Arm’s AI Growth Accelerates as Its Bigger Data Center Bet Begins

Arm reached Google News after reporting record first-quarter revenue of $1.29 billion, up 22% from a year earlier. Profit also rose as AI infrastructure generated more licensing activity and data center royalties.

The result offers stronger evidence that Arm is moving beyond its dependence on smartphones. Data center royalties more than doubled, while demand for the company’s new AGI CPU exceeded $2 billion across fiscal 2027 and 2028. Arm says it has delivered initial products to multiple customers.

The tension sits beneath those numbers. Arm is challenging the x86 architecture used by Intel and AMD while also changing its own business model. The company now wants to sell finished processors alongside the intellectual property that customers use to design their own chips.

That shift opens a larger source of revenue. It also brings manufacturing constraints, higher research spending, customer conflicts, and execution risks that Arm’s traditional licensing business largely avoided.

Arm’s Record Quarter Was Driven by More Than One AI Bet

Arm’s latest results show AI demand reaching both sides of its business: intellectual property and processor royalties.

Revenue for the quarter ended June 30 reached $1.289 billion, compared with $1.053 billion one year earlier. Royalty revenue rose 22% to $715 million, while license and other revenue increased 23% to $574 million.

Royalty revenue comes from chips shipped with Arm technology. Licensing revenue generally arrives when customers secure rights to use the company’s architectures, processor designs, or related technology.

That distinction matters because the two revenue streams respond to different timelines. A license can contribute revenue while a future processor remains under development. Royalties arrive later, after the resulting chips enter products and begin shipping.

Arm attributed the royalty increase to greater adoption of Armv9, its newer instruction-set architecture, and Arm Compute Subsystems. A compute subsystem packages processor cores and supporting components into a more complete design that customers can adapt.

The company also identified cloud AI as its largest royalty growth driver. Data center royalties more than doubled for another quarter, supported by Arm-based server chips and networking processors.

According to Arm’s quarterly results, Neoverse shipments have passed 1.5 billion processor cores. The most recent 500 million shipped within nine months, while the first billion required six years.

Neoverse is Arm’s family of processor designs for cloud infrastructure, networking, and high-performance computing. Customers can use those designs to build their own server processors.

Amazon uses Arm technology in its Graviton chips. Google uses it in Axion processors, while Microsoft has adopted it for Azure Cobalt. Nvidia’s Grace and Vera CPUs also use the architecture.

Those programs already place Arm inside major AI systems without requiring Arm to manufacture a finished server processor. Each customer designs a chip around its own workloads, software, accelerators, and infrastructure priorities.

Arm also reported annualized contract value of $1.732 billion, up 13%. This measure normalizes licensing commitments that can otherwise fluctuate with the timing of large agreements.

The quarter included $193 million in license revenue from an agreement with SoftBank, Arm’s controlling shareholder. Management expects that agreement to contribute around $200 million per quarter for the rest of the fiscal year.

That concentration deserves attention. The SoftBank contract supports reported licensing growth, but it is a related-party arrangement rather than independent evidence of broad customer demand.

Even so, the royalty figures provide a separate signal. Customers must ship products before those royalties appear. A second consecutive period of data center royalties more than doubling indicates that earlier designs are reaching commercial deployments.

GAAP net income reached $270 million, compared with $130 million one year earlier. Diluted earnings per share rose to $0.25 from $0.12.

Non-GAAP net income increased 28% to $480 million. Non-GAAP earnings per share reached $0.45, above the upper end of Arm’s prior guidance.

The combined picture explains the positive Google News headline. Arm produced higher revenue and profit while its data center presence expanded. Yet the most consequential part of the report concerns revenue that has barely started.

Why Google News Is Focusing on Arm’s Data Center Shift

The quarter matters because Arm is becoming a central processor platform for AI infrastructure, not merely an indirect beneficiary of smartphone demand.

AI servers are commonly described through their accelerators, including Nvidia GPUs and Google TPUs. Those accelerators perform the large parallel calculations required for training and inference.

CPUs still coordinate the wider system. They prepare data, manage storage and networking, run operating services, retrieve information, and direct tasks between tools. Agentic AI increases this coordinating workload because agents repeatedly plan, call services, check results, and continue executing.

Arm argues that these workloads favor processors with many energy-efficient cores. Data centers face fixed limits on electrical power, cooling, rack space, and available construction capacity.

A processor that completes more work within those limits can improve the economics of an AI deployment. That proposition has helped Arm move into territory once dominated by Intel Xeon and AMD EPYC processors.

The latest industry figures support the direction, although they require careful interpretation. IDC data reported by server market analysis put first-quarter global server revenue at $122.6 billion, up 30.4% annually.

