On Wednesday, Nvidia introduced a collaboration with Microsoft to construct a “huge” AI-focused cloud pc. It’ll reportedly use tens of hundreds of high-end Nvidia GPUs for functions like deep studying and enormous language fashions. The businesses purpose to make it probably the most highly effective AI supercomputers on the planet.
In flip, the brand new supercomputer will function hundreds of items of what’s arguably the world’s strongest GPU, the Hopper H100, which Nvidia launched in October. Nvidia can even present its second strongest GPU, the A100, and can use its Quantum-2 InfiniBand networking platform, which may switch information at 400 gigabits per second between servers, linking them collectively in a robust cluster.
In the meantime, Microsoft will contribute its Azure cloud infrastructure and ND and NC sequence digital machines. Nvidia’s AI Enterprise platform will tie all of it collectively. The businesses can even collaborate on DeepSpeed, Microsoft’s deep studying optimization software program.
In an announcement, Nvidia talked about the functions the joint supercomputer might serve:
“As a part of the collaboration, Nvidia will use Azure Scalable Digital Machine Situations to additional analysis and speed up developments in generative AI, a quickly rising space of AI the place foundational fashions just like the Megatron Turing NLG 530B are the muse. for unsupervised self-learning algorithms. to create new textual content, code, digital photos, video or audio”.
The previous yr has seen a speedy rise in generative AI fashions similar to Secure Diffusion and DALL-E, which may synthesize novel photos on demand. Comparable fashions have appeared that may create video, synthesize voices, and carry out transcriptions, amongst different makes use of. Because the computational demand for generative AI will increase, Nvidia and Microsoft intend to be there to satisfy it.
As soon as Nvidia and Microsoft’s cloud computing is on-line, clients can deploy hundreds of GPUs on a single cluster to “practice even probably the most huge giant language fashions, construct probably the most complicated recommender programs at scale, and allow generative AI at scale,” in accordance with Nvidia. .
The businesses didn’t present particulars on when the brand new supercomputer can be prepared, however mentioned the announcement marks the beginning of a “multi-year collaboration.” The cloud pc is more likely to improve in capability over time.