Why Arm Fell 9.7% in September 2026 Despite Record First-Quarter Revenue

Arm shares dropped 9.7% to $239.01 in Monday's session even as the company had posted record first-quarter revenue in August, with investor attention shifting to a 148.9% year-on-year surge in capital expenditure to $545 million in FY2026 — a sharp departure from the asset-light model that long defined the platform.

Arm Holdings shares fell 9.7% to $239.01 in Monday's session, erasing weeks of goodwill earned from a record fiscal first quarter. The decline is a reminder that Arm's valuation — built on the premise of royalties compounding as AI chips proliferate across edge and data-center platforms — carries expectations that a single strong quarter cannot fully anchor. For a stock that has long traded at multiples more typical of high-margin software than traditional semiconductors, any signal of rising structural costs can produce outsized reactions.

The record Q1 result, announced via Arm's official newsroom on August 9, was driven by licensing fees from multiple semiconductor vendors incorporating AI functionality into their chip designs. That report reinforced the thesis that Arm's architecture occupies a structurally advantaged position in the AI buildout: the dominant IP layer for mobile compute, and increasingly adopted in hyperscaler custom silicon and purpose-built inference platforms licensed by Qualcomm, Apple, Amazon, and others. A September 10 Form 6-K filing with the SEC — the most recent regulatory disclosure before Monday's move — adds to the body of information the market is now digesting; the specific content of that filing warrants close reading to understand what, if anything, it contributed to the selloff.

What Slicast's own compilation of SEC XBRL filings makes concrete is the scale of investment now underpinning Arm's platform position. The company spent $545 million in capital expenditure in fiscal year 2026, up 148.9% from $219 million in FY2025, equivalent to 11% of its $4.92 billion in FY2026 revenue. Among fifteen chip-sector peers in Slicast's dataset, Arm ranks fifth by capex intensity, behind category leader Micron. The heaviest single period in the dataset was the quarter ended June 30, 2026, at $197 million. For a business historically structured around royalties and licensing — never owning fabs, never carrying significant physical plant — this trajectory represents a meaningful strategic inflection that the market is still pricing in.

Part of that investment footprint now extends to strategic equity stakes. In August, multiple reports confirmed that Arm had participated in the $312 million (€270.5 million) funding round for Olix, a European startup developing data-center AI inference accelerators. The rationale is legible: if Olix ships production silicon on Arm cores — the reasonable inference from a strategic investment of this kind — Arm gains royalty upside and a reference architecture for AI inference workloads that compete with GPU-centric alternatives. The risk is subtler. Arm's value as a neutral platform licensor trusted by competing chipmakers has historically depended on its equidistance from the fray. Equity commitments to specific chip ventures, even modest ones, raise legitimate questions about whether that neutrality can be maintained as the stakes in the AI race rise.

Those questions carry historical weight. In September 2020, Nvidia announced it would acquire Arm from SoftBank for up to $40 billion, a transaction that would have merged Arm's IP with Nvidia's GPU ecosystem in a single entity. Regulators in the United States, European Union, and United Kingdom blocked the deal by early 2022, and Arm subsequently listed independently on Nasdaq in September 2023. The episode clarified a structural reality: the breadth of Arm's licensee base — spanning rivals who would not tolerate a Nvidia-controlled Arm — is itself a competitive asset. That independence is worth preserving, and any strategy that incrementally erodes it deserves sustained scrutiny from investors and licensees alike.

The outlook is genuinely two-sided. The bull case rests on royalty compounding: every AI chip design win, from custom cloud silicon to edge inference modules, adds to the cumulative unit base from which Arm collects per-chip royalties at rates structurally higher than legacy mobile. The bear case is equally structural: if AI compute migrates toward architectures where Arm's IP is thinner — proprietary GPU clusters, x86-centric inference servers, or entirely new paradigms — the royalty mix could soften just as the capital bill grows. Three signals are worth tracking: the specific disclosures in the September 10 6-K and whether they foreshadow any revision to forward guidance; whether FY2027 capex intensity holds at or above 11% of revenue, which would distinguish a sustained investment cycle from a one-time upgrade; and the granularity of royalty revenue disclosure in upcoming quarters, which will confirm or challenge the AI licensing thesis that Arm's current premium presupposes.