Advancing AI Infrastructure for the Era of Agentic AI

This video discusses the evolution of AI infrastructure, focusing on the demands of agentic AI and NVIDIA's latest innovations. It explains how the Vera Rubin platform, along with new CPUs, networking, and software, is designed to handle the complex workloads of AI agents, offering significant performance and efficiency gains.

■ The Shift to Agentic AI
- The changing nature of AI workloads, from simple chat interactions to complex agentic tasks
- The increased demands on compute, memory, and networking for agentic AI

■ Vera Rubin: The Next-Generation AI Platform
- Overview of the Vera Rubin platform and its components (GPU, CPU, networking, storage)
- Performance improvements in AI factory throughput and cost per token

■ Innovations for Agentic Computing
- The Vera CPU for fast tool calling and low-latency performance
- NVLink Fusion for scaling AI infrastructure
- MaxLPS for optimizing data center power and increasing compute density

■ Real-World Impact and Adoption
- Benchmark results and early customer deployments (e.g., Perplexity, Clickhouse)
- Partnerships and availability of the new technologies

This video is for data center operators, cloud providers, and technology enthusiasts interested in the future of AI infrastructure. Viewers will gain insights into the challenges of agentic AI and the strategies NVIDIA is employing to address them, including hardware and software co-design.