The CPU Architecture Powering the AI Compute Buildout

Arm Holdings plc is a semiconductor intellectual property company headquartered in Cambridge, United Kingdom, and listed on Nasdaq. It develops and licenses CPU architectures, core designs, and related system IP — including graphics, interconnect, and security components — that other companies integrate into their own chips.

How it makes money: Two primary streams. Licensing revenue is collected upfront when a partner signs an agreement to use Arm technology, and royalty revenue is collected per chip shipped, calculated as a percentage of the chip's selling price. The royalty stream is the larger and more valuable component because it scales with partner success without requiring Arm to invest in manufacturing or inventory.

Customers: Arm's IP is embedded in the overwhelming majority of smartphones worldwide, and its customer list spans mobile (Apple, Samsung, Qualcomm, MediaTek), cloud and data center (Amazon, NVIDIA, Microsoft, Google), automotive, and embedded/IoT. This breadth is a strength, but it also means the largest customers carry outsized negotiating leverage.

Scale: With a $264.2B market cap, 1,068.00M shares outstanding, and $0.98 in trailing EPS, Arm is priced as a franchise asset rather than on near-term earnings power. The public float of 142.62M shares is a small fraction of shares outstanding, reflecting the company's recent spin-out from SoftBank and the resulting concentrated ownership structure.

Near-term (next 4-8 quarters):

• Royalty rate expansion from v9 adoption. As more shipped chips move to Armv9, the blended royalty rate per unit rises. This is the most visible and most predictable near-term lever.
• Smartphone unit stabilization. The mobile base is mature, but a replacement cycle tied to on-device AI features could lift unit volumes modestly above the flat-to-low-single-digit trend.
• Automotive and IoT design wins converting to royalties. Automotive design cycles are long, but wins booked several years ago begin contributing royalty revenue as those platforms enter production.

Medium-term (3-5 years):

• Data-center CPU share gains. Arm-based server CPUs are taking share from x86 in hyperscaler deployments, and each server socket carries a far higher royalty dollar content than a smartphone core.
• AI accelerator and custom silicon attach. As hyperscalers design custom AI silicon, Arm IP becomes a candidate for the control and host CPU portions of those systems, expanding content per deployment.
• Edge AI inference. On-device inference across phones, PCs, and embedded devices expands the number of Arm-based chips per user, compounding the royalty base.

The key uncertainty is not whether these drivers exist but whether the royalty economics on data-center and AI silicon hold at the rates the current valuation implies.

Projected figures are illustrative estimates consistent with the growth and margin trajectory described below; trailing EPS of $0.98 and current price of $244.71 are as reported.

The model is driven by three things: royalty rate expansion as v9 and data-center mix rises, operating leverage as R&D spend grows slower than revenue, and a gross margin that is already near its ceiling and therefore contributes little incremental upside. Revenue growth in the mid-20s with operating margin expanding from the mid-teens toward the high-30s produces EPS growth that outpaces revenue growth — that operating leverage is what justifies a premium multiple, though not necessarily the current one. The critical sensitivity is royalty rate: a 100bp compression in blended royalty rate would reduce FY2027E EPS by roughly 8-10%, which at a 177x forward multiple translates into a disproportionate price impact.

Market context: Arm's addressable market is effectively all computing — mobile, data center, automotive, and embedded. The relevant TAM is measured in hundreds of billions of dollars of annual chip revenue, of which Arm captures a low-single-digit percentage as royalty. The bull case is not that the TAM grows, but that Arm's capture rate within it rises as chips become more expensive and more Arm-based.

Competitive positioning:

• Architectural entrenchment. Arm's instruction set is the default for mobile and is deeply embedded in toolchains, operating systems, and developer ecosystems. Switching costs are high and measured in years.
• The RISC-V threat. RISC-V is an open, royalty-free alternative that is gaining traction in embedded and microcontroller applications and is being evaluated for higher-performance use cases. It is the most credible long-term threat to Arm's pricing power, though it currently lacks the ecosystem maturity to displace Arm in premium mobile or data-center designs.
• Customer vertical integration. Apple, Qualcomm, and hyperscalers all have the resources to develop proprietary architectures. Their incentive to do so scales directly with the royalty rate Arm charges.

Named comparables: NVIDIA (accelerated computing, AI data-center exposure), Qualcomm (mobile silicon and licensing, closest business-model analogue), Synopsys and Cadence (EDA and IP licensing, similar asset-light model), and Broadcom (custom silicon and infrastructure).

DCF discussion: A discounted cash flow analysis on Arm is unusually sensitive to terminal assumptions because the near-term free cash flow is small relative to the $264.2B market cap. Using a discount rate of roughly 9-11% (justified by a beta of 3.89 but moderated by the durability of the licensing franchise) and a terminal growth rate of 4-5%, the model only supports the current price if royalty revenue compounds in the low-to-mid 20s for at least a decade. Under a more conservative 15% long-term growth assumption, the DCF implies fair value closer to $180-200 per share — roughly 20-25% below the current price. The stock is therefore a bet on the durability and slope of the AI royalty ramp, not on near-term cash generation.

Comparable multiples:

Arm trades at a substantial premium to every comparable on trailing earnings. The premium is defensible only if the royalty model's operating leverage and the data-center opportunity are genuinely larger than the market assigns to Qualcomm's licensing business or to the EDA/IP peers. We think the premium is justified in direction but stretched in magnitude at $244.71.

• Valuation compression. At roughly 250x trailing EPS and 177x our FY2027E estimate, the stock has no margin of safety. A single quarter of decelerating royalty growth or a guidance miss could compress the multiple sharply, independent of the long-term thesis.
• Customer concentration and licensing renegotiation. A handful of customers account for a large share of royalty revenue. Any renegotiation that lowers the royalty rate, or a dispute that disrupts a major agreement, would hit revenue and the multiple simultaneously.
• RISC-V and architectural substitution. Open-source alternatives are improving and are being actively evaluated for higher-performance applications. Adoption is slow, but the direction of travel erodes Arm's long-term pricing power.
• AI capex cyclicality. Arm's data-center opportunity is tied to hyperscaler capital spending, which is cyclical and sentiment-driven. A pause in AI infrastructure investment would delay the royalty ramp that the valuation depends on.
• Float scarcity and volatility. With only 142.62M shares in public float and 11.52% of float shorted, the stock is prone to violent moves in both directions. The 7.59% decline on this session on 1.13M shares — well below the 6.27M average volume — illustrates how thin the marginal liquidity is relative to a $264.2B market cap.