en.Wedoany.com Report, On March 24th, Arm announced the extension of its product portfolio from IP licensing to mass-produced chips, officially launching its first self-designed data center CPU—the Arm AGI CPU. This marks Arm's first complete chip product developed in-house since its founding, representing a critical step in the company's evolution from a foundational architecture licensor to a chip-level solution provider.
It is understood that this chip is specifically built for agentic AI infrastructure, integrating 136 Neoverse V3 cores per chip with a thermal design power of 300 watts, supporting both air-cooled and liquid-cooled deployments. At the system level, a single rack can accommodate over 45,000 cores. Arm states that its single-rack performance more than doubles that of x86 platforms, potentially saving up to $10 billion in capital expenditure per gigawatt of AI data center compute power.
Regarding customers and ecosystem partnerships, Meta is an early partner and co-developer of this chip. Meta plans to deploy the Arm AGI CPU in conjunction with its self-developed Meta Training and Inference Accelerator to build more efficient computing clusters. Additionally, Arm has confirmed further commercial collaborations with customers such as OpenAI, Cerebras, and Cloudflare. On the hardware ecosystem front, Arm is collaborating with ODM manufacturers including ASRock, Lenovo, and Quanta, with early systems already available and broader commercial deployment expected in the second half of this year.
The launch of this chip represents a strategic upgrade for Arm in the data center computing domain. For a long time, Arm architecture has dominated the mobile terminal and IoT sectors due to its high energy efficiency advantages. However, in the data center CPU market, the x86 architecture has long held a dominant position. In recent years, with the explosion in demand for cloud computing and AI computing power, Arm architecture has begun to gradually penetrate the server market, with Amazon's self-developed Graviton processor and NVIDIA's Grace CPU both based on Arm architecture. Arm's direct entry with its own data center CPU will further enrich the product choices within the Arm ecosystem for the data center field.
From a market positioning perspective, the Arm AGI CPU targets the emerging track of agentic AI infrastructure. Unlike traditional AI training chips, agentic AI emphasizes the model's ability to autonomously execute tasks, invoke tools, and perform multi-step reasoning, placing higher demands on the orchestration efficiency and energy efficiency ratio of computing resources. The high core density and high energy efficiency characteristics of the Arm AGI CPU align well with this demand.
With the launch of the Arm AGI CPU, the competitive landscape of the data center CPU market is becoming more diverse. Multiple technology paths, including x86, Arm, and self-developed chips, are now competing on the same stage, providing cloud computing providers and AI infrastructure builders with richer choices. Arm stated that it will continue to advance the commercial deployment of this chip and iteratively enhance product capabilities around the needs of agentic AI.
