On Tuesday, British semiconductor and software design firm Arm Holdings (NASDAQ: ARM) unveiled the Arm AGI CPU. The launch marks a milestone as the first in-house, production-ready semiconductor in Arm’s 36-year history.
Since its founding in 1990, the company has generated revenue by licensing its intellectual property to companies like Apple, Nvidia, and Qualcomm.
ARM Launches ARM AGI CPU, an AI Inference Chip for Data Centers Developed with Meta
The Arm AGI CPU, developed in partnership with Meta, is designed to run inference in AI data centers. The chip, based on the Arm Neoverse family of CPU IP cores and fabricated by Taiwan Semiconductor Manufacturing Co (TSMC) on a 3-nanometer (nm) technology, will work seamlessly with Meta’s training and inference accelerator.
Made from two distinct silicon pieces that operate as a single unit, it aims to address data-crunching needed for a specific type of AI model that, instead of responding to queries as a chatbot, can act on behalf of the user with minimal oversight.
Arm’s focus on CPUs rather than graphics processing units (GPUs), which are widely used to train and run AI models, is also notable. The company claims its chip is able to handle thousands of distributed tasks, such as managing memory and storage, scheduling workloads, and moving data across systems.
Meta is among the first customers for the chip, alongside OpenAI, Cloudflare, SAP, Cerebras, and SK Telecom. The launch comes amid growing popularity for agentic AI, which has led to increased demand for central processing units (CPUs), the majority of which are manufactured by industry titans like Arm, Intel, and AMD.
During the event in San Francisco, Arm said CPUs have become the “pacing element” of modern inference infrastructure and are responsible for keeping distributed AI systems operating efficiently at scale. This puts new demand on CPUs and requires an evolution of the processor, according to Arm.
Arm Expects New AI Chip to Generate $15 Billion in Revenue by 2030
For decades, the chipmaker relied only on its intellectual property for revenue by licensing its designs to companies and collecting a royalty payment based on the number of units sold. But last year, Arm signaled to investors it was focusing on making its own chips and had hired key executives to aid with the project.
Arm began work on the AGI CPU with Meta back in 2023. Mohamed Awad, who oversees the project and is the head of Arm’s cloud AI business, said the company will put the chip into production in the second half of 2026 and has already received test units from TSMC that function as expected.
He also noted that Arm is working on additional designs that it plans to release over the next 12-18 months. In addition to the chip itself, the company is partnering with Lenovo and Quanta Computer to offer complete systems for AI servers.
Speaking to the media, Arm CEO Rene Haas said the company is forecasting the AGI CPU to generate roughly $15 billion in annual revenue over the next five years. It is targeting $9 in earnings-per-share (EPS) and $25 billion in revenue.
Though ARM closed Tuesday’s session down 1.4%, the stock jumped 6.5% during after-market trading. This came after financial analysts reiterated the company’s projection for fiscal 2026. According to the London Stock Exchange Group’s (LSEG) estimates, Arm is expected to generate a net profit of $1.75 per share on revenue of $4.91 billion this year, primarily aided by the sales of AGI CPU.
Arm Holdings Inc. (ARM) is trading at $152.60 in pre-market – up 13.07%.
