Arm Q1 forecast beats estimates on AI demand

UK-based chip design firm Arm Holdings forecast first-quarter revenue of US$1.26 billion and adjusted earnings of 40 cents a share on May 6, above Wall Street estimates.

The company benefited from rising demand for its chip designs in AI data centers.

Fourth-quarter revenue reached US$1.49 billion, while royalty revenue of US$671 million missed estimates and licensing revenue of US$819 million beat them.

Arm shares rose 12% after the outlook but later fell 5.49% in extended trading after executives said they had supply for US$1 billion of demand for its new AGI CPU, but not for a second US$1 billion.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Arm moves from licensing chip designs to selling its own silicon

The AGI CPU is Arm’s first chip sold under its own design, shifting from a 35-year model centered on licensing chip technology to other companies 1.
That change gives Arm a new path to revenue through sales of its own central processing unit (CPU) chips, placing it closer to server chip sellers such as Intel and AMD in data centers 2.
Meta, the parent company of Facebook and Instagram, is the lead partner. It plans to spend as much as US$135 billion on capital expenditures this year. Other customers include OpenAI, the company behind ChatGPT 1. SAP, the German business software company, is also a customer 1.
Arm aims for US$15 billion in yearly silicon revenue by 2031. That target is above the roughly US$4.91 billion Wall Street expects for Arm in the current fiscal year 3.

Competition in server CPUs is changing data center costs

By entering chip sales, Arm may compete with some of its biggest licensing customers and partners, moving away from its long-standing neutral role 2.
Energy efficiency matters for buyers such as Meta, where electricity is a “very scarce resource” 1.
Arm says the AGI CPU delivers about twice the performance per watt of an x86 rack configuration, a standard server setup built on the chip architecture used by Intel and AMD. That could cut running costs for AI infrastructure 1.
For hyperscalers, large cloud and internet companies that run massive data centers, the chip adds another option beside Intel-based or AMD-based servers. The push also tracks a wider move toward CPU-heavy “agentic AI” workloads, where AI systems handle multi-step tasks with more autonomy 4.