Intel CEO Lip-Bu Tan revealed a new collaboration with NVIDIA during a Q4 2025 earnings call. The partners are developing a custom Xeon CPU that fully integrates NVIDIA NVLink technology. This aims to deliver superior x86 performance for AI host nodes, marking a deepening of their enterprise partnership beyond the previously announced RTX iGPU project.

This move signals Intel's aggressive push into the AI datacenter market, where NVIDIA's GPUs currently dominate. By integrating NVLink directly into a Xeon design, Intel could offer a tightly-coupled CPU-GPU solution that reduces bottlenecks in large-scale AI training and inference workloads, challenging AMD's EPYC platforms in the process.

The custom silicon is part of a broader server roadmap simplification, which also includes the upcoming 16-channel 'Diamond Rapids' Xeon. While Tan didn't specify the underlying Xeon generation, the collaboration is expected to influence future designs like 'Coral Rapids,' which is rumored to reintroduce Hyper-Threading for datacenters by late 2027.