Arm and Meta have a long-term partnership to improve AI software on Arm Neoverse chips, but reports do not confirm a new Arm-made CPU or Meta as first buyer.
In short: Reports do not support the claim that Arm is making its first in-house CPU with Meta as the first customer, but they do describe a multi-year partnership to make AI run better on Arm-based systems.
Some recent coverage has suggested Arm is producing its own CPU (a computer’s main “brain” chip) for the first time, and that it built it with Meta, with Meta as the first buyer.
However, other current reports about Arm and Meta’s relationship describe something different. In October 2025, Arm and Meta announced a multi-year strategic partnership focused on improving how Meta’s AI software runs on Arm’s existing Neoverse CPU designs. Neoverse is Arm’s server chip design used by partners, not a physical chip Arm manufactures itself.
In this partnership, Meta plans to use Arm Neoverse cores in its data centers. This includes Nvidia Grace CPUs, which are based on Arm Neoverse designs, inside Nvidia GB200 and GB300 systems. The goal is better performance per watt, meaning more work for the same electricity, for AI tasks like ranking and recommendations in apps such as Facebook and Instagram.
Arm and Meta also said they would work on speeding up Meta’s open-source AI tools, including PyTorch and vLLM, on Arm-based hardware. This is like tuning the engine and the gearbox so the same car runs faster and uses less fuel.
For most people, this is about cost and power use. If big services can run AI with less electricity, they can scale features to more users and expand data centers with less strain on power grids.
Source: TechCrunch AI
54
Productivity & Workflow43
AI Infrastructure & MLOps31
Software Development42
Automation & Workflow34
Data & Analytics25
Marketing & Growth27
Voice & Speech25
Photography & Imaging26
Customer Support18
Writing & Content Creation26
Operations & Admin13
Design & Creative19
Sales & Outreach17
Research & Analysis16