At the Arm Everywhere China conference held in Shanghai on September 8, Arm unveiled a comprehensive refresh across three product lines, covering mobile, cloud, and physical AI. The company introduced its second-generation compute subsystem, CSS for Mobile 2, for the mobile segment, launched the Neoverse CSS N4 for cloud applications, and continued advancing the AGI CPU it first unveiled in March, while also expanding its Total Design ecosystem for physical AI.
The most noteworthy shifts, however, are taking place in the cloud computing arena.
Historically, Arm has generated revenue primarily through architecture licensing and IP royalties, with customers like Qualcomm, Apple, Amazon, and Google handling their own chip designs. Now, with a roadmap spanning from IP to CSS and extending to a delivery-ready AGI CPU, Arm is progressively absorbing the stages of the design process that were once left to its customers into its own product portfolio.
A number of Chinese companies are also lining up to join the Arm camp. Volcano Engine is planning to bring Arm's AGI CPU into its data centers for next-generation AI services, Lenovo is developing AI infrastructure around the AGI CPU, and Arm's collaboration with New Tsinghua Unigroup covers both the AGI CPU and CSS.
As the Arm server ecosystem continues its expansion, the x86 stronghold, long dominated by Intel and Advanced Micro Devices, is facing mounting pressure on market share. For Arm, the new challenge has evolved from getting more chips to adopt its architecture to selling more chips directly to customers itself.
A Three-Pronged Strategy
Arm's most profitable business model has always been enabling others to mine for gold.
For years, companies such as Qualcomm, MediaTek, Apple, and, more recently, cloud players like Amazon and Google have designed their own chips based on Arm's architecture and IP. Arm has not needed to bet on which player would ultimately win the market; as long as more chips adopted its technology, it could share in the growth through licensing fees and royalties.
This core business remains highly lucrative. In the first quarter of fiscal 2027, Arm posted revenue of $1.289 billion, up 22.4% year-over-year, with licensing revenue at $574 million and royalty revenue at $715 million, achieving a non-GAAP operating margin of 41.2%.
Behind this model, smartphones have remained Arm's largest revenue engine. In fiscal 2025, smartphone application processors still contributed roughly 45% of Arm's royalty revenue, yet the corresponding market share has already reached 99%, leaving little room for further growth.
It is against this backdrop that Arm has been aggressively seeking new growth avenues.
AI is clearly the largest incremental market on the horizon. Under Arm's long-term plan announced this year, the combined market opportunity for Edge AI, Physical AI, and Cloud AI is expected to expand from approximately $535 billion today to over $1.5 trillion by fiscal 2031, with cloud AI serving as one of the most significant sources of incremental growth.
This is also a key driver pushing Arm to move its business further down the value chain, beyond its traditional upstream position.
From IP, to CSS that already handles substantial system integration for customers, to the production-ready AGI CPU introduced for the first time in March, Arm's deliverables to customers are becoming increasingly complete.
Arm's Chief Marketing Officer, Ami Badani, explained in discussions with media, including Wall Street News' All-Weather Tech, that this trajectory is largely driven by customer demand. After Arm introduced its compute subsystems (CSS), a growing number of customers expressed a desire for Arm to go one step further and offer products closer to finished chips.
"That's a key reason we launched the Arm AGI CPU," Badani said, indicating this represents a further extension of Arm's business model.
At the September 8 Arm Everywhere China conference, Arm placed this expansion of its business boundaries squarely on the table.
In the mobile market, where Arm is most established, the company unveiled its second-generation mobile compute subsystem, CSS for Mobile 2, integrating the new C2 CPU cluster, Mali G2-Ultra NX GPU, system IP, physical implementation solutions, and software support into a single platform.
A notable feature of the Mali G2-Ultra NX is that it is the first Mali GPU to incorporate a dedicated neural network accelerator. Arm is also extending its partnerships further down the application chain: vivo is bringing the related technology to its flagship smartphones, NetEase's "Where Winds Meet" is advancing neural supersampling implementation, and Tencent Games and Unity China are participating in adaptation efforts from the game middleware and engine levels, respectively.
The cloud-side changes are even more pronounced.
The newly released Neoverse CSS N4 is Arm's most configurable cloud CSS generation to date, supporting up to 128 cores on a single die, LPDDR6 memory, and PCIe Gen 7, with the potential to further compress chip development timelines.
