Jeff Bezos has a message for regulators eyeing tough limits on the explosive growth of AI infrastructure. Outlawing data centers solves nothing. The Amazon founder and executive chairman delivered the sharp analogy during a CNBC interview Thursday, just as his new venture closed a massive funding round.
“You don’t want to accidentally outlaw the knife because it can be used in a bad way,” Bezos said. “Knives are important tools and yes, every once in a while they get misused by someone, but you don’t say the solution to that isn’t to say, ‘OK, no more data centers. No more knives.’ That’s not a smart approach to regulation.”
The comments landed amid rising alarm over electricity demands. Data centers already consume hundreds of terawatt hours annually. Projections show that figure doubling or more by 2030. States from California to Nebraska now push laws forcing operators to cover new power plant costs or face delays. Yet Bezos argues for balance. Government rules serve real purposes. He pointed to the FAA and FDA as models that protect safety without killing progress.
Power Strain Meets Industrial Ambition
Bezos isn’t watching from the sidelines. He co-leads Prometheus, the AI startup he founded in late 2025 with Vik Bajaj, former co-founder of Alphabet’s Verily. The company disclosed a $12 billion Series B on Thursday, lifting its valuation to roughly $41 billion. Investors included JPMorgan, BlackRock, Goldman Sachs, DST Global and Arch Venture Partners. Bezos backed both rounds heavily.
Prometheus builds AI systems trained on specialized engineering data. Not the internet-scraped text that powers chatbots. The goal? An “artificial general engineer” that compresses design and manufacturing timelines. Bezos expects gains of 10 times or greater in some workflows. The tools could reshape chip development, aerospace components, automobiles and more. They might even help hyperscalers optimize their own data centers.
“It’s easy to imagine Amazon or any hyperscaler using the kinds of tools” that Prometheus is developing to improve their data centers, Bezos told CNBC. He added that Prometheus itself could become an AWS customer as its compute needs grow. The companies will stay at arm’s length. Prometheus pulls compute from multiple providers.
This vision sits at the heart of the tension. AI needs vast compute. That compute demands power. And the power grid strains under the load. The International Energy Agency projects data center electricity use rising from about 485 TWh in 2025 to 950 TWh in 2030. That would equal roughly 3% of global electricity. Some estimates put 2026 consumption near 1,050 TWh, enough to rank data centers as the world’s fifth-largest energy user if treated as a country. (IEA)
But the solution isn’t prohibition. Bezos made that clear. Regulation should target specific applications and risks, not the foundational infrastructure. His stance echoes broader industry pushback even as local governments act. At least 27 states have considered or passed bills on data center energy costs, reporting requirements or outright pauses. California, Ohio and Utah already enforce laws that go further than voluntary federal pledges. (Multistate)
Communities feel the pinch first. Electricity prices, water consumption, noise and land use spark protests. In Utah, investor Kevin O’Leary scaled back a giant project after local resistance. Nebraska lawmakers debate requiring new data centers to build dedicated generation. These moves reflect real pressure on ratepayers and grids. Yet they also risk slowing the very innovation that could ease bottlenecks elsewhere.
Prometheus offers one path forward. Its technology aims to accelerate the invention loop. Faster engineering means better hardware, more efficient chips, perhaps even designs that reduce energy intensity over time. The company maintains just 120 to 150 employees but commands attention because of its founders and capital. Bajaj and Bezos spoke publicly together for the first time in the CNBC appearance, signaling a shift from earlier secrecy.
Bezos also sees broader economic effects. AI will create labor scarcity in some areas. That scarcity, he argues, will drive higher productivity and living standards. The work shifts to higher levels. People focus on creativity and oversight while machines handle routine complexity. Still, he cautions against overregulation that could stall these gains.
The data center debate has gone global. The European Union prepares a Data Centre Energy Efficiency Package for early 2026. It pairs with AI and digitalization roadmaps targeting carbon-neutral operations by 2030. Reporting on power usage effectiveness and water metrics becomes mandatory in many places. In the U.S., no comprehensive federal standard exists, but the conversation has turned bipartisan. Both parties worry about grid reliability and consumer costs. (White & Case)
Bezos has voiced related thoughts before. In May he called two-to-three-year timelines for space-based data centers overly ambitious. The idea intrigues him. Moving heat-producing infrastructure off-planet could ease earthly environmental strain. Yet physics, launch costs and maintenance challenges make near-term deployment unlikely. He expects it will happen eventually. Just not on the most optimistic schedules.
Prometheus itself ties into this long-term thinking. Its focus on physical AI and manufacturing could improve spacecraft design alongside terrestrial hardware. The company explores a holding structure to acquire firms positioned to benefit or face disruption from these tools. Details remain sparse. The public comments Thursday offered the clearest window yet into strategy.
Critics worry that massive funding rounds like Prometheus’ simply concentrate power among a few players. Hyperscalers already dominate compute. Now a Bezos-led entity with $41 billion valuation enters the design layer. Success there could influence everything from next-generation processors to factory automation. The “artificial general engineer” concept sounds abstract until applied to real bottlenecks in chip fabrication or energy system planning.
And the energy question won’t fade. Recent analyses show data center growth outpacing overall electricity demand by wide margins. Compound annual growth rates for data center power have run four times higher than the global average. Without efficiency gains or new generation, brownouts, higher prices and stalled projects loom. Some regions already see power as the primary constraint on expansion. (Brookings Institution)
Bezos doesn’t dismiss those risks. He acknowledges regulation’s value in safety and product quality. His knife analogy cuts both ways. Misuse happens. Society still needs knives. The same logic applies to data centers and the AI systems they power. Blanket restrictions would harm more than they protect.
Industry insiders watch closely. Prometheus could supply tools that let data center operators design more efficient cooling, layout or chip architectures. Or it could accelerate entirely new computing paradigms. Either outcome influences the regulatory conversation. If engineering speed doubles or triples, the case for measured rules grows stronger. Faster iteration might solve some problems before they compound.
Bezos and Bajaj emphasized they aren’t hiding. The funding announcement and interviews mark a deliberate step into view. With $18 billion-plus raised across rounds and a lofty valuation, expectations run high. Delivery on the engineering acceleration promise will determine whether Prometheus reshapes industries or joins a list of well-funded AI experiments.
For now the tension persists. Surging demand for AI capability collides with limits on power and public tolerance. Regulators at state and federal levels weigh options. Some favor mandates for self-generation or cost absorption. Others eye efficiency standards and transparency. Bezos offers a principle. Regulate the misuse, not the tool. Keep the knives. Improve how they get made and used.
That stance aligns with his track record. Amazon built vast data centers to support cloud computing and e-commerce. Now as executive chairman he advises the company on AI while pouring personal capital into Prometheus. The interests overlap but remain distinct. He sees mutual benefit without full integration.
The coming months will test these ideas. New state laws take effect. EU proposals advance. Power purchase agreements multiply as operators lock in nuclear, gas and renewable capacity. And Prometheus hires engineers, gathers proprietary datasets and refines its models. The datasets differ sharply from those used in large language models. They require access to hard-to-obtain engineering records and simulation results.
Success won’t come easily. Training systems on physical world data brings unique challenges. Yet the prize matches the scale of investment. An AI that masters engineering could compress decades of iterative design into years. That acceleration matters when energy constraints tighten and competition for talent intensifies.
Bezos has bet big. His comments Thursday framed the bet in practical terms. Society faces trade-offs with powerful technologies. Banning the infrastructure that enables them counts as a poor choice. Better to guide development, enforce safety where needed and reap the productivity gains. The knife stays in the drawer. But it stays sharp.
WebProNews is an iEntry Publication