ARM控股

Arm's first-quarter results met consensus expectations, but investors were disappointed by the lack of firmer revenue guidance around the Arm AGI CPU and questions looming around Arm’s ability to secure enough supply.

Why it matters: Arm’s management raised its demand outlook for the AGI CPU to “above $2 billion” for 2027-28, up from $2 billion last quarter. However, the demand statement was not accompanied by full confirmation of additional supply capacity. While management noted progress on the supply side, there were no concrete commitments, leaving some investors wanting more. Given the lofty expectations priced into Arm’s stock, we believe investors were looking for more concrete demand and supply numbers. Management tempered its fiscal 2027 smartphone royalty growth guidance to high teens, down from 20% previously—a risk we have flagged in prior notes.

The bottom line: We maintain our $190 per share fair value estimate for wide-moat Arm. Our 2031 revenue and EPS forecasts—$28.6 billion in revenue and $10 in EPS—sit above management’s long-term guidance from March, given CPU market expectations have expanded since then on agentic artificial intelligence opportunities. We caution against assuming a larger CPU market automatically translates into proportionally larger demand capture for Arm. Peer AMD now forecasts a $200 billion CPU market in 2030 versus $120 billion previously—but a larger pie doesn’t mean Arm’s slice grows linearly. Arm still needs to secure supply and compete not only against x86 providers AMD and Intel, but also against its own royalty-paying. For now, we model $17 billion in 2031 CPU sales versus management’s initial $15 billion guidance. Capturing an incremental CPU sale through a royalty-paying licensee versus through Arm’s own silicon accrues the same value to the Arm ecosystem but radically different economics to Arm Holdings. Which channel captures more demand will have real implications for Arm’s 2031 revenue ambitions.

Our fair value estimate is $190 per share.

We model $2.20 in adjusted earnings per share in fiscal 2027 and $2.80 in 2028, representing forward P/E multiples of 85 times and 68 times, respectively. Arm's profitability inflection should occur later in the decade, once its own CPU business has reached scale. We model adjusted EPS of $6.50 and $9.90 in fiscal 2030 and 2031, respectively.

We model Arm royalty and licensing revenue to grow at around 20% over the next five years. The contribution of the CPU business is insignificant until 2028, when it reaches $1.5 billion, and it scales to $17 billion by fiscal 2031, above the outlook given at the Arm Everywhere event in March 2026, as we believe the CPU outlook has improved since then. Overall, we model a 40% compound annual revenue growth rate over the next five years, with revenue reaching $28.6 billion by fiscal 2031 and an operating margin of 35%.

We expect royalty revenue to grow in the 20%-30% range and licensing revenue in the mid- to high-single-digit range over the long term. We expect Arm’s blended royalty rate to expand to more than 5% in 2030 and to more than 8% over 10 years, as v9 and compute subsystems account for a larger share of revenue. We expect adoption of these architectures to be faster in smartphones, cloud, and automotive, while the Internet of Things remains at lower royalty rates given its lower computing demands. After our 10-year explicit period, we model returns on new invested capital above 30% and high-single-digit profit growth for another 10 years in stage two of our model, in line with the modeling for a wide-moat firm.

By market, we expect Arm to maintain a very high share in mobile processors. Mobile revenue might face pressures in 2026 and 2027, as smartphone demand is expected to compress as handset prices rise due to higher memory costs, although this will be partially offset by higher royalty rates as customers migrate to v9 and compute subsystems. We believe Arm will continue to gain market share in cloud computing and automotive. Energy consumption is one of the data center’s main expenses, so energy efficiency is paramount. In the automotive industry, the transition to electric vehicles is driving demand for more efficient chips. In sensors and the Internet of Things, we expect more pressure from RISC-V, as this is where the competing architecture is placing more emphasis.

The CPU and royalty businesses have different profitability profiles, given their distinct nature and engineering intensity. We model a 95% gross margin for the royalty business over the long term, while the CPU business will reach gross margins of around 50% once fully scaled, in line with peers Qualcomm, AMD, and MediaTek.

