TL;DR · 30-second read
The Short Version
Arm is a British company that designs the blueprints for computer chips. It collects a small fee on every chip built from those blueprints.
Its fees from chips in big data centers, the warehouse-sized buildings that run websites and artificial intelligence, doubled in a year. Arm says its designs now power more than half the servers at the biggest cloud companies. In 2024 that figure was about one in seven.
The catch is that Arm’s total per-chip fees grew only 8%. Arm is winning the data center faster than the money is arriving.
Arm Holdings reported record fiscal second-quarter revenue of $1.44 billion on November 4, up 27% year over year, for the three months ended September 30, 2026. Its shareholder letter, furnished to the SEC on a Form 6-K, shows license and other revenue up 50% to $773 million and royalty revenue up 8% to $670 million. Within royalties, data center royalties doubled year over year. Arm said its share of servers at the top hyperscalers, the largest cloud operators, rose from 15% in calendar 2024 to more than 50% in calendar 2025. It expects that share to exceed 60% by the end of calendar 2026.
GuruFocus reported that Arm shares rose 4.01% to $173.41 on the morning of November 6, and that GAAP operating income for the quarter was $201 million, against non-GAAP operating income of $638 million. Arm also said its confidence in reaching up to $2 billion of revenue from its own Arm AGI CPU across fiscal 2027 and 2028 has increased.
Executive Summary
The headline numbers tell a clear story about where Arm’s growth is coming from. Data center royalties doubled. Licensing, the upfront fees chip designers pay to use Arm’s technology, grew 50%. According to Arm’s shareholder letter, Google Axion, AWS Graviton, Microsoft Cobalt and NVIDIA’s Arm-based platforms have pushed Arm’s share of top-hyperscaler servers past half. That share stood at 15% in 2024.
Read together, the figures also expose a lag. Total royalty revenue, the per-chip fees that make up most of Arm’s long-term economics, grew only 8%. A tripling of server share has not yet produced anything like a proportional rise in royalty dollars. Arm itself frames licensing growth as a leading indicator, and says it expects royalties to accelerate once the new designs reach production.
That gap matters to anyone pricing compute, planning capacity, or valuing the companies that supply data center processors. Server market share and server revenue capture are different things, and Arm’s own figures show the difference.
Share Tripled, Data Center Royalties Doubled, Total Royalties Rose 8%
Arm’s shareholder letter gives three growth rates that are worth setting side by side. Arm-based server share at the top hyperscalers more than tripled, from 15% in calendar 2024 to more than 50% in calendar 2025. Data center royalties doubled year over year in the July–September quarter. Total royalty revenue grew 8%, to $670 million. When one segment doubles and the total barely moves, the segment is still small relative to the whole, or other segments are shrinking.
A simple calculation shows the scale. Royalties of $670 million after 8% growth imply roughly $620 million a year earlier, an increase of about $50 million. Suppose Arm’s other royalty streams, which are historically dominated by smartphones, were flat. Then a doubling would mean data center royalties went from roughly $50 million to roughly $100 million a quarter, about 15% of the total. Arm does not break the figure out. If its other royalties grew, the data center share is smaller than that. If they fell, it is larger. Under any plausible split, the segment where Arm holds a majority of hyperscaler servers still produces a minority of its royalty income.
The mechanism explains the gap. Arm earns a royalty on each chip as it ships, typically priced as a fraction of the chip’s value. Server share is a deployment metric, and the 50% figure is for calendar 2025, while royalties are booked as chips leave the factory. Hyperscalers design these chips in-house to lower their own costs, so Arm’s slice is a small cut of the chip’s value, not the price of a merchant processor. Arm points to the 50% jump in licensing as the forward signal, saying royalty revenue should accelerate as these designs enter production. That is a reasonable expectation. On these results, it remains a forecast.
What Hyperscaler Adoption Means for the x86 Incumbents
The share gains Arm describes come from the cloud giants building their own processors. Arm’s letter says Amazon reported that Graviton revenue commitments rose more than 20 times year over year. Microsoft highlighted strong adoption of Cobalt, and Google called Axion a critical component of its AI infrastructure. NVIDIA’s Arm-based Vera CPU is generally available on Oracle Cloud Infrastructure and deployed in production with OpenAI. The servers Arm is not counted in at these operators are largely x86 machines built on Intel and AMD processors.
