Better Artificial Intelligence Stock: Arm Holdings vs. ASML

Arm generated fiscal 2026 revenue of $4.9B, up 22.8% year over year, with $904M of net income and an 18.4% margin; its latest fiscal Q1 revenue rose 22% to $1.3B. ASML reported 2025 revenue of €32.7B versus €28.3B a year earlier and €9.6B of net income, for a 29.4% margin, while generating €11.1B in free cash flow. The comparison favors ASML for valuation and dividends, at 27.9x forward P/E versus Arm's 125.0x, while Arm offers faster growth but faces smartphone concentration, RISC-V competition, and elevated valuation risk; ASML remains exposed to China export restrictions and semiconductor-cycle order volatility.

Analysis

The investable distinction is duration of cash flows versus valuation fragility. ARM needs sustained royalty-rate expansion, data-center share gains, and new end-market penetration to justify a multiple that embeds several years of near-flawless execution; any evidence that AI inference remains concentrated in custom silicon, x86, or RISC-V architectures would compress the multiple before reported royalties weaken. ASML’s economics are more exposed to wafer-fab capex timing, but its installed-base service stream and sole-source position create a clearer path from leading-edge capacity additions to earnings power.

The key second-order variable is not aggregate AI demand but the mix of AI silicon. NVDA and hyperscaler custom-ASIC demand are unambiguously positive for ASML only when it translates into incremental leading-edge wafer starts and new-node transitions; inventory rebuilding at mature nodes would not carry the same implication. For ARM, higher chip unit volumes are less important than royalty mix: premium compute subsystems and server designs must become a material share of licensing and royalty revenue to prevent smartphone-linked demand from anchoring growth.

Over the next 1-3 months, ASML is likely to trade on order-book composition, China shipment visibility, and customer acceptance of next-generation systems rather than trailing revenue. ARM is more vulnerable around results because elevated expectations leave little tolerance for deferred licensing deals, weaker smartphone units, or stock-based-compensation dilution. Over 6-18 months, export-control tightening is ASML’s principal left-tail risk, while RISC-V adoption by Chinese vendors and internal hyperscaler silicon teams is ARM’s more gradual but structurally important risk.

Consensus correctly recognizes ASML’s monopoly but may underprice the cyclicality of customer capex concentration; a single delayed leading-edge fab ramp can move quarterly bookings sharply. Conversely, consensus may be too willing to capitalize ARM’s optionality as if it were recurring royalty revenue. The more attractive expression is relative: own the bottleneck at a normalized valuation while funding it with the architecture company priced for a much broader AI monetization outcome.

Key Decisions for Investors

• Initiate a 6-12 month pair trade: long ASML / short ARM, dollar-neutral. Target 15-25% relative outperformance as ARM’s valuation premium normalizes or ASML converts backlog into earnings; reassess if ARM shows two consecutive quarters of material royalty-rate acceleration driven by server and premium AI compute.
• Add ASML on any post-earnings decline tied to quarterly order volatility rather than cancellations, provided management maintains leading-edge customer demand and service-revenue outlook. Thesis is invalidated by broad order deferrals from TSMC, Samsung, and Intel or a meaningful expansion of export restrictions that reduces China revenue without offsetting non-China demand.
• Avoid outright ARM long exposure ahead of the next earnings release unless disclosed design-win and royalty-mix data demonstrate that non-smartphone revenue is scaling faster than the core handset base. A more tactical alternative is limited-risk ARM put spreads dated 3-6 months out if implied volatility is not already elevated; the catalyst is guidance de-risking rather than a deterioration in current reported revenue.
• Use NVDA and TSM as confirmation indicators, not direct substitutes: sustained AI accelerator demand plus announced leading-edge capacity additions supports ASML, while ARM requires evidence of Arm-based server CPU or custom-ASIC adoption. Set an alert for downward revisions to global smartphone units or increased RISC-V deployment at major Chinese OEMs, which would weaken ARM’s royalty-duration thesis.