Arm's Historic Move: First In-House Chip in 35 Years

Arm Holdings, a venerable name in the semiconductor and software industry, is embarking on a significant new chapter by manufacturing its own chips. This strategic shift comes after nearly 36 years of primarily licensing its designs to prominent companies such as Nvidia and Apple.

At a recent event held in San Francisco on Tuesday, the company unveiled the Arm AGI CPU, a production-ready processor specifically engineered to handle inference tasks within AI data centers. The UK-based firm developed this chip by leveraging its sophisticated Arm Neoverse family of CPU IP cores, a project undertaken in close collaboration with Meta.

Meta has also been announced as the inaugural customer for the Arm AGI CPU, which is meticulously designed to integrate seamlessly with the tech giant's existing training and inference accelerators. Arm has further secured a formidable roster of launch partners, including OpenAI, Cerebras, and Cloudflare, among others.

Arm's transition into silicon production has been a subject of industry anticipation for some time. According to CNBC reporting, the company initiated the development of these chips in 2023, and the processors are now readily available for order.

TechCrunch has sought further information from Arm regarding the precise timeline for the chip's development and subsequent release.

While perhaps expected, this move represents a historic deviation from Arm's long-standing tradition of exclusively licensing its designs to other chip manufacturers. The company, which is predominantly owned by the Japanese conglomerate Softbank Group, will now find itself in direct competition with many of its established partners.

Notably, Arm's decision to produce a CPU, as opposed to a GPU, is significant. While Graphics Processing Units (GPUs) have commanded considerable attention for their critical role in training and running AI models, Central Processing Units (CPUs) remain an equally indispensable component within a data center rack.

In its advocacy for CPUs, Arm emphasizes their capability to manage thousands of distributed tasks, encompassing memory and storage management, workload scheduling, and the movement of data across complex systems. The company stated that the CPU has evolved into the "pacing element of modern infrastructure — responsible for keeping distributed AI systems operating efficiently at scale."

These evolving demands place new pressures on CPUs, necessitating a continuous evolution of the processor's capabilities, Arm asserted.

Furthermore, the availability of CPUs has become increasingly constrained.

In March, Intel and AMD reportedly informed their customers in China of extended wait times for their products, attributing these delays to ongoing CPU shortages, a development initially reported by Reuters. Concurrently, computer prices have begun to rise amidst this growing scarcity.