Microsoft and OpenAI are collaborating on a data center project, anticipated to require an investment of up to $100 billion.
The project includes the development of an artificial intelligence supercomputer named “Stargate,” slated for launch in 2028, The Information reported.
The escalating adoption of generative artificial intelligence technology has driven a surge in demand for AI data centers capable of handling more advanced computational tasks than conventional data centers.
Microsoft is likely to finance the project, the report stated. The estimated cost, which surpasses that of many existing data centers by a factor of 100, reflects the ambitious scope of the endeavor.
The proposed supercomputer, based in the United States, would mark the largest in a series of such projects planned by the two companies over the next six years, as per the report.
Although The Information did not disclose the identities of its sources, it attributed the projected $100 billion cost to individuals familiar with the matter, including one who spoke directly to Altman, and another who reviewed initial cost assessments from Microsoft.
The supercomputer initiative, spearheaded by Altman and Microsoft, is structured into five phases, with “Stargate” representing the final phase. A smaller, fourth-phase supercomputer for OpenAI is also in development, expected to debut around 2026.
Presently in the third phase of the plan, Microsoft and OpenAI are allocating a substantial portion of the budget for the next two phases toward procuring the necessary AI chips, which are typically priced at a premium. Nvidia CEO Jensen Huang revealed earlier in March that the latest “Blackwell” B200 artificial intelligence chip would carry a price tag ranging from $30,000 to $40,000.
Microsoft’s introduction of custom-designed computing chips in November last year also underscores its commitment to advancing AI capabilities. The forthcoming project aims to be compatible with chips sourced from various suppliers, providing flexibility and scalability in its implementation.
