Accelerating AI Innovation With Intel Xeon

Author: James McFaden, Supply Chain Quality & Engineering, Lenovo; Milan, Intel; David Handley, Supply Chain, Lenovo Collection: 2025 OCP Global Summit - Expo Hall Sessions

Summary

This joint Lenovo-Intel presentation covers how Intel Xeon 6 CPUs serve as critical host processors in GPU-based AI servers, alongside Lenovo's vertically integrated manufacturing and supply chain capabilities for rapid deployment. The SR680A V3 (with B200 GPUs and 5th-gen Intel CPUs) is currently shipping at hundreds of units per week, while the SR680A V4 with Intel Xeon 6 and B300 GPUs ships from early December 2025. The liquid-cooled SR780A V3 pulls only 80 kW fully loaded and enabled a 20-rack, 1,000+ GPU cluster to go live within 3 weeks of delivery.

Intel positions the Xeon 6 as essential for AI server workloads including data preprocessing, checkpointing, KV cache offload from GPU to system memory (critical as context sizes grow), and resource management for agentic AI. Xeon 6 offers 32 DIMM slots (versus 24 from competitors), MRDIMM support for higher memory bandwidth, Priority Core Turbo for optimized GPU data transmission, leading PCIe 5.0 support (up to 192/176 lanes 2S, 136 1S), built-in AMX AI acceleration, and Intel TDX for confidential computing. Lenovo operates 30+ factories globally, was ranked #8 in Gartner's 2025 Global Supply Chain, and is bringing up a new factory in Whitsett, North Carolina alongside existing facilities in Monterrey, Mexico and Budapest, Hungary.