Arm Holdings Sets Records as AI Reshapes Its Ambitions

With $4.92 billion in revenue, dominant hyperscale positions, and a bold move into silicon, Arm Holdings is converting architectural advantage into something far more valuable: indispensability.

Key Takeaways

• Arm reported full-year revenue of $4.92 billion, up 23%, marking three consecutive years of 20%-plus growth since its 2023 IPO, with data-centre royalties more than doubling year-over-year.
• The company’s first production chip in 35 years, the AGI CPU co-developed with Meta, has already attracted more than $2 billion in forward demand and is orderable through major server partners including Lenovo and Supermicro.
• Arm now holds roughly 50% of CPU compute among top hyperscalers, with Google, Microsoft, NVIDIA, and AWS all expanding their reliance on Arm-based designs for AI infrastructure.

Records Built on Architecture

For three consecutive fiscal years, Arm Holdings has grown revenue by more than 20%. That consistency, rare in semiconductors and almost unheard of across cycles, tells a story that the headline numbers alone do not fully capture. The company is not surfing a single product wave. It is becoming load-bearing infrastructure for the computing era now taking shape.

Fiscal year 2026, ended March 31, delivered revenue of $4.92 billion, up 23% on the prior year. The fourth quarter alone contributed $1.49 billion, a 20% year-on-year increase that came in above the midpoint of guidance. Licensing revenue, the forward-looking metric that reflects customer confidence in future deployments, rose 25% for the year to $2.31 billion. Royalty revenue, the recurring income generated as chips ship, reached $2.61 billion, up 21%. Both are records. Non-GAAP diluted earnings per share hit $1.77 for the year, also a record, on operating margins of roughly 49%.

The figure that most clearly marks the moment, however, is the doubling of data-centre royalties year-over-year. That single data point reflects the degree to which hyperscale AI infrastructure has begun to absorb Arm’s architecture at scale, and it foreshadows the trajectory ahead.

The Architecture That Won the Cloud

Arm’s original proposition was elegant in its simplicity: license a power-efficient instruction set architecture to chip designers, collect royalties on every unit shipped, and let the ecosystem compound. Over 35 years and 350 billion chips, the model worked extraordinarily well in mobile and embedded markets. What has changed, with remarkable speed, is where the architecture is now winning.

The hyperscalers, once largely committed to x86 for server workloads, have shifted. Google’s Axion, Microsoft’s Cobalt, AWS’s Graviton, and NVIDIA’s Vera are all Arm-based. The company now claims approximately 50% of CPU compute capacity among top hyperscalers, a figure that would have seemed improbable five years ago. The drivers are not philosophical: power efficiency and total cost of ownership have become the defining constraints of data-centre economics, and Arm’s architecture has structural advantages on both.

As hyperscalers race to build gigawatt-scale facilities, the arithmetic becomes compelling. Arm’s own projections suggest its designs can deliver more than twice the performance per rack compared with x86 alternatives, with potential capital expenditure savings of up to $10 billion per gigawatt of capacity. These are figures that command serious attention from infrastructure planners, regardless of incumbent relationships or legacy software estates.

Silicon: The Boldest Move in a Generation

The most consequential development accompanying the results was not financial. In late March 2026, at its “Arm Everywhere” event, the company announced the AGI CPU: its first production silicon product in 35 years. Co-developed with Meta as the lead partner, the chip is purpose-built for the agentic AI workloads that define the next phase of inference at scale.

The decision to enter the silicon market represents a deliberate expansion of Arm’s commercial model. Historically, the company sold architecture: blueprints that customers licensed and then built into their own designs. The AGI CPU extends that offering into finished hardware, meaning customers can now engage with Arm as an IP licensor, as a provider of integrated compute subsystems, or as a direct silicon supplier. The same architecture runs through all three tiers, preserving ecosystem cohesion while broadening the addressable opportunity.

Early demand has exceeded initial projections. Arm reported more than $2 billion in customer interest for the AGI CPU across fiscal years 2027 and 2028, more than double the figure cited at the product’s launch. Commercial systems are already orderable through ASRock Rack, Lenovo, Quanta, and Supermicro. The speed of that ramp reflects both the urgency of AI infrastructure buildout and the trust that enterprise customers place in Arm’s ecosystem.

Beyond Hyperscale

The narrative is predominantly one of cloud and AI, and rightly so. But Arm’s expansion is also broadening at the enterprise tier in ways that reinforce its long-term positioning. SAP has moved core workloads onto Arm-based infrastructure. Cloudflare has deployed across its global network. These wins matter not because of their individual scale, but because they signal that Arm’s architecture is proving capable across diverse, demanding workloads, not merely in environments purpose-built by hyperscalers with the resources to optimise aggressively.

The developer community, now 22 million strong, provides a compounding advantage that competitors cannot easily replicate. Software compatibility and developer familiarity reduce the friction of adoption, and as the ecosystem matures, the argument for moving to Arm becomes easier to make at every level of the stack.

Pricing the Execution

Arm’s shares rose approximately 13.6% on results day, closing at $237.30. The market’s enthusiasm was measured, however, by the premium already embedded in the valuation. At these levels, investors are not simply rewarding past performance; they are pricing continued high-teens to low-twenties percentage revenue growth and successful monetisation of the AGI CPU programme. Supply constraints on the new chip and the inherent lumpiness of licensing revenue are live concerns, and execution through the ramp will matter as much as the strategic logic behind it.

Management guided for continued 20%-range growth across both licensing and royalty lines in fiscal 2027. The pipeline supports that confidence, though the company operates in a competitive environment where x86 incumbents are not standing still, and custom silicon efforts by hyperscalers carry the potential to reduce third-party dependence over time.

The Inflection Point

Three record years. A first silicon product in a generation. Fifty percent of CPU compute at the world’s largest hyperscalers. These are not incremental milestones. They reflect a company that entered the AI era with structural advantages and has used them deliberately.

The question for investors and industry observers alike is whether Arm can convert architectural dominance into durable margin expansion as the silicon business scales. The conditions are unusually favourable: AI infrastructure spending shows no credible signs of slowing, power efficiency remains a paramount constraint, and Arm’s ecosystem advantages are difficult to displace quickly. The coming quarters will test whether the AGI CPU can shift from strong early demand to sustained, profitable volume. If it does, the record fiscal 2026 results may come to be seen not as a peak, but as a foundation.