Arm Holdings eyes full-chiplet CPUs, but analysts warn near-term headwinds remain

Bank of America expects Arm to unveil its first in-house standalone merchant CPU at the "Arm Everywhere" event in San Francisco, marking a strategic shift from IP-licensing and royalty models to a product-based offering. Analysts say the new chiplet could materially expand Arm's addressable market and earnings potential, allowing the company to compete across a roughly $60 billion agentic and AI CPU segment by 2030.

Analysis

A move by Arm into merchant CPUs changes the economics from low-capex, recurring-license cashflow toward higher-capex, product-driven revenue — that shifts valuation multiples and introduces binary execution risk. If Arm converts ~10–20% of current licensing relationships into product revenue over 3 years, incremental gross margin could expand materially but free cash flow becomes more lumpy and dependent on NRE and inventory cycles.

Competitive dynamics favor firms that already operate in a chiplet/packaging ecosystem: advanced foundries and OSATs will see higher ASPs per wafer and per-package realized value, while legacy x86 incumbents face pressure in low-latency, power-efficient segments. Hyperscalers could become both customers and gatekeepers — they may accelerate adoption if Arm’s product reduces total cost of ownership for inference, but they can also vertically integrate (incentivized to protect cloud margins), limiting TAM capture.

Second-order winners include EDA/IP tool vendors and CXL/interconnect specialists since an Arm-branded CPU accelerates demand for heterogeneous integration and coherence tooling; expect capex tails at TSMC/packagers to be larger and stickier if chiplet economics prove superior. Conversely, long-cycle suppliers to a single-vendor licensing model (pure-licensing service firms, some fabless partners) face revenue cannibalization and potential churn in the 12–36 month window.

Key risks are binary and timeline-driven: partner litigation/contract disputes, hyperscaler refusals to adopt, and software-ecosystem lock-in that slows server uptake. Near-term catalysts to watch are product silicon demos, foundry packaging reveals, and hyperscaler procurement trials over the next 3–12 months; a failure or slow ramp within 12–18 months materially compresses the bullish case.

Key Decisions for Investors

• Long ARM equity or LEAP calls (12–30 months): Buy ARM Jan-2028 LEAP calls (limit position to 2–3% NAV). Rationale: asymmetric upside if Arm captures even a small share of the agentic/AI CPU spend; downside limited to premium. Hedge with 25% notional short in a rival SoC vendor if execution risk rises.
• Pair trade — long advanced packaging/foundry vs short incumbent CPU vendor (6–18 months): Long TSM (+AMKR for packaging exposure) / Short INTC. Position size 1:1 notional; target 30–50% upside on the long if chiplet adoption accelerates, with the short protecting against broad market declines. Stop-loss: 20% adverse move on the pair.
• Long AMD (12 months) via call spread: Buy AMD 12-month call / sell higher strike call to finance (~1.5–2.0x leverage on upside). Rationale: AMD’s chiplet IP and ecosystem readiness should capture OEM wins if Arm’s move validates modular CPU economics. Risk: if Arm monopolizes OEM relationships, upside reduced; cap losses to net premium.
• Event-driven volatility play (0–3 months): Sell a portion of ARM post-event rally >15% using short-dated call spreads to monetize likely over-exuberance around demos. Reward: collect premium on probable buy-the-news fade; risk: product credibility surprises could gap higher — cap risk via defined-width spreads.