Collaboration between Microsoft and OpenAI aims to develop a $100 billion AI supercomputer project, featuring "Stargate" set for 2028 launch. (AP)
Microsoft and OpenAI, the parent company of ChatGPT, are collaborating on an ambitious venture that entails the development of a groundbreaking data-center project, potentially amounting to a staggering $100 billion investment. This initiative includes the creation of an artificial intelligence supercomputer dubbed "Stargate," slated for launch in 2028, as disclosed in a recent report.
Requests for comments from both companies regarding the project remained unanswered at the time of Reuters' inquiry.
The surge in the adoption of generative artificial intelligence technology has sparked an unprecedented demand for AI data centers capable of handling more sophisticated tasks compared to conventional data centers.
According to insights shared by individuals familiar with private discussions about the proposal, Microsoft is anticipated to spearhead the financing of the project. The projected cost is estimated to be 100 times higher than some of the most substantial existing data centers, as per sources cited in The Information report.
Reportedly, the envisioned supercomputer, to be situated in the United States, will mark the pinnacle of a series of installations planned by the collaborating entities over the forthcoming six years.
The estimated cost of $100 billion was disclosed by a source who engaged in discussions with OpenAI CEO Sam Altman and another source privy to initial cost approximations by Microsoft. However, the identities of these sources were not disclosed.
Altman and Microsoft personnel have outlined the development of supercomputers across five phases, with "Stargate" representing the fifth phase. Preceding this, Microsoft is actively engaged in crafting a smaller, fourth-phase supercomputer tailored for OpenAI, aiming for a debut around 2026.
Presently entrenched in the third phase of the comprehensive plan, Microsoft and OpenAI are chiefly focusing on the procurement of AI chips, which constitute a significant portion of the anticipated costs for the subsequent phases.
The report highlights the substantial price tags associated with AI chips, with Nvidia CEO Jensen Huang indicating that the latest "Blackwell" B200 artificial intelligence chip is priced between $30,000 and $40,000. In a bid to bolster their AI capabilities, Microsoft had previously unveiled a pair of custom-designed computing chips in November of the preceding year.
Furthermore, the project aims to accommodate chips from various suppliers, ensuring flexibility and scalability.
Frank Shaw, a spokesperson from Microsoft, emphasized the company's commitment to advancing AI capabilities through ongoing infrastructure innovations.
The projected costs for these endeavors could surpass $115 billion, exceeding threefold the capital expenditures allocated by Microsoft in the previous year for servers, infrastructure, and related equipment, according to the report's findings.
