OpenAI Execs Urge Massive GPU Expansion Amid Insatiable AI Compute Demand

OpenAI executives continue to highlight the company’s relentless demand for computing power, with CEO Sam Altman pushing to expand the company’s GPU capacity to over 1 million by the end of the year. This surge in need underscores the central role of hardware in advancing artificial intelligence. During a recent interview on Peter Diamandis’ “Moonshot” podcast, OpenAI’s Chief Product Officer Kevin Weil compared the impact of increased GPU availability to the evolution of internet bandwidth. “Every time we get more GPUs, they immediately get used,” Weil said. He likened the situation to how improved internet speed enabled the rise of video streaming—what was once impossible became everyday reality because the infrastructure could support it. Altman echoed this urgency, joking on X in July that the team should “better get to work figuring out how to 100x that” after reaching a milestone in GPU deployment. His comment came amid growing competition, particularly from Elon Musk’s xAI, which recently revealed its Colossus supercluster of over 200,000 GPUs used to train Grok4. Musk has since set an ambitious goal: building a system equivalent to 50 million Nvidia H100 chips within five years. For OpenAI, the race for compute is not just about training models—it’s foundational to every aspect of its operations. Weil noted that more GPUs can reduce latency, accelerate token generation, expand product access from pro to free users, and enable more rapid experimentation. “The more GPUs we get, the more AI we’ll all use,” he said. This demand has driven OpenAI to launch Stargate, a $500 billion joint venture with Oracle and SoftBank. Announced at the White House in January, the project aims to create a massive AI infrastructure capable of supporting the U.S. in achieving artificial general intelligence. CFO Sarah Friar told CNBC that the company is constantly constrained by compute limits, which is precisely why Stargate was created. Even within the company, managing GPU allocation is a constant challenge. Weil acknowledged that research teams have “basically infinite demand” for computing resources. Balancing these needs with product development and scalability remains a top priority. As AI models grow more complex and user demand expands, the race for GPUs has become a defining feature of the industry. For OpenAI, the message is clear: more compute means faster progress, broader access, and greater innovation.