The same data indicated that Arm-based systems represented more than 45% of server revenue. X86 systems retained about 52%.

However, revenue share does not equal CPU shipment share. Expensive accelerated systems can contain Arm CPUs alongside large numbers of GPUs, memory components, networking devices, and storage products.

A rack’s total selling value can therefore rise sharply without a comparable increase in Arm processor units. Intel and AMD remain deeply embedded in conventional servers and enterprise software environments.

Arm’s own royalty data provide a narrower measurement. Data center royalties more than doubled, but the company does not disclose the dollar amount or its share of total royalties.

That omission prevents readers from calculating the exact contribution. A rapidly growing segment can still be relatively small when measured against Arm’s established smartphone business.

Management acknowledged that smartphones remain important. Higher memory costs have reduced handset demand, with Arm describing the market as potentially down by a double-digit percentage.

Arm expects higher royalty rates from Armv9 and compute subsystems to offset much of that weakness. In effect, the company can earn more from each advanced chip even when fewer devices ship.

Cloud AI adds another offset. CFO Jason Child said during the earnings call that Google Axion deployments, Amazon Graviton systems, and Nvidia Vera were exceeding earlier plans.

That is the quarter’s central reversal. Arm’s smartphone exposure remains a drag, but data center growth is becoming large enough to change the overall revenue narrative.

The shift also spreads Arm through several layers of AI infrastructure. Its cores appear in host CPUs, data processing units, SmartNICs, and controllers alongside the main accelerators.

A DPU is a processor that handles data movement, networking, security, or storage tasks previously assigned to a server CPU. A SmartNIC performs similar offloaded functions through an advanced network interface.

Arm says its technology appears in nearly all leading products across these categories. That claim is difficult to verify from the reported financial figures, but the customer list shows broad architectural adoption.

The pressure falls most directly on Intel and AMD. Both companies sell finished processors and depend on x86 software compatibility as an advantage.

They now face custom Arm chips from their largest cloud customers. They also face Nvidia’s Arm-based CPU strategy and Arm’s decision to become a processor vendor itself.

Arm’s AGI CPU Turns a Partner Into a Competitor

Arm’s new processor creates a larger commercial opportunity, but it also changes the relationships that made Arm successful.

For most of its history, Arm supplied designs rather than competing directly with the companies using them. That model let Apple, Qualcomm, Amazon, Google, Microsoft, and other customers differentiate their processors.

The AGI CPU changes that boundary. Announced in March 2026, it is Arm’s first production silicon product for AI data centers.

The processor targets head nodes, general-purpose servers, and agentic AI workloads. Head nodes coordinate accelerators and other resources inside an AI cluster.

Arm has said the current design uses 128 cores. The product is intended to provide the CPU capacity needed around GPUs and custom accelerators rather than replace those accelerators.

Customers identified around the launch included Meta, OpenAI, Cloudflare, Oracle, and Cerebras. Arm also named manufacturing and server partners supporting system availability.

Arm initially described an opportunity of more than $1 billion across fiscal 2027 and 2028. It now says customer demand exceeds $2 billion for that period.

Demand is not recognized revenue. It can include customer interest, pipeline value, expected orders, or commitments with different commercial conditions.

Arm has not published a detailed reconciliation showing how much of the figure represents firm, noncancelable orders. Readers should therefore treat it as a management demand indicator.

The company says it has secured enough manufacturing capacity for the initial $1 billion opportunity. It is pursuing more wafers, memory, substrates, packaging, and test capacity to serve additional orders.

That supply work is unfamiliar territory compared with licensing intellectual property. A chip vendor must coordinate fabrication schedules, component purchases, assembly, quality control, inventory, and customer delivery.

Arm’s annual filing said the AGI CPU had no material effect on fiscal 2026 revenue. The latest quarter did not provide a separate silicon revenue figure.

Management plans to give another update with its third-quarter results, when it expects better visibility into the final quarter of fiscal 2027 and fiscal 2028.

This creates a timing gap. Investors can see rising development costs today, while the processor’s financial contribution remains largely ahead.

The direct processor business also complicates Arm’s relationship with licensees. Amazon, Google, Nvidia, and Microsoft use Arm intellectual property to develop processors that can overlap with the AGI CPU.

Qualcomm has announced plans to enter the AI data center CPU market with an Arm-based chip. Every additional licensee expands Arm’s platform while adding another potential competitor to its silicon product.

That tension is not necessarily fatal. The semiconductor industry often supports overlapping suppliers, custom designs, and reference products.

