System-level ‘coopetition’: Why Nvidia’s DGX Rubin NVL8 runs on Intel Xeon 6

Nvidia has selected Intel’s Xeon 6 processors as the host CPUs for its Nvidia DGX Rubin NVL8 systems. The DGX Rubin NVL8 is part of Nvidia’s next flagship AI system portfolio, designed to help companies accelerate agentic AI adoption.

The DGX Rubin NVL8 systems are designed for large-scale AI workloads, combining eight Rubin GPUs with high-bandwidth memory and interconnects to support high-throughput inference and data movement. The systems are powered by Intel Xeon 6776P processors as host CPUs. The platform also uses NVLink technology to enable fast communication between GPUs for parallel processing.

The Xeon 6 CPU will provide architectural continuity and scalability for GPU-accelerated AI systems as workloads shift toward massive, real-time inference, Intel said during Nvidia GTC 2026.

A split architecture

Analysts say the choice of Intel CPUs is closely tied to enterprise compatibility and deployment requirements.

“As AI shifts toward real-time inference and agentic workloads, the CPU’s role becomes even more critical as managing complex workflows and feeding data efficiently to GPUs can become a bottleneck,” said Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting. Nvidia is optimizing for the best host CPU ecosystem—performance, compatibility, supply, and enterprise readiness—and x86 continues to dominate data center infrastructure. Xeon 6, with its high memory bandwidth (MRDIMM) and strong x86 compatibility, helps ensure GPUs remain fully utilized without data delays, he added.

Enterprise environments still rely heavily on x86 ecosystems for operational tooling, security frameworks, and lifecycle management. “Nvidia is choosing to retain x86 compatibility, which allows enterprises to integrate these systems into existing environments without rearchitecting their entire infrastructure stack. The cost of forcing a new CPU paradigm today would be slower adoption, higher integration risk, and operational friction,” said Sanchit Vir Gogia, chief analyst at Greyhound Research.

Not a strategic alliance

Despite working together at the system level, the relationship between the two companies does not amount to a formal strategic alliance.

“The Intel–Nvidia dynamic is best understood as system-level coopetition. Long-standing collaboration persists across data center and PC ecosystems, with Intel CPUs paired alongside Nvidia GPUs forming standardized AI server architectures and enabling deeper integration,” said Manish Rawat, semiconductor analyst at TechInsights.

However, competition is accelerating structurally.

Even though Nvidia dominates the GPU space, the company is also expanding its presence across more layers of the data-center stack. It has been developing its own CPUs, such as the Grace CPU, aimed at tighter integration between compute, memory, and interconnect. The company has also launched Vera CPU, purpose-built for agentic AI at GTC 2026.

This reflects Nvidia’s broader approach of building more of the system in-house, spanning both hardware and software, even as it continues to incorporate external components where required.

“Nvidia’s push into CPUs (Grace, Vera) and tightly integrated, NVLink-based systems signals a shift toward full-stack ownership spanning compute, networking, and software. This challenges Intel’s traditional dominance in CPUs and system control. In essence, Nvidia is partnering tactically to sustain ecosystem adoption while strategically positioning to displace incumbents and capture greater control of next-generation AI infrastructure,” added Rawat.

On the other hand, Intel is also pushing into the GPU and AI accelerator with offerings such as its Xe-based GPUs and Gaudi accelerators. But it continues to lag behind Nvidia in terms of market adoption and ecosystem maturity.

The use of Intel CPUs in Nvidia’s DGX Rubin NVL8 is strategically important for Intel.

Even as it lags in AI accelerators, Xeon’s presence in Nvidia’s flagship systems keeps Intel embedded in AI infrastructure economics, allowing it to capture value from control-plane and data-movement layers while avoiding full displacement by ARM-based alternatives like Grace. Even though Intel may be losing the GPU battle, it remains relevant in the broader system stack, Jain highlighted.

Nvidia’s $5 billion Intel stake

The development also comes after Nvidia disclosed that it had purchased $5 billion worth of shares in Intel in December 2025. Back then, Nvidia founder and CEO Jensen Huang had stated in the press note that the historic collaboration tightly couples Nvidia’s AI and accelerated computing stack with Intel’s CPUs and the vast x86 ecosystem.

Rawat explained the investment provided balance sheet support to Intel and signalled confidence to markets. But the deeper intent was architectural leverage. By securing tighter alignment with Intel’s x86 ecosystem, Nvidia can drive CPU–GPU co-design across AI infrastructure, data centers, and emerging AI PCs. This acts as a hedge on multiple fronts, reducing dependence on ARM-based platforms, countering AMD in x86, and limiting ecosystem fragmentation.

Experts also view this as a strategic move towards securing long-term control.

“This investment should be viewed as a strategic move anchored in supply chain resilience, manufacturing alignment, and long term ecosystem stability. Availability of advanced packaging, access to fabrication capacity, and geopolitical considerations around semiconductor supply chains are becoming decisive factors. Nvidia’s investment in Intel signals an intent to secure optionality in these areas,” added Gogia.

Jain noted the investment to be tactical for providing securing future supply chain collaboration. While not comparable to Nvidia TSMC deep strategic alliance, it reflects strategic optionality, not integration.