Arm Holdings reported record quarterly revenue, record earnings and roughly doubled demand for its new data centre processor in the space of six weeks, and investors responded by wiping more than 11% off the share price.
The reaction to Tuesday's fiscal fourth-quarter results tells a story not about a company that missed expectations but about one that may have outrun them.
Arm, the Cambridge-headquartered chip designer whose architecture underpins everything from smartphones to the servers running artificial intelligence workloads, posted revenue of $1.49 billion for the three months to 31 March, up 20% year on year, with licensing income rising 29% and non-GAAP earnings per share hitting $0.60, ahead of consensus estimates.
Data centre royalties more than doubled for the second consecutive quarter, and the company now claims roughly 50% market share among the top hyperscale cloud providers.
None of that was the problem.
The sell-off centred on the AGI CPU, Arm's first production silicon product, a 136-core processor designed for the agentic AI workloads that are expected to require more than four times today's CPU capacity by 2030.
Customer demand for the chip has surged to approximately $2 billion across fiscal 2027 and 2028, more than double the figure Arm cited at its launch in March.
Yet chief executive Rene Haas acknowledged on the earnings call that Arm had secured supply for only the first $1 billion of that demand, with constraints across wafers, substrates, memory and packaging limiting its ability to capitalise on the rest.
Management left its near-term royalty and chip revenue outlook unchanged, a decision that prompted Raymond James analyst Simon Leopold to note that supply constraints had prevented the company from raising its forecast.
For a stock trading at roughly 69 times calendar-year 2028 earnings and already up 84% in 2026, that was enough to trigger a sharp reversal.
The deeper question is whether the AGI CPU represents a natural extension of Arm's business model or a fundamental transformation of it, and whether the market has already priced in the answer.
Arm has spent four decades as the architect's architect, licensing chip designs and collecting royalties every time a semiconductor manufacturer ships a product based on its blueprints.
The model is capital-light, high-margin and almost uniquely scalable: Arm-based chips shipped more than 30 billion units in fiscal 2026 alone.
The AGI CPU changes that equation by putting Arm into the business of selling finished silicon, a move that carries manufacturing risk, supply chain exposure and the potential to compete directly with the very customers who pay Arm licensing fees.
Amazon Web Services builds its own Graviton processors on Arm architecture, Google has its Axion CPU, and Microsoft deploys its Cobalt chip, all derived from Arm's intellectual property.
These companies are now simultaneously Arm's licensees and its competitors in the data centre CPU market, a tension that Bank of America analyst Vivek Arya pressed Haas on during the earnings call.
Haas was characteristically bullish, arguing that efficiency per core gives Arm a structural advantage in high core-count designs and predicting that Arm-based CPUs would hold the largest market share of any architecture by the end of the decade.
He pointed out that every major AI accelerator, from Nvidia's Blackwell and Rubin platforms to Google's TPU and Amazon's Trainium, now pairs with an Arm CPU.
BofA was less convinced, modelling a 5% to 7% share of the AGI CPU market for Arm by fiscal 2031, well below management's 15% target.
The bank cited intensifying competition from x86 incumbents AMD and Intel, the proliferation of custom Arm-based designs by hyperscalers themselves, and multi-sourcing agreements that could constrain Arm's pricing power.
BofA nudged its estimates slightly higher and raised its price target to $245 from $180 but retained a neutral rating, noting that the stock is already pricing in a successful transformation to a mixed IP-and-chip company.
The smartphone market added a further wrinkle: Haas disclosed that handset unit growth turned negative last quarter, concentrated at the lower end but sufficient to weigh on royalty income in what remains Arm's largest revenue source by volume.
Counterpoint Research projects that Arm-based CPUs could account for 90% of host CPU deployments in custom AI server configurations by 2029, up from roughly 25% in 2025, a trajectory that supports the bull case for the architecture even if it does not resolve the question of who captures the economics.
If hyperscalers continue to design their own Arm-based chips, Arm collects a royalty; if they buy the AGI CPU off the shelf, Arm captures a far larger share of the value.
Related reading
- Google in talks with Marvell to develop two new AI chips as it builds multi-supplier TPU strategy
- Nvidia launches open-source quantum AI models as it deepens grip on chip design workflows
- Meta partners with Broadcom on MTIA AI chips
The risk for investors is that Arm ends up in a position where the architecture wins comprehensively, but the company's share of the spoils proves more modest than the valuation implies.
At 69 times 2028 earnings, the stock needs not just a growing data centre market but concrete evidence that Arm can convert demand into supplied revenue and defend its pricing against customers with the resources and incentive to
The recap
- Arm's shares fell over 11% after Q4 and June-quarter results.
- Data center revenue grew more than 100% year-over-year.
- Management reiterated 15% AGI CPU market goal by fiscal 2031.
