The new multi-year collaboration aims to build an AI supercomputer to handle some of the most sophisticated AI models.
NVIDIA will be collaborating with Microsoft to build one of the most powerful supercomputers dedicated to artificial intelligence training and inference.
The multi-year partnership will see NVIDIA supply tens of thousands of DGX A100 and H100 GPUs, with software optimized to get the best performance out of AI workloads.
Through the use of Microsoft Azure’s scalable virtual instances, NVIDIA will be able to run large-scale research projects to further accelerate the development of AI.
In particular, NVIDIA sees the supercomputer as a benefactor to the emerging field of generative systems, which have been popularized with the launch of OpenAI’s ChatGPT and DALL-E.
“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,” said Manuvir Das, vice president of enterprise computing at NVIDIA. “Our collaboration with Microsoft will provide researchers and companies with state-of-the-art AI infrastructure and software to capitalize on the transformative power of AI.”
Of course, generative systems are just one area of AI, and we expect the supercomputer will be used for a lot more projects. NVIDIA is a key partner with many self-driving vehicle operators, so we suspect that this supercomputer will be helpful in running massive simulations to teach quickly teach autonomous vehicles how to drive and edge cases to spot.
“AI is fueling the next wave of automation across enterprises and industrial computing, enabling organizations to do more with less as they navigate economic uncertainties,” said Scott Guthrie, executive vice president of the Cloud and AI Group at Microsoft. “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.”
As artificial intelligence research grows, as does the requirements needed to have another breakthrough. Already, we see much of the infrastructure available even five years ago is not enough to power some of the largest foundational models, which are the backbone of the generative systems and other recent AI pursuits.
This new supercomputer looks to push infrastructure once again into the driver’s seat, although with the ever increasing need for more resources, it may not take long for AI models to require even more juice than NVIDIA’s supercomputer provides.