Microsoft and NVIDIA join forces to build a massive AI supercomputer in the cloud

  • 15

We live in a world where artificial intelligence (AI) plays a critical role in robotics, video games, autonomous driving, programming, and natural language processing, among other things. But for AI to continue advancing and begin to gain ground in other areas, it is necessary to have supercomputers with great computing power and the right software to make the most of its capabilities.

Companies like Tesla and Meta can allocate millions of dollars to build their own AI supercomputers to develop their projects. They are huge companies and the investment could be justified. Microsoft and NVIDIA, for their part, propose an alternative. Why spend millions of dollars building an AI supercomputer when you can pay to use AI infrastructure in the cloud?

Microsoft and NVIDIA team up to build a tremendously powerful AI supercomputer

The two US technology companies announced on Wednesday a multi-year collaboration to build “one of the most powerful AI supercomputers in the world”. There are not too many details of this future technological beast, but it is known that AI technologies from Microsoft and NVIDIA will be combined to make the capabilities of the supercomputer available in the cloud.

At the hardware level we will find the latest in technology. The Microsoft and NVIDIA supercomputer will incorporate “tens of thousands” of NVIDIA H100 GPUs, a technological prodigy built on TSMC’s 4nm photolithography with more than 80 billion transistors. There will also be older generation NVIDIA A100 GPUs with 54 billion transistors.

The icing on the cake will be NVIDIA’s InfiniBand interconnection architecture in its seventh generation. The devices Quantum-2 They promise to provide Remote Direct Memory Access (RDMA) and advanced acceleration engines to support high workloads. In this case, a bandwidth up to 400 Gb/s. All of this will work with the NVIDIA AI Enterprise operating system.

Quantum 2 Nvidia

As we said, this will be an AI supercomputer in the cloud. This means that through Microsoft’s Azure, customers will be able to use different scalable virtual machine instances for their projects. The companies point out that the main strength of this system will lie in providing tools to develop generative AI, which is key for digital image, code, video and audio projects.

Frontier, the world's most powerful new supercomputer, has reached a milestone: breaking the exascale barrier

For its part, Microsoft will optimize its DeepSpeed ​​library in order to reduce computing power and memory usage during AI training workloads. For now, it is unknown how much money these companies will invest in the supercomputer and from what date it will start working. It is time to wait to find out how this project will progress and if it will finally meet its objectives.

Images: Microsoft | NVIDIA

In Xataka: Europe will have an exascale supercomputer: Jupiter, the machine that will rival the best of China and the US

We live in a world where artificial intelligence (AI) plays a critical role in robotics, video games, autonomous driving, programming,…

We live in a world where artificial intelligence (AI) plays a critical role in robotics, video games, autonomous driving, programming,…

Leave a Reply

Your email address will not be published.