The economics are asymmetric. When a hyperscaler moves a workload from an x86 chip it buys to an Arm-based chip it designs, the x86 vendor loses a whole processor sale. Arm gains a royalty worth a fraction of that sale. That is one reason Arm’s royalty line moves more slowly than its share statistics. The value the incumbents lose does not transfer one-for-one to Arm. Much of it stays with the hyperscaler as lower cost, and some goes to foundries and other suppliers.
For data center operators and enterprise buyers, the practical effect is more processor choice inside the clouds, plus pressure on x86 suppliers to compete on performance per watt and price. It also brings a software consideration. Arm notes that its IP and silicon businesses share a common software ecosystem, but moving existing applications from x86 to Arm still has to be tested workload by workload.
Selling Silicon to the Companies That License It
Arm’s second growth engine, the Arm AGI CPU, changes its role from designer to chip vendor. Arm says purchase orders from multiple customers extend through calendar 2027. First commercial deliveries are on track for the fiscal fourth quarter of 2027, which runs January to March 2027. Oracle Cloud Infrastructure plans AGI CPU instances in 2027. Arm attributes its higher confidence in the up-to-$2 billion target to additional supply and customer mix.
This puts Arm in a position that needs careful handling. The hyperscalers whose in-house chips drive its data center royalties are also, in principle, buyers who could choose Arm’s finished silicon or keep designing their own. Arm presents the two as complementary, built on one platform and offered side by side on OCI. Whether licensees see it that way is the open question. A licensee that feels it is competing with its supplier has options, including other instruction sets, and that is the risk to watch as AGI CPU volumes ramp.
The Power-Efficiency Claim Deserves a Close Read
Arm ties its pitch to the industry’s tightest constraint, electricity. The letter cites a Morgan Stanley projection of a 38-gigawatt global power shortfall through 2028. It cites independent testing by Signal65 at the Open Compute Project summit, showing the AGI CPU delivering roughly twice the performance per rack of AMD’s EPYC 9755. And it states that this could reduce the projected shortfall by half.
The first two claims are specific and checkable. The third is an extrapolation. Halving a global shortfall would require the 2x per-rack result to hold across the general-purpose workloads data centers actually run, and it would require CPUs to account for most of the shortfall. In AI facilities, accelerators, networking and cooling take a large share of the power budget. A per-rack benchmark against one competing part, with workloads and configuration not detailed in the letter, is a useful signal for buyers to test. It does not settle a question about the global grid. Arm’s Neoverse CSS N4, rated at up to 2x performance per socket and 25% better performance per watt than its predecessor, points in the same direction for partner-designed chips. These are vendor-stated figures.
Background
Arm Holdings, based in Cambridge, England, and listed on Nasdaq, does not traditionally make chips. It designs processor architectures and cores and licenses them to companies such as Apple, Qualcomm, NVIDIA, Amazon, Google and Microsoft. These companies pay upfront licensing fees and then royalties on each chip shipped. Arm’s power-efficient designs have dominated smartphones for years. The data center server market, by contrast, was long dominated by x86 processors from Intel and AMD.
That has changed as the largest cloud operators began designing their own Arm-based server processors to control costs and power use. Arm has also moved beyond licensing, introducing compute subsystems that deliver more complete designs and, from March 2026, the Arm AGI CPU, its own silicon sold directly to customers. Arm’s fiscal year ends March 31, so its second fiscal quarter of 2027 covers July through September 2026. Source: Arm Holdings Stocks Rise 4.01% as Data Center Royalties Double (GuruFocus): Arm’s share move after its record second quarter, with data center royalties doubling. Primary sources: Arm Holdings Form 6-K, quarterly financial report for the period ended September 30, 2026 (filed November 4, 2026); Arm Holdings Form 6-K, November 2026 (filed November 4, 2026); Exhibit 99.1: Arm Holdings plc Reports Results for the Second Quarter of the Fiscal Year Ending 2027; Exhibit 99.2: Arm Holdings Q2 FY2027 Letter to Shareholders.Sources