Arm can present the finished processor as another way to adopt its platform. Customers that want differentiated silicon can continue licensing designs, while others can buy a completed CPU.

Still, decisions around product roadmaps and customer access will become more sensitive. A licensee will want confidence that Arm does not reserve its best technology for its own processor.

Arm must also avoid appearing to use one customer’s market information against another. Its neutral platform role has been an important reason competitors could adopt the same architecture.

The AGI CPU therefore tests whether Arm can remain an ecosystem supplier while becoming a merchant silicon vendor. Higher revenue alone will not settle that issue.

The Real Contest Is Arm Versus x86 Economics

Arm’s challenge to Intel and AMD depends on total deployment economics, not architecture labels or benchmark headlines.

X86 processors benefit from decades of enterprise software, administrative experience, and established purchasing relationships. Many workloads can move between newer systems, but migration still requires validation.

Cloud providers have reduced that obstacle by controlling the full environment. They can optimize operating systems, compilers, applications, and hardware together.

Amazon used this approach with Graviton. Google followed with Axion, while Microsoft developed Cobalt for Azure.

These processors let hyperscalers reduce dependence on third-party CPUs and tailor systems around their own services. Arm gains royalties even when it does not sell the completed chip.

The AGI CPU addresses buyers that lack a hyperscaler’s silicon engineering organization. A finished processor gives cloud providers and AI companies another route to Arm-based infrastructure.

Arm’s advantage is clearest when customers prioritize core density and energy efficiency. Agentic workloads can create large numbers of parallel tasks that need coordination around accelerator clusters.

However, the processor cannot succeed through core count alone. Buyers evaluate memory bandwidth, latency, networking, accelerator connections, software support, reliability, and sustained performance.

They also consider procurement simplicity. Intel and AMD offer mature server roadmaps supported by large original equipment manufacturers and extensive enterprise software certification.

AMD has steadily expanded its data center position with EPYC, while Intel continues investing in Xeon products and its manufacturing process. Both vendors can respond through pricing, packaging, performance, and supply agreements.

Arm also competes with its own customers’ custom processors. Google can pair Axion with TPU infrastructure, while Amazon can integrate Graviton with AWS software and networking.

Nvidia can optimize Vera around its accelerator roadmap and NVLink interconnect. NVLink is Nvidia’s high-speed connection for moving data between processors and accelerators.

Arm’s standalone processor needs credible paths into these mixed systems. Haas noted that Nvidia’s NVLink Fusion could connect third-party Arm CPUs to Nvidia accelerators, although Arm announced no specific integration.

That answer identifies an important technical dependency. AI customers rarely choose a CPU in isolation. They select a system architecture built around accelerators, memory, networking, and software.

Intel and AMD have their own integration strategies. AMD combines EPYC processors with Instinct accelerators, while Intel can coordinate Xeon CPUs with networking and accelerator products.

Arm’s business model provides a different form of leverage. Even when a customer selects Amazon, Google, Microsoft, Nvidia, or Qualcomm silicon, Arm can still collect a license fee or royalty.

This makes the opponent map more nuanced than Arm against one chip company. The central contest is Arm architecture against x86 economics across the broader data center.

Arm does not need every customer to buy its AGI CPU. It benefits when an Arm-based design wins, regardless of which partner supplies the finished processor.

The finished chip increases Arm’s potential value capture from selected deployments. It also gives the company a product that can establish performance and implementation standards for its wider platform.

If the strategy works, Arm can earn through three layers: processor intellectual property, compute subsystems, and completed silicon. Each layer offers more content and revenue per deployment.

The risk rises with each layer. Finished processors require more capital, operational coordination, and commercial judgment than an architecture license.

That tradeoff explains why the latest earnings deserve more than a celebratory AI-growth headline. Arm is no longer only enabling the challenge to x86. It is joining that challenge directly.

What the Higher Profit Does Not Show

Arm’s profit increased, but its GAAP operating figures reveal the cost of expanding into more complete products.

GAAP operating expenses rose 28% to $1.162 billion. Research and development expense increased 29% to $838 million.

GAAP operating income fell 20% to $91 million despite the 22% increase in revenue. The GAAP operating margin declined to 7.1% from 10.8%.

Net income still more than doubled, supported by items outside operating income. That difference makes operating performance an important counterweight to the headline profit figure.

On a non-GAAP basis, the picture was stronger. Operating income rose 29% to $531 million, and operating margin increased to 41.2%.

The gap reflects exclusions including share-based compensation and other adjustments. Neither presentation should be viewed alone.

GAAP results capture expenses borne by shareholders, while non-GAAP figures can help compare underlying operations. The growing difference becomes more important as Arm adds engineers and invests in silicon.

