Arm announced on Monday that CPUs built on its technology will be integrated with AI chips using Nvidia's NVLink Fusion technology. The move will make it easier for customers of both companies who prefer custom solutions for their infrastructure, especially hyperscale data center operators, to pair Arm-based Neoverse CPUs with Nvidia's dominant GPUs.


It’s the latest example of Nvidia partnering with nearly every major tech company through collaborative deals — and now Nvidia is at the heart of the artificial intelligence industry. The announcement indicates that Nvidia is opening up its NVLink platform to integrate with a variety of custom chips, rather than forcing customers to use its own CPUs.

Currently, NVIDIA has launched an artificial intelligence product called "Grace Blackwell", which combines multiple GPUs with an NVIDIA-branded Arm architecture CPU. Other configurations include servers with Intel or AMD CPUs.

However, Microsoft, Amazon and Google are all developing or deploying Arm-based CPUs in their cloud businesses to have better control over hardware configurations and reduce costs.

Arm doesn't make the CPUs itself, but licenses the instruction set technology needed for its chips. In addition, the company also sells chip design solutions to help partners develop chips based on the Arm architecture more quickly.

As part of Monday's announcement, Arm said the custom Neoverse chips will feature a new protocol that enables seamless data transfer between the chip and the GPU.

Historically, the CPU has been the most important component in a server. But generative AI infrastructure centers around AI acceleration chips, which in many cases are NVIDIA GPUs. In an AI server, you can use up to 8 GPUs with 1 CPU.

In September this year, Nvidia announced that it would invest $5 billion in leading CPU manufacturer Intel. One of the core purposes of the deal is to enable Intel CPUs to be integrated into artificial intelligence servers through Nvidia's NVLink technology.

In 2020, Nvidia reached an agreement to acquire Arm for $40 billion, but the deal failed in 2022 due to regulatory issues in the United States and the United Kingdom. As of February, Nvidia still held a small stake in Arm, which is majority owned by SoftBank.

Meanwhile, SoftBank sold out all of its Nvidia shares earlier this month and is providing support for OpenAI’s “Stargate Project.” In addition to using chips from Nvidia and AMD, the plan also plans to use Arm technology.