Arm announced at the 2025 Supercomputing Conference (Supercomputing ’25) that it has joined the NVLink Fusion ecosystem in collaboration with Nvidia. This marks a significant advancement for the technology, with the number of core microarchitecture developers supporting it reaching 2 and a total of 4 CPU developers. For Nvidia, this means that Arm customers will be able to develop processors that can work in conjunction with Nvidia’s AI accelerators; at the same time, Arm can also design CPUs that compete with Nvidia’s proprietary processors or Intel’s processors in Nvidia-based systems.
“Arm is integrating NVLink intellectual property (IP) to enable its customers to create CPU system-on-chips (SoCs) that can connect to Nvidia’s GPUs,” Dion Harris, Nvidia’s head of data center product marketing, stated, “With NVLink Fusion, large-scale data center operators can significantly reduce design complexity, save development costs, and accelerate time to market. The addition of Arm customers provides more options for specialized semi-custom infrastructure.”
Arm is a diversified large enterprise, covering instruction set architecture (ISA) and intellectual property (IP) licensing, as well as developing customized CPUs and system-on-chips (SoCs) for large clients. Supporting NVLink Fusion brings corresponding advantages for each type of business.
Advantages as an IP Provider
By supporting NVLink Fusion, Arm gains a significant new competitive edge in the data center market. By directly integrating NVLink IP into its architecture product portfolio, Arm can provide a ready path for licensees to create CPUs that can natively access Nvidia’s AI accelerator ecosystem. Theoretically, this makes Arm ‘s architecture design more attractive to two types of customers: large-scale data center operators looking to customize CPUs, and sovereign cloud builders needing compatibility with market-leading Nvidia GPUs to support AI and high-performance computing (HPC). Previously, the only processors that supported connecting to Nvidia’s GPUs via NVLink were Nvidia’s own Grace CPUs.
Advantages as a CPU Developer
Although Nvidia only mentioned Arm ‘s role as an IP provider, Arm as a CPU developer targeting large-scale data center operators and sovereign entities can also benefit. Specifically, Arm gains the ability to compete directly in Nvidia-based systems. With native NVLink Fusion integration, future Arm -designed server CPUs can compete head-to-head with Nvidia’s Grace, Vera processors, and Intel’s Xeon processors in systems centered around Nvidia’s GPUs. If Nvidia grants permission (which is still uncertain), Arm CPUs could become core participants in rack-level NVLink solutions.
Advantages as an ISA Licensor
Supporting NVLink Fusion also strengthens Arm ‘s position as an ISA licensor — this makes Arm ‘s architecture more attractive to two groups: large-scale data center operators and chip designers looking to create customized CPUs that are deeply integrated with Nvidia’s GPUs. By ensuring that Arm ‘s architecture CPU designs can work with the consistent NVLink architecture and Nvidia’s GPUs (rather than being limited to PCIe interfaces), Arm gains ecosystem appeal and relevance for the future, which is currently unmatched by competing architectures like x86 and RISC-V. This undoubtedly poses risks for AMD and Intel: AMD has little interest in supporting NVLink , while Intel will take years to develop a custom Xeon CPU that supports NVLink . However, considering chip development cycles, by the time the Arm architecture CPUs supporting NVLink are ready, Intel’s custom Xeon CPUs may also have been released.
Benefits for Nvidia
Arm supporting NVLink Fusion can significantly expand the pool of CPUs that can natively adapt to Nvidia’s dominant AI systems without Nvidia having to create all these CPUs. By allowing companies like Google, Meta, and Microsoft to directly integrate NVLink into their SoCs, Nvidia ensures that future Arm architecture processors are either designed around Nvidia’s GPUs or at least compatible with them. On one hand, this may weaken the appeal of open alternatives like UALink; on the other hand, it will overall reduce the competitiveness of companies like AMD, Broadcom, and Tenstorrent in the AI accelerator market.
Additionally, this can strengthen Nvidia’s position in sovereign AI projects that adopt Arm CPUs (at least for the next few years): governments and cloud service providers looking to use customized Arm CPUs for control plane or data loading tasks can now achieve this without leaving the Nvidia GPU ecosystem.
In summary, Arm joining the NVLink ecosystem is a win-win for Arm, Nvidia, and their many partners, but it may pose significant risks for AMD, Intel, and Broadcom.
Welcome to join the EETOP WeChat group
Registration for the AMD Technology Summit: FPGA/SoC+x86 Embedded
