Arm Holdings Enters Chip Manufacturing After 36 Years of Licensing, Unveils Arm AGI CPU for AI Data Centers

Storied semiconductor and software company Arm Holdings is embarking on a monumental shift in its business model, moving beyond its nearly 36-year tradition of exclusively licensing its designs to companies like Nvidia and Apple, and beginning to manufacture its own chips. This strategic pivot was highlighted by the unveiling of its first production-ready chip, the Arm AGI CPU, designed specifically for running inference in an AI data center. The announcement was made at a significant industry event in San Francisco, CA, which ran from October 13-15, 2026, with the specific revelation occurring on a Tuesday during the conference.

This move marks a historic deviation for the UK-based company, which has long been the architectural backbone of billions of devices worldwide, primarily through its intellectual property (IP) licensing model. For decades, Arm’s core business involved developing CPU architectures and core designs, then licensing them to other semiconductor companies. These licensees would then integrate Arm’s IP into their own system-on-chips (SoCs), tailoring them for various applications ranging from smartphones and tablets to embedded systems, automotive components, and enterprise servers. This approach allowed Arm to maintain a pervasive presence across the technology landscape without the capital-intensive burden of fabrication. Its designs are renowned for their power efficiency and performance, which have made them indispensable, particularly in the mobile computing sector, and increasingly relevant in the data center and artificial intelligence arenas.

The newly introduced Arm AGI CPU represents a direct foray into silicon production, signaling Arm’s intent to become a more direct player in the hardware market. The chip has been developed leveraging Arm’s advanced Neoverse family of CPU IP cores, a line specifically engineered for cloud, data center, and infrastructure applications. The Neoverse platform provides a scalable and efficient foundation for high-performance computing, making it a natural choice for an AI data center-focused processor. Crucially, the development of the Arm AGI CPU was undertaken through a strategic partnership with Meta, one of the world’s leading technology companies with vast AI infrastructure requirements.

Meta is not only a development partner but also the inaugural customer for the Arm AGI CPU. The chip is meticulously engineered to integrate seamlessly and work harmoniously with Meta’s existing training and inference accelerators. This deep collaboration suggests a highly optimized solution tailored to Meta’s specific workload demands, aiming to enhance the efficiency and performance of their expansive AI operations. The involvement of Meta as both a partner and first customer lends significant credibility and market validation to Arm’s new offering. Beyond Meta, Arm has also announced a robust roster of launch partners, including prominent AI and cloud infrastructure firms such as OpenAI, Cerebras, and Cloudflare. The inclusion of such influential players underscores the Arm AGI CPU’s potential to address critical needs within the rapidly evolving AI ecosystem. OpenAI’s involvement, in particular, highlights the chip’s relevance for cutting-edge AI model deployment, while Cerebras and Cloudflare signify its applicability across specialized AI hardware and broad cloud service provision.

The transition to manufacturing its own silicon has been a subject of industry anticipation for some time. Reports, including those from CNBC, indicate that Arm initiated the development of these proprietary chips as early as 2023. The rapid progression from conceptualization to a "production-ready" state, with processors already available for order, speaks to the urgency and strategic importance Arm places on this new venture. While TechCrunch has reached out to Arm for more granular details regarding the precise timeline of the chip’s development and release, the swift market entry is notable.

This strategic shift is a profound departure from Arm’s established identity and business model. For nearly four decades, Arm thrived by providing the foundational blueprints for others to build upon, meticulously avoiding direct competition with its diverse customer base. Now, by producing its own CPUs, Arm will inevitably find itself competing alongside many of its long-standing partners in certain market segments. This move suggests a re-evaluation of market opportunities and a desire to capture a larger share of the value chain, particularly in high-growth areas like AI data centers. Arm Holdings itself is majority owned by the Japanese conglomerate Softbank Group, whose strategic vision likely plays a significant role in empowering such a transformative corporate direction.

A particularly noteworthy aspect of this launch is Arm’s decision to produce a CPU (Central Processing Unit) rather than a GPU (Graphics Processing Unit). In the current technological landscape, GPUs have garnered immense attention and investment due to their unparalleled capabilities in training and executing complex AI models, owing to their highly parallel processing architectures. However, Arm emphasizes that CPUs remain an equally, if not more, critical component within a modern data center rack, particularly for orchestrating and managing the vast, distributed workloads that characterize contemporary AI infrastructure.

In its compelling "pro-CPU pitch," Arm meticulously outlines the indispensable role of these processors. CPUs are tasked with managing thousands of distributed tasks crucial for the seamless operation of large-scale AI systems. These responsibilities include managing memory and storage resources, intelligently scheduling diverse workloads across various computing units, and efficiently moving vast amounts of data across complex systems. Arm posits that the CPU has evolved into "the pacing element of modern infrastructure – responsible for keeping distributed AI systems operating efficiently at scale." This highlights the CPU’s role as the central conductor, ensuring that all components of an AI data center work in concert, preventing bottlenecks, and maximizing overall system throughput and responsiveness.

The company further argues that the escalating demands of distributed AI place unprecedented pressure on CPUs, necessitating a significant evolution in processor design and capabilities. This evolution likely encompasses advancements in core architecture, enhanced input/output (I/O) capabilities, specialized instructions for AI-related tasks, improved power efficiency, and greater scalability to handle ever-increasing computational loads.

The timing of Arm’s entry into the CPU market with its own silicon is also highly relevant given the current global supply chain dynamics. CPUs, once readily available, have become increasingly difficult to procure. In March, leading chip manufacturers Intel and AMD reportedly informed their customers in China of extended wait times for their CPU products, a situation initially reported by Reuters. This scarcity, driven by a confluence of factors including surging demand from cloud computing and AI, along with lingering supply chain disruptions, has begun to exert upward pressure on hardware costs. Reports indicate that computer prices have already started to rise amid this growing shortage. Arm’s direct entry into manufacturing could potentially offer an alternative supply source and introduce new competitive dynamics, which could prove beneficial for data center operators grappling with procurement challenges.

This strategic pivot by Arm represents a bold step that could redefine its role in the global semiconductor industry, moving it from a foundational IP provider to a direct competitor in the high-stakes AI hardware market.

Becca Szkutak is a senior writer at TechCrunch, specializing in venture capital trends and startups. Prior to joining TechCrunch, she covered the same beat for Forbes and the Venture Capital Journal. You can contact or verify outreach from Becca by emailing [email protected].