NVIDIA and Intel deepen partnership with $5B investment, NVLink-enabled x86 plans, and flexible chipmaking strategy
NVIDIA’s latest partnership with Intel is more than a headline-making investment. Following a joint webcast between the companies’ leaders, the two outlined how the deal will reshape AI infrastructure across both ARM and x86 platforms, without altering NVIDIA’s existing ARM roadmap. NVIDIA confirmed it invested $5 billion in Intel’s common stock as part of a broader collaboration that spans datacenter and consumer CPU segments.
The technical centerpiece is all about scale. Today, NVIDIA can build NVLink 72 rack-scale systems using its own ARM-based Vera CPU, enabling massive, tightly coupled AI supercomputers. By contrast, x86 datacenter CPUs that connect over PCIe typically top out at NVL8 configurations, limiting scale-up potential. The plan is to integrate NVLink directly with Intel’s datacenter CPUs, bringing x86 into the same high-bandwidth, rack-scale ecosystem. NVIDIA estimates this unlocks a $30 billion untapped market by offering customers true choice between ARM and x86 architectures for large-scale AI deployments.
This collaboration is designed to flow both ways. NVIDIA expects to be a primary consumer of Intel’s datacenter CPUs, while also supplying RTX GPU chiplets that Intel plans to use in a PC-oriented system-on-chip. That means deeper co-design across platforms, with each company contributing core technology to the other’s product stack.
On manufacturing, NVIDIA emphasized that its ARM roadmap remains unchanged and that TSMC continues to be the primary foundry for upcoming chips. The company acknowledged ongoing work with Intel Foundry Services, but signaled that near-term production will stay anchored with TSMC. Even so, Intel could play a meaningful role on the advanced packaging front. Technologies like Foveros are being considered to combine RTX GPU chiplets with Intel CPU chiplets for PC SoCs, highlighting how packaging innovation may be the first area where the partnership shows up in shipping products. Longer term, process nodes such as 18A and 14A could influence that calculus, but the companies indicated they are comfortable with the current split: TSMC for core fabrication, Intel for potential packaging collaboration.
What this means for customers and the market is straightforward. Expect rack-scale AI systems that bring NVLink’s advantages to both ARM and x86, broader platform choice for hyperscalers and enterprises, and a faster path to heterogeneous PC chips that marry high-performance CPU and RTX GPU chiplets. The strategy balances ecosystem flexibility with manufacturing pragmatism—keeping NVIDIA’s proven ARM plans intact while expanding x86 capabilities through tight integration with Intel.
