Arm's Earnings Shine, but Valuation Narrative Needs a New Story

In April-June 2026, Arm delivered a performance report with revenue of $1.289 billion, net profit attributable to parent of $270 million, and a YoY growth rate of 107.69%. These numbers are impressive for any semiconductor IP company, but as a PE investor, what I see isn't just "growth," but the "quality of growth" and "sustainability of growth."

Short Term: Leverage Effect of Licensing Model Begins to Show

Arm's revenue structure consists of two parts: License fees and Royalties. The former is the fee customers pay once to buy architectural design licenses; the latter is the commission charged per chip shipped.

Behind the data for Q1 of Fiscal Year 2027, two points are worth breaking down:

• Royalty income growth far exceeds shipment volume growth: Global smartphone shipments grew only single digits YoY, but Arm's royalty income growth reached double digits. This implies the Average Selling Price (ASP) of royalties per chip is rising, driven by high-value chips like data center CPUs, automotive ADAS chips, and AI accelerators starting to contribute royalties. Penetration of Arm v9 architecture jumped from ~15% in 2025 to over 25% in 2026, with v9 royalty rates being 2-3 times that of v8.
• License fees maintain a high base: Of the $1.289 billion revenue, license fees account for ~40%, maintaining above $500 million for four consecutive quarters. This reflects a key signal: Even during the semiconductor downturn, design firms are locking in Arm architectures early for "products 2-3 years in the future."

[!note] Short-term judgment: Arm's financial performance aligns with the typical leverage model of "IP licensing + royalty" companies—license fees provide stable cash flow, while royalties bring excess profits when chip shipment volumes scale up. Current doubling of net profit mainly comes from improved gross margin on royalty income (slightly rising from 92% to 94%) and control of operating expense ratios.

Long Term: Architectural Moat and the "Collector" Role in the AI Era

Arm's long-term valuation logic depends on whether it can evolve from a "mobile CPU IP supplier" to the "foundation of all computing architectures." Currently, this logic faces two key challenges:

1. Competitive Barrier: Is the Threat from RISC-V Overestimated?

RISC-V's open-source architecture has attracted many Chinese manufacturers and startups in the past two years, but commercial implementation remains concentrated in low-compute scenarios like MCUs and IoT. In high-performance fields like server CPUs, mobile APs, and GPUs, Arm's software ecosystem (compilers, OS, toolchains) and hardware compatibility (backward compatibility across generations of instruction sets) constitute substantial barriers.

Key Data Comparison:

• Arm architecture software compatibility: 25 years of accumulation, over 15 million developers
• Number of certified RISC-V software packages: Less than 5% of Arm's
• Market share of Arm architecture in server CPUs reached 15% in Q1 2026 (vs 10% in the same period 2025), while RISC-V server chips are still in the lab stage

2. Exit Path: SoftBank's IPO Pricing and Arm's "AI Narrative"

When SoftBank pushed Arm to list in 2023, it gave a market cap of $35-40 billion, corresponding to ~20x PE for FY2024 net profit. Now, Arm's market cap approaches $80 billion, but the P/E ratio remains high at over 50x. This valuation is built on the assumption that "all AI inference chips will adopt Arm architecture."

But reality is:

• NVIDIA's GPUs use proprietary architectures, not relying on Arm
• Google TPU, Amazon Trainium, and other custom AI chips use Arm's Cortex-X series as control cores, but the compute part remains proprietary
• The most optimistic institutional forecasts suggest Arm architecture penetration in AI inference chips will only reach 40%-50% by 2030

Investment Judgment: Valuation Reasonable, But Needs Catalysts

At the current stage, Arm is not a high-cost-performance buy target.

Dimension	Analysis
Business Model	Extremely asset-light, high gross margin (>90%), stable cash flow, but growth depends on client shipment volume growth
Competitive Barrier	Software ecosystem unmatched, but RISC-V is eroding the low-end market with a "lower license fee + open source" strategy, which is a concern
Valuation Logic	Current 50x PE assumes net profit growth >25% for FY2027-2028. If AI compute demand slows or RISC-V breaks through in server CPUs, valuation faces restructuring
Exit Path	SoftBank still holds ~90%, limiting liquidity; institutional investors focus more on the next AI chip mass production cycle (2027-2028)

My suggestion: If Arm can prove the following two points, the current valuation is acceptable:

• Before the end of FY2027, royalty income growth consistently exceeds 20%, and v9 architecture penetration surpasses 35%
• Secure architectural licensing from at least one major AI chip manufacturer (like OpenAI, Meta) for their proprietary chips

Otherwise, the market may quickly discover that the $270 million net profit is merely a phased dividend under the "AI bubble," rather than the beginning of structural growth.

[!quote] Open Question: When AI inference chips shift from "customized ASICs" to "generalized CPU + accelerator hybrid architectures," does Arm's "collector" role become more solid, or will it be marginalized?

Original Link: https://www.ithome.com/0/983/440.htm