Jeff Bezos thinks the current rush to build private data centers for AI is a massive waste of resources—and he’s drawing parallels to a 110-year-old mistake. Speaking at the 2026 New York Times DealBook Summit, the Amazon founder argued that companies like Meta and OpenAI constructing their own massive computing facilities mirror early 20th-century factories that generated their own electricity before power grids existed.“Right now, everybody is building their own data center, their own generators essentially. And that’s not going to last,” Bezos said, recounting a visit to a 300-year-old Luxembourg brewery that once operated its own power plant. He predicts the industry will inevitably shift toward centralized cloud computing services like Amazon Web Services (AWS), which offer far greater efficiency and scale than isolated infrastructure projects.
AI’s exploding power demands are reshaping global energy consumption
The stakes are enormous. Data centers consumed 415 terawatt-hours of electricity in 2024—about 1.5% of global demand. By 2030, that figure is expected to more than double to 945 TWh, roughly equivalent to Japan’s entire consumption. The US could see data center electricity use jump from 4% to 12% of national demand.A single ChatGPT query uses nearly 10 times the energy of a Google search, while training one large AI model consumes as much power as 200 American homes annually. This has triggered a scramble for solutions, from nuclear power deals to space-based data centers.
Tech CEOs agree: Energy, not chips, is the real bottleneck
Jeff Bezos isn’t alone. Microsoft CEO Satya Nadella has acknowledged that GPUs sit idle due to insufficient power, noting “I don’t have warm shells to plug into.” OpenAI’s Sam Altman has called for breakthroughs in fusion or cheaper solar, while Google CEO Sundar Pichai labeled power “the long-term bottleneck for AI.”Meta recently secured nuclear power deals to generate 6.6 gigawatts by 2035 for its Prometheus supercluster in Ohio. Alphabet acquired Intersect Power for $4.75 billion to build co-located energy and computing infrastructure.Bezos positioned AI as a “horizontal enabling layer” like electricity itself—but warned the infrastructure needs the same centralized efficiency revolution that transformed industry a century ago.
