Keynote: gRPC, A Foundation for AI
summary
Claims concrete wins: 400 gig interconnect for faster ingestion, GKE clusters up to 65,000 nodes and an RDMA VPC at 3.2 Tbps for GPU/TPU hops, an inference gateway that reports 60% lower latency and 40% higher throughput, and an intelligent traffic manager that claims ~30% lower TCO by boosting GPU utilization. MCP and A2A get gRPC transport (replacing JSON‑RPC/HTTP), gRPC runs on serverless/cloud run, and a demo shows TLS and model‑armor/third‑party firewall integration without proxy boxes.
AI-written summary · about
chapters (4)
- 0:00The future of gRPC in AI
- 15:32Agentic AI architecture
- 16:54The gRPC agent mesh
- 19:19Demonstrating gRPC with MCP
related talks
06mKeynote: gRPC Project Update
39mKeynote: gRPC Opening Keynote - Abhishek Kumar, Google & Arunkumar Jayaraman, Google
16mKeynote: Overview of gRPC
20mKeynote: Overview of gRPC
19mKeynote: What's New in gRPC?
43mKeynote: Building AI in the Open: How the Cloud Native Community Shapes What Comes Next
11mKeynote: State of gRPC
09mKeynote: Kubernetes for GenAI: The New Operating System of the AI-First World