OpenAI's 30GW Bet
- OpenAI announced plans to scale to 30GW of compute by 2030, building on a prior 10GW commitment. - The company said it has already identified 8GW toward that target in its recent disclosure. - This signals major capital intensity for future models and infrastructure needs, with OpenAI publicly framing long-term compute expansion as strategic (x.com).
OpenAI said on April 22 it plans to build 30 gigawatts of AI compute by 2030, tripling the 10-gigawatt target it laid out last year. (x.com) The company said it committed in January 2025 to generate 10 gigawatts of compute and has already identified more than 8 gigawatts toward that earlier goal. OpenAI announced the original Stargate plan on January 21, 2025 as a project that intended to invest $500 billion over four years in United States AI infrastructure. (x.com) (openai.com) OpenAI’s public buildout has moved through Stargate, the infrastructure platform it formed with SoftBank, Oracle and other partners. In September 2025, OpenAI said Stargate had nearly 7 gigawatts of planned capacity across sites in Texas, New Mexico, Ohio and Wisconsin, with more than $400 billion of investment over three years. (openai.com) A gigawatt is a power-scale measure more familiar from electric grids than software products. OpenAI has increasingly described electricity, chips and data centers as the core inputs for training and running future AI systems at the scale it wants. (openai.com) That framing has hardened as OpenAI’s usage has grown. In an October 27, 2025 policy submission, the company said its weekly user base had doubled in seven months from more than 400 million to more than 800 million people, and tied that growth to the need for much more energy. (openai.com) The 30-gigawatt target also extends a string of large infrastructure deals OpenAI signed in 2025. In September, OpenAI and Nvidia said they planned at least 10 gigawatts of Nvidia-powered AI data centers, and in October, OpenAI and Broadcom said they would work on 10 gigawatts of OpenAI-designed accelerators by 2029. (openai.com 1) (openai.com 2) OpenAI has also widened the financing behind that expansion. On March 31, 2026, the company said it had raised $122 billion in new funding to expand frontier AI globally, invest in next-generation compute and meet demand for ChatGPT, Codex and enterprise products. (openai.com) The bottleneck is not only money or chips. OpenAI’s own policy papers have argued that the United States is short on power, pointing to China’s 429 gigawatts of new electricity capacity in 2024 versus 51 gigawatts added in the United States. (openai.com 1) (openai.com 2) OpenAI is presenting the next phase of AI as an infrastructure race measured in substations, turbines and transmission lines as much as in models. Its 30-gigawatt goal turns that argument into a number investors, utilities and rivals can now measure against. (x.com) (openai.com)
