Sam Altman's AI empire needs trillions and a quarter of US power grid

Sam Altman, CEO of OpenAI, is spearheading an unprecedented AI infrastructure buildout that could consume a staggering 250 gigawatts of power by 2033—equivalent to a quarter of the U.S. electrical grid. This revelation came during a media event at OpenAI’s Stargate data center project in Abilene, Texas, where Altman emphasized the need for "brute industrial force" to scale AI capabilities.

The Stargate Project: A Texas-Sized Undertaking

The Abilene site, a collaboration with Oracle and SoftBank, spans 800 acres and employs 6,400 workers. It’s just the beginning: OpenAI plans to construct five more massive data centers across the U.S., aiming for 10 gigawatts of capacity by 2025. Altman likened the project to historic feats like the Hoover Dam and the Apollo program, underscoring its scale and ambition.

Industrial Compute: OpenAI’s New Frontier

In an internal Slack note obtained by Sources, Altman revealed OpenAI’s "audacious long-term goal" of 250 gigawatts by 2033. The company has formalized an "industrial compute team" to manage this expansion, led by Peter Hoeschele. Altman wrote that achieving this would require "trillions" in investment and a reimagining of how compute infrastructure is built.

Public Awareness and Skepticism

Altman admitted that most people don’t grasp the physical scale behind AI tools like ChatGPT. Critics question whether this infrastructure binge is necessary or reckless, especially given the environmental and economic costs. Some compare it to California’s high-speed rail, while others see it as essential for maintaining U.S. competitiveness in AI.

Broader Implications

The AI infrastructure boom isn’t just an OpenAI phenomenon. Companies like Nscale are raising billions to build "AI factories," while partnerships like OpenAI’s $100M deal with Databricks aim to integrate AI agents into enterprise workflows. Meanwhile, Elon Musk’s xAI is joining other AI giants in securing nominal-fee contracts with federal agencies.

Key Figures:

• $923 billion: Estimated U.S. economic output from AI-driven capital expenditures in 2025.
• 2.7 million jobs: Supported by this spending, per IMPLAN analysis.

Altman’s vision is clear: AI’s future hinges on gargantuan infrastructure. Whether it’s a boon or a boondoggle remains debated, but one thing is certain—the race for AI supremacy is now measured in gigawatts.