On Wednesday, Nvidia, the US chip designer and computing company, announced a collaboration with Microsoft to construct a substantial computer tailored for handling high-intensity artificial intelligence (AI) computing tasks in the cloud. This AI-powered computer is slated to operate on Microsoft’s Azure cloud platform, leveraging tens of thousands of graphics processing units (GPUs), including Nvidia’s potent H100 and A100 chips. While the financial details of the partnership were not disclosed by Nvidia, industry insiders estimated the cost of each A100 chip to be between $10,000 and $12,000. Ian Buck, Nvidia’s general manager for Hyperscale and HPC, noted the increasing adoption of AI in the enterprise and emphasized the significance of providing accessible AI services for business applications.
In addition to supplying Microsoft with the necessary chips, Nvidia revealed plans to collaborate with the software and cloud giant in developing AI models. Buck stated that Nvidia would also utilize Microsoft’s AI cloud computer as a customer, actively participating in the creation of AI applications to deliver services to end-users. The soaring demand for faster and more powerful computing infrastructure, driven by the rapid growth of AI models like those employed in natural language processing, has become increasingly evident.
Furthermore, Nvidia highlighted that Azure would be the inaugural public cloud platform to integrate its Quantum-2 InfiniBand networking technology, boasting a remarkable speed of 400Gbps. This high-speed networking technology plays a crucial role in efficiently linking servers, a critical requirement for the collaborative functioning of thousands of chips across multiple servers, particularly in heavy AI computing workloads.
