Gilad Shainer, NVIDIA | theCUBE + NYSE Wired: AI Factories - Data Centers of the Future
Show Notes
In this episode of theCUBE, Gilad Shainer, NVIDIA’s Senior Vice President of Marketing, discusses the pivotal role of networking in shaping the future of AI technology and data centers. Joined by host John Furrier at the New York Stock Exchange, Shainer emphasizes how networking frameworks like InfiniBand and Spectrum-X are becoming fundamental to AI infrastructures, effectively acting as the "operating system" for supercomputers. He details the importance of low-latency, jitter-free interconnects and explains how extreme co-design between hardware and software leads to optimally integrated AI systems. Furthermore, the conversation dives into the significance of density, energy efficiency, and reliability in building next-generation AI factories, covering innovations like liquid-cooled GPUs and co-packaged optics. Shainer also elaborates on the transformative roles of devices like the BlueField-4 data processing unit in edge deployments and storage systems, illustrating the evolving landscape of distributed AI technologies.
Key Topics Covered:
- The concept of networking as the "operating system" for AI-scale infrastructures.
- Importance of low-latency and jitter-free interconnects for effective GPU communication.
- Extreme co-design principles integrating hardware and software for optimal performance.
- Role of density and energy efficiency in the design of AI factories.
- Innovations in GPU cooling and connectivity to enhance data center performance.
- Functions of BlueField-4 as a vital component for storage, security, and edge computing.
- Partnerships with major tech companies like Microsoft, Meta, and Cisco to enhance AI solutions.
- Future visions around the convergence of networking, AI, and edge computing.
