According to a recent report, Microsoft and OpenAI will construct a $100 billion AI supercomputer named Stargate
Microsoft and OpenAI are collaborating to develop an AI supercomputer named “Stargate,” with an estimated cost of up to $100 billion. The Information reported this on Friday, citing sources who spoke to Sam Altman and others who reviewed initial cost estimates. The project, which could cost over 100 times more than some of the largest data centers, is expected to be completed by 2028, with Microsoft likely financing the endeavor.
According to the report, Stargate will be the largest in a series of supercomputers that Microsoft and OpenAI intend to construct over the next six years. The supercomputer initiative is presently in its third phase, with Stargate slated for the fifth and final phase. Acquiring AI chips is expected to account for a substantial portion of the cost for the upcoming phases.
There is currently a bottleneck in the AI development process due to a shortage of graphics processing units (GPUs), essential for processing data in AI models. The surge in demand for these chips, driven by the boom in generative AI, has caused supply shortages. Additionally, Nvidia, a major manufacturer of high-performance AI chips, essentially holds a monopoly on the market. However, it is facing challenges in meeting the growing demand, leading to delays in fulfilling orders. Furthermore, the cost of Nvidia’s chips is continuously rising.
Following Nvidia’s GTC 2024 event, CEO Jensen Huang confirmed that the company’s newest AI chip, the Blackwell B200, will be priced between $30,000 and $40,000 per unit. Nvidia has invested over $10 billion in the research and development of this chip, which significantly outperforms all other chips currently available on the market.
However, several technology companies, including Intel, Qualcomm, Google Cloud, Arm, and Samsung, have joined forces to establish the “United Acceleration Foundation.” This open standard accelerator programming model seeks to compete with Nvidia’s software and hardware dominance in AI.