Wednesday 24 June 2026
Michael Hochberg /

The AI Race Isn’t About Electricity

Specialized silicon will solve the AI-driven energy panic.
A presenter in a leather jacket gestures toward a technical diagram of NVIDIA's Vera Rubin GPU architecture, labeled with "Grace" and "Blue" components, displayed on a black background.
Nvidia CEO Jensen Huang introduces the Vera Rubin GPU in Taipei on June 1, 2026. (Photo by I-Hwa Cheng / AFP via Getty Images)

Twenty-nine jet engines sit in a West Texas field. Crusoe, an infrastructure partner to OpenAI, ordered them in two batches from late 2024 through mid-2025. Each unit produces about 35 megawatts of electricity; together they produce a gigawatt, about the same output as a nuclear reactor.

Such investments reflect a fashionable worry that AI’s appetite for electricity will choke the current AI boom—that the United States will run out of power before it runs out of demand for AI computation.

Michael Hochberg is a visiting scholar at the Centre for Geopolitics at Cambridge University, a California Institute of Technology-trained physicist, and a serial semiconductor entrepreneur.

Related Content