Arm released its first in-house processor — Meta became its first customer

Arm has released its first proprietary processor and immediately named its first customer — Meta. The new chip is called Arm AGI CPU and is designed for inference tasks: performing the computations that power cloud AI services, including agent systems capable of generating increasingly complex chains of parallel tasks. This is a fundamental turning point for Arm.

The British company was founded in 1990 and over three decades has become one of the most influential yet least visible players in the semiconductor world. Throughout this time, Arm never manufactured chips independently — the business was built exclusively on licensing architecture. Companies like Apple, Qualcomm, Samsung, Nvidia, and Amazon paid for the right to design their own processors based on its architecture.

The Arm architecture powers most smartphones and tablets on the planet, as well as a huge portion of cloud infrastructure. But the company never had its own server hardware. Now that is changing.

Arm AGI CPU is oriented precisely at data centers and AI inference tasks — that is, not model training, but their industrial deployment in real services. The chip is specifically designed for agent workloads, where one AI agent can spawn dozens of parallel subtasks, requiring high computational parallelism from the infrastructure. This is where the main demand is concentrated right now: companies are building gigantic clusters to process billions of user queries to language models and multimodal systems.

Inference load volume is growing exponentially as AI applications reach the mass market. In this partnership, Meta acts not just as the first customer, but as a full co-developer. The company stated that it is the leading partner of the project and expects to work with Arm on several generations of data center CPUs.

At the same time, the new processors will take their place in Meta's heterogeneous infrastructure alongside GPUs from Nvidia and AMD, not displacing conventional hardware, but complementing it. The collaboration with Arm is explained not only by interest in the new chip, but also by Meta's accumulated difficulties with its own developments. The company has long been working on its proprietary AI accelerator MTIA, but timelines have been postponed several times, and the chip still hasn't captured a significant share of the internal infrastructure.

Partnership with Arm is a pragmatic solution: instead of building full chip expertise from scratch, Meta gains access to a proven architectural school with decades of engineering experience. For Arm, the transition to its own manufacturing is also a strategic move, and one that has been overdue. The company is consistently moving toward an IPO and is interested in justifying a higher market valuation.

Own products fundamentally change the revenue structure: instead of royalties from licenses — direct equipment sales with higher margins. Analysts have long pointed out that the licensing model limits revenue growth in absolute terms. If AGI CPU proves in demand, Arm will transform from an invisible architectural backend into a full-fledged server hardware manufacturer.

The name itself deserves special attention. AGI stands for Artificial General Intelligence, and despite the fact that modern systems don't yet possess real general intelligence, the abbreviation has long been adopted as a marker of the most advanced developments. Arm is clearly betting that the name will cement the chip's association with infrastructure for next-generation AI systems.

The first shipments of Arm AGI CPU are scheduled for 2026. If the collaboration with Meta proves successful, future customers could well include Amazon, Microsoft, and Google — all three are already building their own data center CPUs based on Arm architecture and have long been familiar with its capabilities.