Anush Elangovan, AMD | theCUBE + NYSE Wired: AI Factories - Data Centers of the Future

In this episode of theCUBE + NYSE Wired, Anush Elangovan, vice president of AI Software at AMD, joins host John Furrier to discuss how AI factories are revolutionizing modern data centers into supercomputers. Elangovan outlines AMD's multi-generational platform that encompasses EPYC CPUs, Instinct GPUs, Pensando NICs, and the ROCm software stack. He emphasizes the importance of GPU memory capacity and bandwidth for handling larger AI models, showcasing the features of AMD's Helios rack-scale system. The discussion highlights the role of open ecosystems and community engagement, particularly through AMD’s upcoming AI DevDay. Elangovan explains the significance of a comprehensive software architecture in ensuring reliable and efficient operations across various environments—cloud, on-premises, and edge. He articulates AMD's vision of integrating sophisticated hardware and software tools to create a robust AI platform that supports enterprises of all sizes.

• The transformation of data centers into AI factories and supercomputers.
• Overview of AMD's multi-generational platform with EPYC CPUs and Instinct GPUs.
• Importance of GPU memory capacity and bandwidth in AI model performance.
• Details of AMD’s Helios rack-scale system, including specifications.
• AMD's commitment to open ecosystems and community engagement initiatives.
• The role of ROCm as an open-source software framework in AI development.
• Strategies for ensuring reliability and performance in AI-scale operations.
• Future events including AI DevDay and AMD's ongoing developer engagement plans.