HPC Infrastructure for AI Outcomes: Learning at Large Scale

Microsoft Azure claims 5 entries in the top 50 supercomputers in the world (Nov 2022). These entries are each smaller parts of larger systems, at separate sites. As a part of public cloud infrastructure, they are used by customers all over the world for running both HPC and large AI models. They are also used by Microsoft itself! Our partnerships (such as with NVIDIA, and OpenAI) help us to run some of the largest AI models in the world, and bring this intelligence directly into the entire suite of Microsoft products, from Windows to Office to Bing (and Azure itself, too). Come & learn some of what we have learned from training, deploying, and running the world’s largest AI models, on some of the world’s largest HPC infrastructure - and our thoughts on how this relates to sustainability & powering the future.