Information & Technology

Microsoft and OpenAI Plan USD 100 Billion Data Center for Advanced AI Models

• By Atul B
• 2025-05-07

The next generation of artificial intelligence models will be powered by Microsoft and OpenAI's alleged collaboration on "Stargate," a ground-breaking supercomputing facility in the United States that might cost up to $100 billion. Launched in 2028 and expected to completely extend by 2030, Stargate is the fifth and largest phase of their multi-year relationship, according to Reuters. It will have an extraordinary 5GW of dedicated power capacity to support millions of specialized AI processors. As a result of the project's increased demand for GenAI infrastructure, new energy options, including small modular nuclear reactors, are being investigated in order to meet its enormous power requirements.

Microsoft and OpenAI’s AI Supercomputing Revolution with Multi-Vendor Silicon Power

The Stargate data center will be able to handle significantly more sophisticated computational operations than traditional hyperscale facilities because of its unique optimization for AI workloads. Millions of specialized AI server chips are anticipated to be housed there, powering OpenAI's upcoming models. This facility will follow a larger industry trend as IT giants like Microsoft, AWS, and Google each invest billions of dollars in new, AI-focused data centers to meet growing consumer demand, even if the precise U.S. location has not yet been confirmed.

The AI chips themselves are a major source of expense; for example, NVIDIA's upcoming Blackwell B200 accelerators are priced between $30,000 and $40,000. The Stargate supercomputer is being designed to support chipsets from several providers, including AMD's most recent processors and Microsoft's proprietary "Azure Maia" and "Cobalt" architectures, in order to minimize vendor lock-in and increase resiliency. In addition to guaranteeing competitive price and supply security, this multi-vendor model allows the project to take advantage of each partner's finest silicon advances.  Together with specially designed AI software stacks and high-bandwidth interconnects, this combination of various hardware highlights Microsoft and OpenAI's goal of establishing a new standard for AI infrastructure performance, adaptability, and scale.

Microsoft and OpenAI’s Bold Investment in Next-Gen AI Infrastructure

Microsoft has committed up to $100 billion, more than 100 times the cost of a typical large data center, to fund the huge "Stargate" supercomputing program in order to support OpenAI and next-generation AI models. Microsoft, which has already invested over $13 billion in OpenAI, will use both NVIDIA GPUs and its own proprietary silicon in five stages, with Stargate being the last stage, which is expected to go live in 2028 and expand fully through 2030. In order to build out the necessary compute capacity, the partners are currently in the third phase of the project, acquiring AI chipsets, which might cost anywhere from $30,000 to $40,000. Prior to the complete Stargate implementation, a mid-cycle, fourth-phase mini-supercomputer project is planned to begin in 2026 as a testbed for networking, power, and cooling infrastructures.

Given the ongoing growth in demand for generative AI, hyperscale data center construction increased by 30% in 2024. Stargate plans to integrate cutting-edge liquid-cooling technology and on-site renewable energy generation to handle its anticipated 5 GW power usage without putting undue strain on regional grids. In addition to petabyte-scale storage, the expansive campuses, which cover hundreds of acres, will have high-bandwidth interconnects to support workloads involving model training and inference in real time. The need of having a strong AI infrastructure in order to sustain global leadership in artificial intelligence is highlighted by this extraordinary investment.

Next‑Gen AI Campus to Leverage Nuclear Power and Ethernet Networking

The ambitious "Stargate" venture from Microsoft and OpenAI will house millions of specialized server chips during a multi-phase rollout, making it particularly tuned for generative AI workloads. Reuters reports that although the supercomputer complex is still in the early stages of planning, it is anticipated to open in 2028 and grow over the course of two years, possibly consuming up to 5 GW of electricity when fully developed. Like Amazon, both businesses are looking at nuclear energy possibilities in light of increasing power demands. They have even hired nuclear experts to speed up the construction of Small Modular Reactors (SMRs).

Additionally, OpenAI intends to modernize its network infrastructure by switching from proprietary InfiniBand to high-speed Ethernet fabrics. Stargate is the fifth phase of their relationship, following a projected Q4 FY26 "Phase 4" supercomputer in Mount Pleasant, Wisconsin, where Microsoft has acquired 1,000 acres of extra property and a $1 billion expansion, possibly for a future $10 billion AI center. The amount of money needed to maintain leadership in AI model development and implementation is highlighted by this next-generation facility.