Microsoft and OpenAI are reportedly looking far beyond today’s AI data centers. According to The Information, executives at the two companies have discussed a massive U.S. supercomputer project code-named "Stargate" that could cost as much as $100 billion and accelerate OpenAI's AI development.
The reported plan is not just another data center expansion. It would involve a supercomputer made up of millions of specialized server processors, require several gigawatts of power, and become the largest in a series of data centers planned over the next six years.
What the Stargate project would be
According to three people who took part in confidential talks, Microsoft and OpenAI executives are shaping plans for a data center built around a new supercomputer. The goal would be to give OpenAI far more computing capacity for developing AI models.
The project is code-named "Stargate." One person who has spoken with OpenAI CEO Sam Altman about it, and another who has seen some of Microsoft's initial cost estimates, said it could cost as much as $100 billion.
If completed, the system would represent an extraordinary bet on AI infrastructure. The source article describes it as potentially one of the largest investments in the history of computer science.
The scale matters because advanced AI development depends heavily on compute. In plain terms, more specialized processors and more data center capacity could give OpenAI more room to train, test, and improve its models. The reported project shows how Microsoft and OpenAI are thinking about infrastructure as a core part of future AI progress, not merely as a support function.
Timing, scale, and power demands
Officials discussed launching Stargate in 2028. They also discussed an expansion by 2030 that could require up to five gigawatts of power toward the end.
The project would be built in the U.S. and would be the largest of several data centers planned over the next six years. The Information reported that the system would need several gigawatts of power, an amount described as the equivalent of at least several large data centers today.
That electricity requirement is one of the central challenges in the plan. The source article says a large share of the cost would go toward procuring chips, but securing enough power sources to operate the system would also be difficult.
Funding would likely come primarily from Microsoft. That detail is important because the plan, as reported, depends not only on technical ambition but also on the willingness to finance a project at a scale far beyond typical data center spending.
Why chips are at the center of the story
The reported Stargate plan fits into a broader concern repeatedly raised by Altman: a shortage of chips for AI development. The source article says he has pointed to that shortage lately and is said to be in talks with numerous investors and project partners about building a global infrastructure of chip factories.
For a supercomputer made up of millions of specialized server processors, chip supply would be a defining issue. Without enough processors, a project of this size cannot reach its intended scale. Without enough power, those processors cannot be operated in the way the plan requires.
Those two constraints, chips and electricity, sit underneath the headline number. A $100 billion AI supercomputer would not be just a single technical build. It would be a supply chain project, an energy project, and a data center project at the same time.
Phase 4 before Phase 5
Stargate is reportedly not the first step. OpenAI and Microsoft are said to be working on a smaller supercomputer known as "Phase 4," expected to be operational in 2026.
Even that smaller system could cost $10 billion when completed, according to the source article. The article notes that this would still be many times the cost of today's data centers.
Stargate is described as "Phase 5." According to one source, Microsoft's decision to proceed with it will depend partly on whether OpenAI can significantly improve the capabilities of its AI and truly achieve some sort of "superintelligence".
That condition makes the project more than a simple infrastructure roadmap. The reported decision point links spending on compute to progress in model capability. In other words, the next stage of infrastructure may depend on whether the current and near-term systems produce enough improvement to justify a much larger build.
The self-improving AI idea
The Information also reported that some OpenAI executives believe more servers could let the company use its existing AI to create the right synthetic training material for better models. The source article describes this as a self-improving AI.
Synthetic training material is significant in this context because it connects infrastructure to model development strategy. If OpenAI can use existing AI systems to help produce training material, then added server capacity could support a feedback loop: more compute enables more generated training material, which may help train better models.
The report also aligns with rumors from OpenAI insider "Jimmy Apples," who recently wrote on X that OpenAI CEO Sam Altman is "building a big special something" "somewhere in the desert." Energy companies Helion Energy and American Clean Power are allegedly involved. The source article adds that "Jimmy Apples" confirmed he was referring to Stargate.
For now, Stargate remains a reported plan based on confidential talks and early cost estimates. But the details point to a clear direction: Microsoft and OpenAI are treating AI infrastructure as a long-term race in chips, power, data centers, and model capability.