The AGI CPU, meanwhile, places Arm in a position much closer to the end product. This data-center-focused CPU is built on a 3nm process, currently featuring a 136-core design, with support for PCIe Gen 6 and CXL 3.0.
With these offerings, Arm now has three distinct product depth tiers: customers with comprehensive in-house chip design capabilities seeking greater differentiation can continue to license IP; those who wish to design their own chips but reduce underlying integration work and shorten development cycles can adopt CSS; and customers prioritizing rapid deployment without the need to design their own CPUs can directly adopt the AGI CPU.
These three product forms correspond to different trade-offs customers make between autonomous design, development lead times, and deployment speed, thereby widening the range of customers Arm can serve.
However, as Arm extends its reach toward finished chips, the once-clear upstream-downstream relationships are beginning to show signs of potential competition.
AWS already offers Graviton, Google has Axion, Microsoft has Cobalt, and NVIDIA has also introduced Vera based on Arm architecture.
Previously, these companies' Arm-based custom CPUs represented a steady stream of IP licensing and royalty revenue for Arm. But with Arm's AGI CPU entering the market, direct competition has emerged.
In response, Badani told All-Weather Tech and other media outlets that the AGI CPU is defined as an extension of the market space, emphasizing that IP, CSS, and the AGI CPU will coexist for the long term.
Badani cited SAP as an example: a portion of that company's cloud business already runs on Arm-based AWS Graviton servers, but a significant amount of enterprise on-premises deployments still run on other architectures. For Arm, entering AWS through IP does not automatically grant access to all of SAP's computing environments.
The AGI CPU provides a direct pathway into these enterprise on-premises deployment scenarios. Arm's ambition is to use this to extend its architecture into markets that were previously difficult to reach with IP and CSS alone, while also encouraging more software to be adapted for Arm.
The Chinese Camp Takes Shape
With the rise of the agent trend, the CPU market is poised for expansion.
Arm's management anticipates that as Agentic AI workloads increase, the configuration ratio of CPUs to GPUs may gradually move closer to 1:1.
This is opening up incremental space for CPUs. JPMorgan forecasts that shipments of CPUs dedicated to Agentic AI will surge from 1.6 million units in 2025 to 27 million units by 2028, representing a compound annual growth rate of 155%.
Several Chinese companies have already announced concrete plans to adopt Arm's in-house chips.
Arm revealed that Volcano Engine is building next-generation cloud services on the Arm platform and plans to launch an agent sandbox solution based on its AGI CPU.
When All-Weather Tech followed up on the progress of this collaboration at the event, Eddie Ramirez, Arm's Vice President of Cloud AI Business Expansion, stated that the two sides are working closely to bring the AGI CPU into Volcano Engine's data centers and run next-generation AI services on top of it. The specific deployment timeline remains to be disclosed by Volcano Engine at a later date.
Lenovo is also developing next-generation AI infrastructure around the AGI CPU. Arm's partnership with New Tsinghua Unigroup covers both the AGI CPU and CSS, extending further into agent orchestration, CXL memory pooling, rack-level system architecture, and edge AI.
However, while the server CPU market is expanding, the existing competitive landscape is also being redrawn.
Leading the charge is the long-dominant x86 camp.
Historically, the server CPU market has been controlled primarily by Intel and Advanced Micro Devices. Today, Arm-based server CPUs are entering data centers from multiple directions.
This means x86 is confronting the expansion of the entire Arm server ecosystem.
JPMorgan estimates that Arm's share of the server CPU market could rise from approximately 22% in 2025 to 43% by 2028. Even if the market expansion driven by Agentic AI allows x86 to maintain growth, Intel and Advanced Micro Devices are likely to face continued pressure on their market share.
Competition within the x86 camp itself is equally intense. Advanced Micro Devices continues to steadily take share from Intel. JPMorgan estimates that Advanced Micro Devices' server CPU share had already reached approximately 44% by the first quarter of 2026, and the company is pushing forward with next-generation products like Venice and Verano to capture incremental demand from AI servers.
For x86, the immediate challenge is defending its share; for Arm, the bigger question is how to convert its architecture dividend into a substantial chip revenue stream.
Agentic AI is putting server CPUs back at the center of the table. Arm already commands a growing architecture base; the next step is proving whether it can build a sufficiently large chip business on top of that foundation.
Comments