Between 4 billion and 15 billion dollars, it is not the growth curve that separates them, but a self-disruption of the business model.

On March 24, Arm launched its first self-developed data center CPU in its 35-year history in San Francisco. This chip, named AGI CPU, is equipped with 136 Neoverse V3 cores, built on TSMC's 3nm process, with a 300W TDP. Meta is the first customer and will deploy it on a large scale within the year. Also announced as partners are OpenAI, Cerebras, Cloudflare, SAP, and SK Telecom.

Arm CEO Rene Haas presented a set of target numbers at the launch event, stating that the chip business aims to achieve 15 billion dollars in annual revenue by 2031, with total company revenue of 25 billion dollars and earnings per share of 9 dollars.

What do these numbers mean? Arm's total revenue for the entire company in FY2025 (ending March 2025) is 4.007 billion dollars, according to Arm's annual report data, with licensing revenue of 1.839 billion dollars and royalty income of 2.168 billion dollars, yielding a gross margin of 97%. In other words, a company with annual revenue of 4 billion dollars needs to rely on a new business to generate close to the scale of Intel's entire data center division within 5 years. According to Intel's Q4 2024 financial report, Intel's DCAI (Data Center and AI) division's total revenue for the entire year 2024 is 12.8 billion dollars.

From 4 billion to 15 billion, the 3.7 times leap represents that Arm is trying to transform from a pure IP licensing company into a hybrid entity that sells both designs and finished products. There is no precedent for this in the chip industry.

Why is Arm taking this risk? The answer lies in its client list.

Over the past three years, Arm's largest data center customers have been doing the same thing. According to publicly available AWS data, Amazon has migrated over 50% of its EC2 computing power to its self-developed Graviton chips, with the latest Graviton5 reaching 192 cores. According to Google Cloud, Google's Axion chips have supported the migration of over 30,000 internal applications, improving energy efficiency by 80%. Microsoft's Cobalt 200 is also based on the Arm Neoverse architecture, built with TSMC’s 3nm process and featuring 132 cores.

These cloud vendors all use Arm's architecture licensing, but design their chips, manufacture them, and deploy them themselves. What Arm earns is licensing fees and royalties, not profits from the chips. As more and more computing power demand is met by these self-developed chips, Arm's revenue ceiling in data centers becomes increasingly clear.

Looking at Arm's revenue structure over the past four years, the outline of this ceiling becomes more specific. According to Arm's historical financial reports, from FY2022 to FY2025, the company's total revenue grew from 2.7 billion dollars to 4 billion dollars, with an average annual growth of about 14%. During this period, royalty income grew from 1.562 billion dollars to 2.168 billion dollars, and licensing revenue increased from 1.141 billion dollars to 1.839 billion dollars. The growth rate of royalties has slowed down to 20% from previous years, and a significant portion of this 20% growth comes from upgrades to the mobile Armv9 architecture, not data centers.

Extrapolating at this growth rate, even if licensing and royalty income maintain an annual growth of around 20%, by 2031 it can only reach about 10 billion dollars. The remaining 15 billion dollars must come from a business that does not currently exist. This is the arithmetic logic behind Arm's decision to manufacture its own chips.

By choosing to manufacture chips, Arm is essentially competing with its own customers. A company that sells architectural blueprints has started building its own buildings, while its blueprint buyers have been building for several years.

This is the true background of the 136-core AGI CPU. According to The Register, this chip has a base frequency of 3.2 GHz, a maximum of 3.7 GHz, 12-channel DDR5 memory, 6 GB/s bandwidth per core, 96 PCIe 6.0 lanes, and supports CXL 3.0. Arm positions it as the "computing foundation of the agentic AI cloud era," focusing on CPU task scheduling and data flow management in AI inference rather than directly competing with GPUs.

The pace of market share changes is also telling. According to Omdia estimates, by 2025, Arm architecture servers are expected to account for around 21% of global shipments, with a growth rate of 70%. However, within hyperscale data centers, this ratio is already close to 50%. The 40-year monopoly of x86 is not collapsing but is being replaced chip by chip.

The risk of Arm's self-developed chips does not lie in technology but in relationships. Meta is willing to be the first customer, in part because Meta does not have as mature a self-developed chip project as Amazon or Google. But how will Amazon, Google, and Microsoft view this situation? If a supplier starts competing with you, will you still grant them the most critical architecture licensing?

Arm's bet is that the overall growth of the data center cake is faster than the decline in customer relationships. Rene Haas clearly believes that the demand for CPUs in the AI era is large enough that self-developed chips and architecture licensing can coexist. The goal of 15 billion dollars is the pricing of this judgment.

After 35 years of selling blueprints, it is the first time to build a building itself. The blueprints are still being sold, and the buildings are being constructed; it just depends on whether there is enough room on the same plot of land.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。