Arm Launches First In-House AI CPU to Compete With Partners
Arm Holdings is shifting its business model by manufacturing its own silicon, marking the first time the company has produced a chip in its 35-year history. Known for licensing architecture to giants like Apple and Nvidia, the U.K.-based firm unveiled the Arm AGI CPU during an event in San Francisco on Tuesday.
A Strategic Shift in Silicon
The new processor is engineered specifically for AI data center inference. Developed using Arm’s Neoverse CPU IP cores, the project began in 2023, according to CNBC. By entering the hardware manufacturing space, the SoftBank-owned company is now positioning itself as a direct competitor to many of its long-term licensing partners.
Meta has signed on as the debut customer for the Arm AGI CPU. The chip is designed to integrate seamlessly with Meta’s existing training and inference accelerators. Other industry players, including OpenAI, Cerebras, and Cloudflare, have also joined as launch partners for the new hardware.
Why CPUs Matter for AI Infrastructure
While the AI industry has been largely dominated by the demand for GPUs, Arm argues that CPUs remain the backbone of modern data center architecture. The company emphasized that its new CPU is built to manage the complex, distributed tasks required for AI at scale, such as:
- Handling memory and storage management
- Scheduling intensive workloads
- Efficiently moving data across large-scale systems
Arm described its new hardware as the “pacing element of modern infrastructure,” essential for keeping distributed AI systems running smoothly. This launch arrives at a critical moment for the hardware market, as industry-wide CPU shortages reported earlier this year have led to longer wait times and rising computer prices globally.
The Arm AGI CPU is currently available for order. Further details regarding the full development timeline remain pending as the company navigates this transition from pure-play design to hardware production.