Arm Enters Production Silicon With the Arm AGI CPU
Arm Holdings has moved into the production silicon business for the first time in its 35-year history with the launch of the Arm AGI CPU, a data center processor built for agentic AI workloads. The move marks a major change for the company, shifting it from a model centered on licensing intellectual property to one that now includes manufacturing its own processors.
Arm described the launch as the next phase of its compute platform and a defining moment for the company.
Arm AGI CPU Specifications and Data Center Design
CPU Core Count and Architecture
The Arm AGI CPU is built with up to 136 Arm Neoverse V3 cores per processor. It is designed to support AI-focused data center environments that need high core density and efficient compute performance.
Memory Bandwidth, Latency, and Power
Arm says the chip provides 6GB/s of memory bandwidth per core with sub-100-nanosecond latency. The processor has a 300-watt thermal design power.
Rack-Level Performance and Density
According to Arm, the chip delivers more than 2x performance per rack compared with x86 CPU platforms. The company also says it could enable up to $10 billion in capital expenditure savings per gigawatt of AI data center capacity.
In air-cooled deployments, the platform supports up to 8,160 cores per rack. In liquid-cooled systems, it can deliver more than 45,000 cores per rack.
Manufacturing Process
The Arm AGI CPU will be manufactured by TSMC using its 3nm process technology.
Arm's AI Chip Development Effort
CNBC reported that Arm built a $71 million laboratory in Austin, Texas, to support development of the new CPU. The facility reflects the scale of the company's investment as it expands beyond chip design and licensing into silicon products for AI data centers.
Meta's Role in the Arm AGI CPU Launch
Meta is the lead partner and co-developer of the chip. The company worked with Arm to build what it described as an efficient compute platform that significantly improves data center performance density.
Meta plans to pair the Arm AGI CPU with its custom MTIA accelerators. Arm and Meta have also committed to working together across multiple generations of the chip.
Commercial Commitments and Ecosystem Support
Companies Committed to the Platform
Arm said it has commercial commitments from:
- Cerebras
- Cloudflare
- F5
- OpenAI
- Positron
- Rebellions
- SAP
- SK Telecom
OEM and ODM Production Partners
Production support includes OEMs and ODMs such as:
- Lenovo
- Supermicro
- Quanta Computer
Broader Ecosystem Backing
Arm also said that more than 50 ecosystem players have expressed support. Those named include AWS, Broadcom, Google, Marvell, Micron, Microsoft, Nvidia, and Samsung.
How the Launch Changes Arm's Business Model
For more than three decades, Arm generated royalties by licensing chip designs to companies such as Apple and Nvidia instead of producing silicon itself. With the Arm AGI CPU, the company is now stepping into direct competition with some of its own licensees.
Reports about the project first emerged in February 2025, when the Financial Times reported that Arm had secured Meta as its first customer for a planned in-house chip.
SoftBank, which holds a majority stake in Arm, has positioned the company at the center of its wider AI infrastructure ambitions.
Why the Arm AGI CPU Matters for AI Data Centers
The launch brings together several major shifts at once: Arm's move into manufacturing, a processor aimed specifically at agentic AI workloads, and a development partnership with Meta. It also places the company more directly inside the fast-growing buildout of AI data center infrastructure, where rack density, memory bandwidth, latency, and power efficiency are central considerations.
FAQs
What is the Arm AGI CPU?
It is Arm's first in-house data center CPU, designed for agentic AI workloads.
Who is Arm's lead partner for the AGI CPU?
Meta is the lead partner and co-developer of the chip.
Where will the Arm AGI CPU be manufactured?
The chip will be manufactured by TSMC on its 3nm process technology.