Microsoft Azure has announced a new virtual machine powered by NVIDIA’s top-of-the-line H100 GPU to accelerate generative AI like ChatGPT.
Delivering the promise of advanced AI to our customers requires supercomputing infrastructure, services, and expertise to keep up with the exponential growth in size and complexity of modern models.
Microsoft is meeting this challenge by leveraging decades of supercomputing experience to create an AI infrastructure capable of delivering performance at scale, supporting the largest AI training workloads. . The Microsoft Azure cloud, especially GPU-accelerated virtual machines (VMs), form the basis for many generative AI advances by Microsoft and its customers.
Generative AI applications are evolving rapidly and offer unique value in nearly every industry. From new AI-powered Microsoft Bing and Edge reinventing search to AI-powered support in Microsoft Dynamics 365, AI is rapidly becoming a ubiquitous component of software, and AI infrastructure is out there and on the road.
Our experience in deploying multiple of his ExaOP supercomputers to Azure customers around the world gives us confidence that our infrastructure can deliver true supercomputing performance. For Microsoft and organizations committed to large-scale deployment of Inflection, NVIDIA, OpenAI, and others, this offering enables a new class of large-scale AI models.
AI at scale is in the DNA of Azure. Initial investments in large-scale language model research such as Turing and technological milestones such as building the first AI supercomputer in the cloud prepared us for the moment when generative artificial intelligence was possible.
Azure services such as Azure Machine Learning give customers access to AI supercomputers for model training, and Azure OpenAI services give customers the power of large-scale generative AI models. Scale has always been the North Star in optimizing Azure for AI. We now provide supercomputing capabilities to startups and enterprises of all sizes without the need for large physical hardware or software investments.