Sam Altman, the CEO of OpenAI, has always aimed high, and his latest proposal underscores that ambition. On Tuesday, he introduced an extensive initiative called “Abundant Intelligence,” detailing a factory designed to generate a gigawatt of AI infrastructure weekly. To provide context, one gigawatt can supply electricity to approximately 876,000 households for a year. Achieving this scale weekly is incredibly ambitious. Altman himself recognized the challenges, stating, “the execution of this will be extremely difficult” and may require years of effort.
Nonetheless, he believes the result could be “the coolest and most important infrastructure project ever.” This announcement follows a major financial development, as Nvidia recently revealed a $100 billion investment in Altman’s projects, granting access to around 10 gigawatts of AI-focused data centers. This single agreement is more substantial than the annual GDP of several small countries and represents just one element of Altman’s broader strategy. A key component of Altman’s vision is Stargate, a collaborative $500 billion endeavor with Oracle and SoftBank. This initiative aims to establish AI “super-hubs” throughout the United States, with the first currently being built in Abilene, Texas.
OpenAI has confirmed the selection of five additional sites, which combined will provide nearly 7 gigawatts of capacity. Although Altman has highlighted that much of the infrastructure will be based in the U.S., he has hinted at upcoming announcements regarding international partnerships and expanded rollout plans. Currently, the initiative resembles both a lofty aspiration and a massive construction endeavor, merging significant investments with unparalleled ambition. However, the grand vision confronts a significant obstacle: energy. Operating even the 10 gigawatts associated with the Nvidia agreement will demand enormous quantities of electricity. Experts express doubts about whether existing power grids can accommodate this scale.
Brad Gastwirth, head of research at Circular Technology, warned that energy supply could become the “silent bottleneck” in the AI race in Silicon Valley. “This will increasingly become a larger issue as years go by,” he stated in a recent interview. The challenge becomes clear when viewed through a numerical lens. Generating a gigawatt every week equates to constructing the energy infrastructure of a major city within days and then repeating that process weekly. This undertaking extends beyond chips, cooling systems, and server racks; it relies on obtaining vast amounts of clean, reliable energy without overburdening power grids or exacerbating climate risks. Looking forward, despite the challenges, Altman remains optimistic.
He perceives the obstacles not as barriers but as chances to redesign global infrastructure for the AI era. He believes the groundwork for this future is already underway as generative systems integrate into daily life. If successful, Altman’s plan would not only propel the advancement of larger AI models but also transform the intersection of technology and energy in the 21st century. Whether society is prepared for “a gigawatt every week” is still uncertain, but Altman appears determined to explore that possibility.