Meta sends Arm a friend request asking for help with Nvidia’s Grace CPUs

No custom Arm CPUs to speak of yet

Meta on Wednesday entered into a partnership with Arm Holdings with the aim of helping its software run more efficiently on the British chip designer's CPUs.

The tie-up signals that Arm will also play a bigger role in the datacenters of Facebook and Instagram's parent company, though not necessarily in the way you might expect. Rather than a custom Arm CPU, like the ones that Microsoft, AWS, and Google designed, Meta tells us the partnership will focus on optimizing the Arm-based silicon that it's already deploying.

Like most hyperscalers and cloud providers, Meta is rolling out large quantities of Arm Neoverse cores across its AI datacenters; they just happen to be part of Nvidia's GB200 or GB300 NVL72 rack systems. Each of these racks is equipped with 72 Blackwell GPUs and 36 of Nvidia's Neoverse-V2-based Grace CPUs. 

"Meta's AI ranking and recommendation systems – which power discovery and personalization across Meta's family of apps, including Facebook and Instagram – will leverage Arm's Neoverse-based datacenter platforms to deliver higher performance and lower power consumption compared to x86 systems," the release reads.

The move from x86 boxes to Nvidia's rack systems presents an opportunity for Arm and Meta to optimize existing codebases and frameworks, like PyTorch or Facebook General Matrix Multiplication (FBGEMM) libraries, to take better advantage of the RISC architecture's extensions. 

This includes things like transitioning workloads that previously relied on Intel or AMD AVX2 or AVX-512 to take advantage of modern Arm vector extensions or optimizing PyTorch code to utilize features like ExecuTorch or Arm's KleidiAI platforms.

Adoption of Nvidia's Grace-Blackwell systems has driven a surge in Arm's datacenter market share. A recent Dell'Oro report estimated that Arm CPUs accounted for 25 percent of the datacenter CPU market in Q2.

A year ago, that share hovered around 15 percent, with adoption by major cloud providers its biggest driver. As of 2025, nearly every major US cloud company is renting virtual machines and services powered by custom Arm processors. 

AWS has Graviton, Google Axion, and Microsoft Cobalt. Oracle, meanwhile, continues to deploy Arm-based CPUs from Ampere Computing. As it so happens, on Wednesday, Oracle announced the availability of yet more instances based on the chip vendor's AmpereOne processors.

Given Meta's scale, it would seem like the next obvious candidate for a custom cloud CPU. The company already deploys a number of custom chips for things like video transcoding, as well as its MTIA line of machine learning accelerators, which power its ad recommendation systems. However, at least for the moment, it doesn't seem a custom Arm CPU is in Meta's future. Or if it is, the company isn't ready to talk about it. ®