Khosla: AI's power crunch needs on-site generation, not grid upgrades

What happened

Vinod Khosla, founder of Khosla Ventures and early OpenAI backer, argues that linear generators—which convert fuel directly into electricity without combustion—offer a faster solution to powering AI data centers than waiting for grid connections. A single large AI data center now consumes as much electricity as a city of 80,000 people, while more than 2,600 gigawatts of proposed projects are waiting to connect to a grid whose total installed capacity is less than half that number.

Why it matters

The existing electricity grid cannot meet AI infrastructure demand at the scale, flexibility and speed required. Grid interconnection queues stretch seven years, forcing companies to rely on 50-year-old gas turbines. Khosla contends that companies adopting on-site linear generation—which can be deployed in months, operate independently of the grid, and run on natural gas, biogas, hydrogen, propane or fuel blends—will have a structural competitive advantage over those waiting for grid upgrades. This shapes which nations and companies will lead in the AI era.

What to watch

Data center executives are already moving: already, a significant percentage say they're either investing or planning to invest in on-site generation, and this technology is being adopted by major companies and data center developers. The infrastructure decisions being made now will shape the competitive landscape of AI for the next twenty years.