Intel and NVIDIA Join Forces: NVIDIA Integrates Intel Xeon 6 Processors into the Rubin Platform

A new phase in the partnership between Intel and NVIDIA is opening the door to more powerful AI infrastructure. This is happening amidst the rapid rise of agentic AI.

Why NVIDIA Chose Intel’s x86 Architecture

At NVIDIA GTC 2026, the company officially announced the integration of Intel Xeon 6 processors into its flagship DGX NVL8 Rubin systems. This strategic move is designed to offload certain workloads from GPUs by shifting them to more capable CPUs.

The Rubin systems will be powered by the Xeon 6776P processor based on performance cores (P-cores). It offers 64 cores and 128 threads for parallel workloads, supports PCIe 5.0 and MRDIMM memory, and provides hardware-level isolation via Intel Trust Domain Extensions (TDX).

This combination improves performance per watt while also reducing overall system costs thanks to Intel’s open software ecosystem.

Additional Chip — NVIDIA Groq 3 LPU

The Rubin platform will also feature the new NVIDIA Groq 3 LPU chip designed for accelerated inference. It uses 500 MB of ultra-fast SRAM instead of traditional HBM and delivers up to 150 TB/s of bandwidth per chip, significantly increasing token throughput and reducing latency.

At present, the Rubin platform consists of six key components: Rubin GPU, Vera CPU, NVLink6 scale-up switches, ConnectX-9 smart NIC, BlueField-4 DPU, and Spectrum-X scale-out switch with integrated optics.

Vera Rubin in Space

The most ambitious part of the presentation was the Vera Rubin space module, designed for deployment in orbital data centers. It delivers up to 25 times more performance compared to previous solutions.

This technology is already being adopted by companies such as Axiom Space and Planet Labs PBC.