Is Arm Holdings Still a Hidden Winner of the AI Chip Race?

Arm's revenue grew 23% and 24% in its last two fiscal years, exceeding its 17%-21% growth rates during the early smartphone era, as Arm-based CPUs gain data-center adoption through Nvidia, Amazon, Microsoft, and Alphabet. Higher-value Armv9 architectures and compute subsystems are increasing royalty revenue per chip alongside AI-driven unit-volume growth. Arm shares have outperformed Nvidia over the past month, year, and three years, though the stock trades at a rich 55x trailing sales and 188x free cash flow.

Analysis

ARM is increasingly a monetization lever on hyperscaler custom-silicon adoption, but the relevant earnings sensitivity is not simply AI-server unit growth. The upside requires a sustained mix shift toward higher-value architecture and subsystem content; large cloud customers also possess meaningful bargaining power and can internalize more design work once volumes scale. That makes ARM's reported royalty-rate progression and licensing backlog more important than aggregate data-center capex headlines over the next 1-3 quarters.

The competitive second-order effect favors AMZN, GOOG and MSFT operationally: proprietary Arm CPUs reduce dependence on merchant x86 supply and can lower inference and general-compute cost per workload. NVDA benefits where Arm-based host CPUs deepen ecosystem integration, but ARM CPU penetration can pressure AMD's data-center CPU share if hyperscaler-designed silicon expands faster than AMD's EPYC adoption. Intel is the more direct omitted exposure: custom Arm deployment reduces the addressable merchant CPU pool even if total server compute spending remains strong.

The stock's valuation leaves little tolerance for a normal licensing/royalty timing mismatch, customer concentration concern, or evidence that accelerator-heavy clusters require fewer host CPUs per dollar of AI capex than investors assume. Near-term upside is therefore more likely to be narrative-driven than fundamental unless ARM demonstrates accelerating royalty revenue and higher effective royalty rates in the next results cycle. Over 6-18 months, the structural thesis is credible, but the market may be pricing an unusually smooth transition from mobile-led to infrastructure-led economics.

Contrarian view: hyperscaler Arm CPU success is not automatically ARM margin expansion. The largest customers' motivation is cost control and design autonomy, which ultimately caps supplier economics; ARM's value capture depends on retaining architectural indispensability rather than merely participating in unit growth. A broad AI-capex slowdown would also hit ARM later than accelerator vendors through royalty reporting lag, creating a delayed de-rating risk.

Key Decisions for Investors

• Do not chase ARM on this article alone; maintain a watch position only until the next earnings release validates royalty-revenue acceleration, effective royalty-rate expansion, and data-center licensing conversion. A guidance raise driven primarily by license revenue rather than royalties would not validate the thesis.
• Express the near-term custom-silicon beneficiary view through a 1-3 month long AMZN / short AMD pair, sized modestly: Amazon captures cloud-margin benefits from Graviton while AMD faces incremental merchant-CPU substitution risk. Exit if AMD reports materially faster EPYC share gains or AMZN commentary indicates custom-chip deployment is not reducing infrastructure unit costs.
• For ARM exposure, prefer waiting for a post-earnings entry after a 15-20% valuation reset or a demonstrable royalty inflection; at current expectations, downside from a merely in-line quarter likely exceeds upside from modest beats. Use a 6-12 month horizon and cap risk with puts rather than unhedged common.
• Monitor NVDA host-CPU attach and hyperscaler capex disclosures over the next two quarters. Rising accelerator shipments without corresponding ARM royalty growth would falsify the assumed CPU-content linkage and is a signal to avoid or reduce ARM exposure.