Arm has launched its first-ever data center CPU, designed to support “agentic AI infrastructure.”

Dubbed the Arm AGI CPU and co-developed with Meta, it is the first time in Arm’s 35-year history that the company has produced its own chip, a departure from its traditional business model of designing and licensing out its chip IP to companies such as Apple, Google, Microsoft, Nvidia, and Qualcomm.

“AI has fundamentally redefined how computing is built and deployed. Agentic computing is accelerating that change,” said Rene Haas, CEO, Arm. “Today marks the next phase of the Arm compute platform and a defining moment for our company. 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.”

Comprising up to 136 Arm Neoverse V3 cores running up to 3.7GHz across two dies, the CPU has a 300W TDP and supports 12 channels of DDR5 memory with speeds of up to 8,800Mtps or 6Gbps per core. For I/O functions, the chip comes equipped with 96 lanes of PCIe 6.0 connectivity and supports CXL 3.0.

Built on TSMC’s 3nm process node, Arm AGI is capable of delivering more than double the performance per rack when compared with x86 platforms in certain configurations, Arm claimed, caveating that this figure is based on the company’s own internal estimates rather than official benchmarks.

Arm said it has validated two different OCP rack designs – a 36kW air-cooled system with 30 compute blades, totaling 8,160 cores per rack, and a 200kW liquid-cooled server capable of housing 336 Arm AGI CPUs for over 45,000 cores. The latter system has been developed in partnership with Supermicro, with ASRock Rack, Lenovo, and Quanta Computer also working with Arm.

Meta will deploy the Arm AGI CPUs alongside its own custom Meta Training and Inference Accelerator (MTIA), and Arm said the partners have committed to collaborating across multiple generations of the Arm AGI CPU roadmap.

In addition to Meta, other launch partners for the CPU include Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom.

“Delivering AI experiences at global scale demands a robust and adaptable portfolio of custom silicon solutions, purpose-built to accelerate AI workloads and optimize performance across Meta’s platforms,” said Santosh Janardhan, head of infrastructure, Meta. “We worked alongside Arm to develop the Arm AGI CPU to deploy an efficient compute platform that significantly improves our data center performance density and supports a multi-generation roadmap for our evolving AI systems.”

Early systems featuring the Arm AGI CPU are available now, with broader availability expected in the second half of the year.