ARM bets $15B on making its own chips, partnering with Meta and OpenAI
Mar 26, 2026
Key Points
- ARM is pivoting from pure licensing to manufacturing its own chips, targeting $15 billion in revenue by 2031 as CPU shortages intensify across AI infrastructure.
- Meta and OpenAI are co-developing ARM-based processors, signaling major cloud operators are willing to abandon x86 dominance for custom silicon alternatives.
- ARM's manufacturing bet compresses margins from 97% to roughly 50%, a structural shift the company must execute flawlessly given its 90x forward earnings valuation.
Summary
ARM is shifting from a pure licensing business to manufacturing its own chips. The stock rose 15% this week on the announcement. ARM plans to reach $15 billion in revenue by 2031 through custom chip development, up from $4 billion last year in licensing revenue.
A CPU shortage has become acute. Intel cannot produce CPUs fast enough to meet demand. Nvidia has started selling its Grace CPU alongside GPUs to help address the constraint. Agentic AI is the primary driver. These systems require constant CPU cycles to feed data to GPUs, make web queries, run Python code, and interact with external systems. Data centers are now seeing measurable degradation in uptime for non-GPU workloads as CPU-only infrastructure strains under the load.
ARM's shift changes its economics fundamentally. Its licensing business generates 97% gross margins, a near-pure software play. Manufacturing chips will compress margins to roughly 50%. The company expects the $15 billion addressable market to more than compensate for that compression. ARM trades at roughly 90 times forward earnings, the highest multiple among semiconductor peers, leaving little room for execution missteps.
Meta and OpenAI are the anchor partners. Meta is co-developing what ARM branded the "ARM AGI CPU," processors purpose-built for AI infrastructure and general compute. These partnerships show that major cloud operators are willing to bet on ARM-based alternatives to x86, which Intel and AMD still dominate.
Competitive dynamics cut both ways. Nvidia and ARM are technically competitors—Nvidia sells ARM-based Grace CPUs—but also allies in displacing x86 from data centers. Software built for Nvidia's ARM chips will likely run on ARM's own chips, creating a shared ecosystem advantage against Intel and AMD. This mirrors the broader GPU market, where multiple chipmakers now compete against CUDA's moat rather than against each other.
ARM's valuation assumes the company can execute a manufacturing operation at scale while maintaining pricing power in a market where Intel and AMD still set the bar. The company is betting that a CPU shortage and agentic AI demand will persist long enough to justify the margin compression and capital intensity of chip manufacturing.