Microsoft and Nvidia have a new collaboration. It’s about building a “massive” AI-driven cloud computer.
MICROSOFT AND NVIDIA: an extreme adventure
A new partnership is concocted between the supercomputer IA Cloud Microsoft and Nvidia. In fact, the two companies plan to manufacture a AI-driven “massive” cloud computer. The product would then have tens of thousands of high-end Nvidia GPUs. And to power applications like deep learning and large language models. That said, the two computer giants aim to make this new product one of the most powerful AI supercomputers in the world.
An ultra-powerful supercomputer: characteristics
The new supercomputer will then consist of thousands of units, which is the most powerful GPU in the world. It’s the Hopper H100deployed recently by Nvidia. The latter will also provide its second most powerful GPU: the100. And that’s not all, it will also implement its network platform Quantum-2 Infiniband. This one which is able to transfer data at 400 gigabits per second between the servers.
For its part, Microsoft is bringing back its infrastructure azure cloud and series virtual machines ND and NC. From then on, Nvidia’s AI Enterprise platform will tie it all together. In addition, Microsoft and Nvidia will work together on DeepSpeedMicrosoft’s deep learning optimization software.
What are the advantages ?
As part of the collaboration, Nvidia said in a statement, the company will use the instances scalable virtual machines from Azure. The goal here is to research and accelerate the evolution of generative AI. This is a burgeoning field of AI, where fundamental models like Megatron Turing NLG 530B are the basis of unsupervised self-learning algorithms. In this case then, the supercomputer will be able to create not only new text and code, but also digital images, video and audio.
Indeed, generative AI models have rapidly increased since last year. This is for example the case of Stable Diffusion and DALL-E who can synthesize new images on demand. Also, similar models have appeared that can create videos, synthesize voices and perform transcriptions. However, despite its different models, the demand continues to increase for generative AI. Although, it is the ultimate goal of Microsoft and Nvidia: to have a solution to answer it.
Customers will then be able to enjoy the benefits of this new cloud computer from Microsoft and Nvidia as soon as it becomes available. In particular, they will be able to deploy thousands of GPUs in a single cluster. According to Nvidia, they will even be able to “train the most massive large language models, create the most complex recommender systems at scale, and enable generative AI at scale.”
Nevertheless, until today, Microsoft and Nvidia have not provided details on the application date of the new supercomputer. Instead, they said the announcement marks the start of a “multi-year collaboration.”