Microsoft, Nvidia partner to build AI supercomputer

Microsoft and Nvidia have announced a “multi-year collaboration” to build an artificial intelligence  (AI) supercomputer, which can be used by enterprises to train, deploy and scale AI models. The AI supercomputer will leverage Microsoft Azure’s complete set of computing, networking storage resources along with Nvidia AI enterprise software suite and will be powered by some of the most powerful Nvidia data centre graphics processing units (GPU) such as NVIDIA A100 and H100 GPUs. 

Nvidia said that it will utilise Azure’s supercomputer infrastructure, which includes scalable virtual machine instances optimised for AI training and inference. It will use them to accelerate research in generative AI, which uses machine learning (ML) to generate artificial content from text, images, and videos. Deepfake is an application of generative AI.

Nvidia said that it will also collaborate with Microsoft to optimise DeepSpeed, a deep learning optimisation software suite. Deep learning is a subset of ML that comprises layers of neural networks modeled to simulate the human brain. Nvidia will provide its H100 Transformer Engine for DeepSpeed to accelerate transformer-based models used for large language models and generative AI models.  

“AI technology advances as well as industry adoption are accelerating. The breakthrough of foundation models has triggered a tidal wave of research, fostered new startups and enabled new enterprise applications,” Manuvir Das, vice president, enterprise computing at Nvidia, said in a statement. 

“Our collaboration with Nvidia unlocks the world’s most scalable supercomputer platform, which delivers state-of-the-art AI capabilities for every enterprise on Microsoft Azure,” Scott Guthrie, executive vice president of Cloud and AI Group at Microsoft, said in a statement. 

Nvidia is working on several supercomputer projects. Many of the most powerful supercomputers in the world including IBM’s Summit and Sierra and HPE Perlmutter use Nvidia’s GPUs. Nvidia is also working with Facebook-parent Meta on an AI supercomputer. In January, Meta announced that its research team has built a supercomputer with over 6,000 Nvidia GPUs, making it one of the fastest supercomputers in the world.