Arm Breaks From Licensing Roots With First In-House AI Chip

Arm has stepped into unfamiliar territory with a move that could reshape its role in the semiconductor industry. Long known for licensing chip designs rather than manufacturing hardware, the company has announced it will begin selling its own processors, marking a decisive shift in strategy.

The company unveiled its first in-house product, a data center processor called the Arm AGI CPU, designed to support agentic AI workloads. The launch represents what Arm CEO Rene Haas described as “a defining moment” for the company, as it extends its platform from intellectual property into physical silicon.

A defining moment, indeed. For more than three decades, Arm’s business model centered on supplying chip blueprints to partners, collecting licensing fees and royalties on billions of devices, from smartphones to servers. That approach helped the company become one of the most pervasive forces in computing, while avoiding direct competition with its customers.

Designed for AI Data Centers

Arm’s processor is aimed at large-scale AI data centers, where demand for compute is accelerating as companies deploy increasingly complex models. The chip includes up to 136 cores and is optimized for energy efficiency, a critical factor as data center power consumption climbs. Arm estimates that its design could significantly reduce the cost of building large AI facilities, where total investment can reach tens of billions of dollars.

Meta is the lead partner on the project and the first confirmed customer. The social media giant collaborated on development and plans to deploy the chip as part of its AI infrastructure. Other early adopters include OpenAI, Cloudflare, and SAP, signaling major interest across the tech sector.

The partnership underscores a growing urgency among tech firms to secure reliable and efficient compute resources. As AI systems evolve from model training to continuous operation, handling inference and ever more data processing, CPUs are regaining importance alongside GPUs. These general-purpose processors are becoming a constraint in AI systems, particularly as workloads diversify.

Arm executives argue that the shift into silicon is driven by customer demand rather than competitive ambition. As they describe it, the chip is intended to complement, rather than replace, existing offerings from partners.

It is likely, though, that Arm’s move will prompt some market friction. Arm’s customer base includes many of the same companies now developing their own processors, including cloud providers and chipmakers. By entering the market directly, Arm risks blurring the line between supplier and competitor.

Yet it is also likely that Arm will find an ongoing source of revenue from its new chip. Data centers are expanding rapidly, and the industry is seeking ways to increase performance without proportionally increasing power consumption. Arm’s architecture, known for its efficiency, has gained traction in this environment, particularly as hyperscalers look for alternatives to traditional x86 processors.

Co-Designed Hardware

The company’s expansion also reflects how AI infrastructure is built. Rather than relying solely on off-the-shelf components, large tech firms are increasingly co-designing hardware tailored to their workloads. Arm’s new role positions it as both a collaborator and a vendor in that process.

Production of the AGI CPU is expected to scale later this year, with manufacturing handled by TSMC using advanced process technology. Early systems are already being tested, and broader availability is expected in the second half of the year.

Whether Arm can navigate the competitive tensions created by this move remains an open question. What is clear is that the company is no longer content to remain behind the scenes. As AI reshapes the economics of computing, Arm is investing heavily in producing silicon.