At GTC 2024, we spoke with John Lee, Microsoft Azure’s Principal Lead, AI Platforms & Infrastructure, about the incorporation of NVIDIA’S new Blackwell GPU architecture within Azure’s AI capabilities portfolio. John is not only a long-time HPC veteran, he’s also an AI and cloud computing enthusiast.
Here he talks about Microsoft as one of the first organizations to bring the power of NVIDIA Grace Blackwell GB200 and advanced NVIDIA Quantum-X800 InfiniBand networking to cloud, delivering trillion-parameter foundation models for natural language processing, computer vision, speech recognition and more.
Citing his two decades of experience in HPC (prior to Microsoft, he was at Intel and Cray), he said, “What makes HPC so different from rest of the industry is scalability, how big you run a job. And in AI, that directly translates because when you’re trying to model these incredibly large and complex large language models, scalability matters. So if you take a look at Azure, we’ve been building up this AI supercomputing infrastructure capability for many, many years.”
