitive abilities across virtually all domains.
Tech giants are already feeling the pressure and taking proactive, often drastic, measures to secure the necessary power. Microsoft, for instance, has entered a 20-year agreement with Constellation Energy to restart the Three Mile Island nuclear plant in Pennsylvania, which ceased operations in 2019, with a target relaunch in 2028. The company’s AI operations have already led to a notable increase in resource consumption, with Microsoft reporting a 34% surge in water usage, consuming 1.7 billion gallons between 2021 and 2022, a trend directly linked by experts to rising AI workloads. Similarly, OpenAI CEO Sam Altman has personally invested in Helion, a nuclear fusion startup, acknowledging that a breakthrough in energy is ‘essential for AI’s future.’ Both Microsoft and AMD have also reportedly urged U.S. policymakers to fast-track energy permits.
Schmidt emphasized that while superintelligence is inevitable and approaching rapidly, with specialized AI tools expected across all fields within five years, the urgent need is for proactive planning of energy infrastructure. He stressed that it’s not enough to build the ‘brains’ of AI; the world must also be able to ‘power them.’ Environmental groups, including Greenpeace, have echoed these concerns, warning that unchecked AI expansion risks undermining global climate goals rather than supporting them.
Also Read:
- AI Startup Sector Faces Significant Uncertainty for 2026 Growth, Raising Concerns Over Market Valuation
- The AI Revolution: Unpacking 2025’s Market Trends, Technological Advancements, and Economic Impact
This energy crunch is not just an infrastructure problem but also carries geopolitical implications. Schmidt drew parallels to the 1938 Einstein letter urging FDR to pursue the atomic bomb, suggesting a similar inflection point for AI. He noted that while U.S. chip restrictions have slowed China, the country is innovating around them, and with open-source models proliferating, AI capabilities could shift leadership eastward. Schmidt’s paper on ‘mutual AI malfunction’ even proposes a deterrence strategy akin to nuclear MAD (Mutually Assured Destruction) for cyber means.
