Arm Joins Nvidia NVLink Fusion Ecosystem for AI Chips

At the Supercomputing '25 conference, Arm Holdings announced it has joined the NVLink Fusion ecosystem to integrate Nvidia's high-speed interconnect technology into its Neoverse platform for AI data center chip designs. This partnership enables Arm's Neoverse CPUs and those designed by its licensees—including hyperscalers such as Google, Meta, Microsoft, and Amazon—to connect natively with Nvidia GPUs via a coherent fabric instead of relying on PCIe.

The integration allows customers to build custom system-on-chips (SoCs) that link directly to AI accelerators, which reduces design complexity and development costs. For Arm, the move is a strategic effort to expand its footprint in the server market, challenging the historical dominance of Intel Xeon and competing against x86 and RISC-V architectures. This shift aligns with a broader strategy by Arm to offer more complete chip designs to its customers.

For Nvidia, the agreement expands the range of compatible CPUs for its AI-centric systems without requiring the company to develop all the hardware internally. Having reached a similar deal with Intel in September, Nvidia is attempting to solidify NVLink as the industry standard for AI infrastructure and limit the appeal of open alternatives like UALink.