May 21, 2025 (Bloomberg) - OpenAI is building one of the world’s largest AI-focused data centers in partnership with Microsoft, a massive project codenamed "Stargate" that marks a strategic shift toward custom-built infrastructure to support the development of artificial general intelligence.
Located in partnership with Microsoft and codenamed "Stargate," this data center project is set to become one of the world’s largest and most advanced facilities dedicated entirely to powering artificial general intelligence (AGI). The scale is unprecedented: an estimated $100 billion investment spread over years, with a physical footprint rivaling or exceeding that of some of the largest hyperscale cloud campuses. For OpenAI, which has until now relied heavily on Microsoft’s Azure infrastructure, Stargate represents a strategic leap into custom-designed systems, tailored silicon, and deeply optimized training environments—engineered to accelerate AI model development far beyond today’s capabilities.
What makes the Stargate initiative uniquely significant is that it marks a convergence of ambition, necessity, and competition. Training the next generation of foundation models—those with trillions of parameters—demands a level of computing power that goes well beyond conventional data centers. Standard GPU clusters are no longer sufficient for OpenAI’s vision. Stargate will reportedly integrate next-generation AI chips co-designed with Microsoft, specialized for model training efficiency and throughput. The facility is expected to host hundreds of thousands—potentially millions—of such accelerators, networked with ultra-low-latency fabrics, purpose-built for massive-scale parallelism. The goal is not simply to keep up with demand, but to create the physical and computational backbone for OpenAI’s future breakthroughs—whether in language, vision, robotics, or yet-undiscovered domains of artificial cognition.
Beyond hardware, Stargate also reveals OpenAI’s evolving philosophy toward infrastructure sovereignty. By anchoring its operations in a megafactory model, the company reduces its dependency on third-party cloud services and gains more control over performance, security, and energy consumption. Power use, in fact, is central to the design: the data center will likely draw on dedicated renewable energy sources, and advanced cooling systems are being designed to handle the immense thermal output from densely packed AI chips. All of this points toward a new kind of facility—more akin to a nuclear research lab or a semiconductor fab than a traditional server farm. It is a physical manifestation of the hypothesis that the road to AGI will require not just smarter algorithms, but entire ecosystems of purpose-built infrastructure operating at planetary scale.
Stargate is also emblematic of a broader industry trend: the re-industrialization of AI. As models grow more capable and complex, the conversation is shifting away from just software and data toward factories, logistics, materials, and electrical engineering. OpenAI’s move signals that future AI leadership will hinge not only on theoretical breakthroughs but on the ability to manufacture, maintain, and operate sophisticated physical systems. In that sense, Stargate is not just a data center—it is a bet that the frontier of intelligence will be shaped as much by steel and silicon as by code and data. As OpenAI pushes forward with this massive undertaking, the implications ripple outward: from geopolitics and energy markets to ethics and existential safety. Stargate, in name and ambition, suggests that the company is no longer just looking to build smarter machines—it’s trying to engineer the infrastructure of a new technological epoch.
