OAKLAND, Calif. (Reuters) – U.S. chipmaker Nvidia said on Wednesday it would partner with Microsoft to build large-scale artificial intelligence (AI) computers in the cloud.
The AI computer makes heavy use of NVIDIA’s H100 and A100 image processing semiconductors (GPUs) and runs on Microsoft’s Azure cloud service.
NVIDIA has not disclosed the financial details of the deal, but industry sources say the A100 chips cost around $10,000 to $12,000 each, while the H100 is much more expensive.
The company’s general manager of hyperscale and HPC, Iann Bach, told Reuters that there is a growing desire and need for AI adoption among companies.
Nvidia not only sells chips to Microsoft, but also develops AI models through partnerships.