Arm has announced its first move into silicon product manufacturing, launching the Arm AGI CPU to target AI data centre workloads.
The company’s new strategy expands its offerings beyond intellectual property licensing and compute subsystems to include chips directly designed by Arm.
The AGI CPU is designed for agentic AI infrastructure, which demands more compute to run always-on AI agents and higher data throughput.
Meta is working as a lead partner and co-developer on the project, planning to integrate the AGI CPU with its applications and custom silicon such as the Meta Training and Inference Accelerator.
Other industry partners, including Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP and SK Telecom have also confirmed plans to use the AGI CPU for tasks such as accelerator management, control plane processing, and cloud application hosting.
According to Arm, demand for CPUs in data centres is rising due to the growing use of agent-driven AI applications.
These systems require an increase in processing capacity without exceeding current power limits.
The AGI CPU features up to 136 Neoverse V3 cores per chip and supports both high-density air-cooled and liquid-cooled server configurations.
This setup enables up to 8,160 cores per rack in air-cooled systems or over 45,000 in liquid-cooled racks.
Arm claims this architecture allows for higher workload density and improved utilisation compared to existing x86-based CPUs while maintaining efficiency within tight power envelopes.
Original equipment manufacturers such as ASRock Rack, Lenovo, Quanta Computer and Supermicro are developing early systems using the AGI CPU.
Broader availability is expected later this year.
Over 50 companies across various sectors, including hyperscale cloud providers, silicon manufacturers, memory suppliers and networking firms, have expressed support for Arm’s entry into silicon products.
Industry leaders backing this initiative include AWS, Broadcom, Google, Marvell, Micron, Microsoft, NVIDIA, Samsung, SK Hynix, and TSMC.
Arm CEO Rene Haas said: “With the expansion into delivering production silicon with our Arm AGI CPU, we are giving partners more choices all built on Arm’s foundation of high-performance, power-efficient computing, to support agentic AI infrastructure at global scale.”
In a related development, Altera has extended its partnership with Arm by integrating Altera’s field-programmable gate arrays (FPGAs) with the AGI CPU.
This collaboration aims to provide system architects with programmable platforms optimised for low latency and scalability in AI data centres.
Altera FPGAs are already present in data centre environments alongside CPUs and GPUs for functions such as data pre-processing and networking. By combining Altera FPGAs with the new Arm processor, both companies intend to enhance real-time performance capabilities across next-generation AI infrastructure.
