For the First Time Ever, Arm Has Built a Chip of Its Own

British chip designer Arm has released its first production-ready silicon processor, breaking a 35-year tradition of licensing its intellectual property exclusively. 

The company introduced the Arm AGI CPU during its “Arm Everywhere” event on March 24, announcing a co-development partnership with Meta, the parent company of social media platforms, to build hardware for large-scale artificial intelligence data centres.

The Arm AGI CPU utilises the company's Neoverse V3 cores to manage continuous AI workloads, known as agentic AI, where software agents operate autonomously. 

The processor functions as the orchestration layer in computing facilities, managing memory, scheduling tasks, and directing hardware accelerators. 

According to Arm's hardware specifications, a single server blade in the company's reference design contains two AGI CPU chips, delivering a total of 272 cores. 

(Image: Arm)

This specific per-chip compute density is the foundational mechanism allowing Arm to achieve its target of 8,160 cores within a standard 30-blade, air-cooled rack setup.

A standard air-cooled 36-kilowatt rack configuration houses 30 server blades containing 8,160 cores. Arm has also detailed a liquid-cooled 200-kilowatt setup built with hardware manufacturer Supermicro that accommodates 336 processors and over 45,000 cores. 

This compute density directly addresses the physical power and space constraints of modern server facilities, alleviating the operational bottleneck in global AI infrastructure. Arm estimates the AGI CPU configuration yields double the rack-level performance of current x86-based systems.

Meta serves as the lead co-developer and will deploy the processors alongside its proprietary MTIA AI accelerators to support its gigawatt-scale infrastructure. 

Meta Head of Infrastructure Santosh Janardhan noted in the announcement that the jointly developed hardware “significantly improves our data center performance density”. 

The collaboration represents a strategic business expansion for Arm into direct chip production, a shift that Arm Chief Executive Officer Rene Haas said: “reflects the next phase of the Arm compute platform”.

Multiple technology organisations have partnered with Arm to launch and deploy the AGI CPU across cloud, enterprise, and networking environments. 

While Meta serves as the lead co-developer and primary customer for the processor, the broader coalition of official launch partners integrating the silicon into their operations includes Cerebras, Cloudflare, F5, OpenAI, Positron, Rebellions, SAP, and SK Telecom.

“OpenAI runs AI systems at massive scale. Hundreds of millions use ChatGPT every day, businesses build on our API, and developers rely on tools like Codex,” said Sachin Katti, Head of Industrial Compute at OpenAI, in the announcement. “The Arm AGI CPU will play an important role in our infrastructure as we scale, strengthening the orchestration layer that coordinates large scale AI workloads and improving efficiency, performance, and bandwidth across the system.”

Commercial systems featuring the Arm AGI CPU are currently available for order from manufacturers including Lenovo, ASRockRack, and Supermicro. 

Later this year, Meta and Arm plan to release their server board and rack designs to the Open Compute Project, an initiative that shares data centre product specifications across the technology industry to establish open hardware standards.