OpenAI is accelerating its ambitious 'Stargate' project, aiming to construct the foundational compute power for an era defined by artificial intelligence. The company announced it has already surpassed its initial 2025 commitment of securing 10 gigawatts (GW) of AI infrastructure in the United States by 2029, adding over 3 GW in just the last 90 days.
This rapid expansion reflects the escalating demand for AI capabilities across all sectors. OpenAI emphasizes that meeting this need responsibly requires building more compute capacity, faster. Compute is framed as the critical input for training advanced models, ensuring reliable service, improving performance, and ultimately lowering costs.
The 'AI flywheel' concept, where more compute enables better models, leading to increased usage and revenue for reinvestment in infrastructure, underpins OpenAI's strategy. While financing and partnership models may adapt, the focus remains on delivering scalable capacity on time and with flexibility for evolving technology and demand.
Building an Ecosystem
Recognizing the immense scale of the undertaking, OpenAI is forging broad partnerships. Building AGI infrastructure necessitates coordination with local communities, utilities, energy providers, chipmakers, cloud providers, construction firms, investors, and skilled trades. This partner-centric approach aims to accelerate progress, enhance efficiency, mitigate execution risks, and maintain adaptability.
Expanding the Footprint
OpenAI and its partners are actively scouting new data center locations nationwide, planning significant compute capacity expansion beyond the initial 10 GW target. Site selection hinges on critical factors like power availability, land, permitting, transmission infrastructure, workforce, community backing, and partner readiness.