Arm employed 9,584 people at the end of fiscal 2026, including 8,058 engineers. Annual research and development spending had already risen 34% to $2.776 billion in that year.

The company says current investment supports next-generation architectures, compute subsystems, and the AGI CPU family. These programs can generate future revenue, but their returns remain unproven.

Supply is another limitation. Management described tight availability across wafers, memory, substrates, and testing equipment.

Securing capacity for a product does not guarantee that customers will deploy it at scale. Qualification, software optimization, system integration, and data center construction can delay shipments or revenue recognition.

The reported $2 billion demand figure also needs clearer definition. Arm has not disclosed the number of processors, customers, delivery schedule, or cancellation protections behind it.

Competition can change before all that demand converts. Intel, AMD, Nvidia, Qualcomm, and custom-chip teams will continue introducing products during fiscal 2027 and 2028.

Arm’s disclosure choices add another uncertainty. Starting this quarter, it stopped reporting remaining performance obligations and the number of Total Access and Flexible Access licenses.

The company said those indicators became less relevant after its expansion into production silicon. That reasoning is understandable, but investors lose consistent measures of the legacy licensing business.

Annualized contract value remains available. Its 13% growth was healthy, although slower than the quarter’s 23% increase in licensing revenue.

Related-party revenue deserves continued scrutiny as well. Arm’s SoftBank agreement contributed roughly one-third of quarterly licensing revenue.

Arm disclosed substantial revenue from SoftBank affiliates in fiscal 2026. These contracts can represent legitimate engineering work, but they make customer concentration and contract terms more consequential.

Smartphone weakness remains the immediate offset. Arm expects newer technology and higher royalty rates to support handset royalties despite falling industry units.

That strategy depends on customers continuing to adopt Armv9 and more complete compute subsystems. A weaker product mix or slower device replacement cycle would reduce the protection.

Cloud growth can compensate, as it did this quarter. Yet Arm still does not provide enough segment data to calculate exactly how much cloud AI must grow to offset mobile weakness.

Readers tracking the story through Google News should therefore separate three claims. Arm’s reported revenue and profit growth are confirmed financial results. Data center royalty growth is confirmed directionally but lacks a disclosed base.

The $2 billion AGI CPU figure is a company measure of demand, not completed sales. Its conversion will determine whether the silicon strategy becomes a material business.

Three Signals Will Test Arm’s AI Growth Story

The next phase depends on converting processor demand, sustaining cloud royalties, and controlling the cost of becoming a chip supplier.

The first signal is Arm’s promised AGI CPU update with its third-quarter results. Management expects better visibility then into fiscal 2027’s final quarter and fiscal 2028.

A higher secured-capacity figure, disclosed revenue, or firmer order classification would strengthen the claim that AGI CPU demand is converting. Another broad pipeline statement without delivery detail would leave the central uncertainty unresolved.

The second signal is data center royalty growth. Another quarter above 100% would show that Arm-based processors are still moving from design announcements into deployed systems.

The quality of that growth matters too. Evidence spanning Amazon, Google, Microsoft, Nvidia, networking chips, and independent cloud customers would reduce reliance on any single program.

A sharp slowdown would not erase the architectural shift. It would suggest that the current comparison benefited from a small base, uneven shipment timing, or concentrated hyperscaler deployments.

The third signal is operating leverage. Arm guided to second-quarter revenue of $1.38 billion, plus or minus $50 million, and non-GAAP operating expenses near $780 million.

Management expects non-GAAP earnings per share of $0.47, plus or minus $0.04. Royalty growth is expected in the low teens, while licensing revenue should rise about 30%.

Those figures will show whether licensing can offset weaker smartphone royalties without making results increasingly dependent on large, uneven contracts.

GAAP research spending and operating margin will reveal the deeper cost. If expenses continue rising faster than revenue before silicon sales appear, the strategy will demand more patience.

Arm’s quarter confirms that AI demand is already benefiting its established platform. It does not yet prove that Arm can manufacture and sell processors at the scale suggested by customer interest.

For developers, the immediate question is software portability. More Arm capacity across major clouds makes multi-architecture testing, dependency management, and performance measurement increasingly relevant.

Infrastructure buyers should compare complete system economics. Processor efficiency matters, but so do accelerator compatibility, available instances, software support, supply guarantees, and migration work.

Teams following these developments can preserve earnings notes, architecture comparisons, and deployment findings inside a searchable technical knowledge base. That record helps separate confirmed adoption from changing vendor claims.

The latest Google News cycle marks a real milestone, not the end of the contest. Watch what Arm ships, what customers deploy, and how much the company spends to get there.