Nvidia has secured arguably its biggest ever partner for its NVLink Fusion backplane: Arm.

NVLink Fusion, unveiled in May, allows partners to build semi-custom AI infrastructure linking non-Nvidia accelerators with Nvidia hardware in rack-scale solutions.

While the likes of MediaTek, Synopsys, and most recently, Fujitsu, are already onboard, the addition of Arm sees Nvidia add one of the most widely deployed compute platforms across the industry in Neoverse.

According to Arm, its Neoverse offering is projected to reach 50 percent market share across the top hyperscalers worldwide in 2025, with AWS, Google, Microsoft, Oracle, and Meta among those operating on more than one billion cores already deployed.

By integrating Neoverse with NVLink Fusion, Arm aims to strengthen its already dominant architecture, a move the vendor claims will “deliver intelligence per watt, shaping the architecture of the AI era.”

“Extending the Arm Neoverse platform with Nvidia NVLink Fusion brings Grace Blackwell-class performance to every partner building on Arm, a milestone that reflects the incredible momentum we're seeing in the data center,” said Rene Haas, Arm’s CEO.

Building the ‘connective fabric of the AI era’

Nvidia’s NVLink Fusion is an interconnect technology capable of bringing together a multitude of partner hardware, including network interface controllers (NICs), data processing units (DPU), or scale-out switches.

The custom hardware concept allows custom silicon and CPUs to be connected to Nvidia GPUs, with chiplet-based interconnect (UCIe) providing the bridge.

While CPU offerings from Fujitsu and Qualcomm present somewhat niche additions to Nvidia’s ecosystem, adding Arm’s Neoverse could be a real coup.

Neoverse is the basis for a boatload of custom hardware from major cloud providers. Google’s Axion processors, for example, are built on Arm’s Neoverse V2 designs, as is AWS’s Graviton4. Meanwhile, Microsoft’s Azure Cobalt 100 CPUs and Alibaba Cloud’s Yitian 710s use the Arm Neoverse N2 core.

By making it possible for chips built on Arm’s Neoverse designs to be interoperable with Nvidia hardware, NVLink Fusion is cemented as the interconnect standard for AI infrastructure.

“NVLink Fusion is the connective fabric of the AI era, linking every CPU, GPU, and accelerator into one unified rack-scale architecture,” said Nvidia CEO Jensen Huang. “Together with Arm, we’re extending this vision across Neoverse to empower innovators everywhere to design the next generation of specialized AI infrastructure.”

Arm signing up its Neoverse platform to be an NVLink Fusion is poignant given the interconnect’s origins.

The tech was initially designed to interface with AMBA CHI C2C (Coherent Hub Interface Chip-to-Chip), a protocol invented by Arm designed to enable chip-to-chip and chiplet-to-chiplet communication.

In 2023, the design culminated in the development of Nvidia’s Arm-based Grace CPU, which was then brought together with its then-flagship Hopper GPUs to create GH200 Superchips – a CPU+GPU platform capable of some 200 exaflops of processing power.

Nvidia’s NVLink Fusion Arm expansion looks to further dent traditional x86+GPU setups.

In recent years, Arm has represented an existential threat to the long-standing x86 architecture to such a degree that AMD and Intel moved to shore up their platforms' future through a dedicated Ecosystem Advisory Group aimed at improving its interoperability.