We assign Arm Holdings a wide Morningstar Economic Moat Rating based on intangible assets and switching costs. Arm is the IP owner and developer of the Arm architecture, which powers 99% of the world’s smartphone CPU cores. It also has a high market share in other battery-powered devices like wearables, tablets, and sensors and is gaining market share in the data center segment.

The instruction set architecture is the set of instructions that dictates how the processor's hardware behaves when it receives a software instruction. Examples of architecture instructions are arithmetic instructions (addition, subtraction, and multiplication), memory instructions for the transfer of data between CPU registers and memory, and data pathway instructions, which define the speed at which data moves between the CPU and registers (temporary storage locations). Software written for a certain architecture won’t work straightforwardly in a different architecture, creating switching costs.

Arm architecture is optimal for battery-powered devices, since it is simpler and consumes less power than x86 architecture. Traditionally, x86 has been associated with higher computational performance, but this comes at the expense of increased power consumption. In contrast, Arm architecture consumes less energy but normally offers less computational power, making it ideal for battery-powered devices. For years, and at various levels of urgency, the two architectures have been striving to meet in the middle. Intel initially strove to develop x86 mobile processors but ultimately backed away from these efforts. Meanwhile, Arm and its customers have been more successful at improving the processing power of its chips, notably with Apple’s M-series processors as well as data center processors like Amazon’s Graviton, Google Axion, or Nvidia's Grace/Vera.

Once a company commits to developing a chip in a particular architecture, it is highly likely that the next version of the chip will be designed in the same architecture. Changing the ISA requires repurposing all designs and software, which can be time-consuming and inefficient from an engineering perspective, resulting in high switching costs. On the contrary, making small, evolutionary improvements in the same architecture is much easier. Apple provides a good example of the challenges of switching architecture.

Arm licenses its architecture to chip designers for a fee, offering different types of licenses depending on the flexibility the customer needs. Large clients like Apple or Qualcomm buy architectural licenses, which give them the freedom to modify the architecture and add or delete instructions to tailor the chips to their specific needs. In off-the-shelf agreements, Arm licenses a ready-to-use portfolio of CPU designs that the client can incorporate into its devices. The portfolio ranges from simple to more advanced CPUs. Both off-the-shelf and architectural customers pay a royalty fee per chip shipped. Arm's royalty rates for its newest architectures, compute subsystems, and v9 are significantly higher than for previous generations, at around 10% and 5%, respectively. This means if a $1,000 smartphone has $100 of Arm content at a 10% royalty rate, Arm will get $10. Arm’s business model creates a predictable royalty revenue stream while remaining capital-light, with a focus on R&D and engineering.

By licensing Arm, customers gain immediate access to a rich ecosystem with multiple off-the-shelf CPU designs, customer support, hundreds of hardware and software vendors developing tools in Arm architecture, and a community of 15 million developers. Clients also accelerate time to market and reduce engineering complexity, as designing CPUs from scratch can take immense time and cost hundreds of millions of dollars. Arm’s competitive position in smartphone cores is unmatched, and it could take decades for a competitor to even scratch at its wide moat. As applications become more complex, they need more cores and more expensive ones, giving Arm capital-light revenue upside.

Aside from its dominance in the smartphone and wearables markets, Arm is making inroads into the data center market. Historically, data centers were built with monolithic architectures. The data center relied on a single, large computing network with a single code that handled all tasks. With the arrival of cloud computing and the need to scale, data center architectures are evolving toward containerized services; data center workloads are divided into discrete units, with tasks broken into smaller pieces. The trend toward containerization is favoring Arm-based CPUs, as they have enough computational performance in a containerized environment while consuming less power than x86 chips. Energy consumption is a data center’s main operating expense, so having more efficient CPUs allows it to fit more of them, increasing throughput and reducing unitary power consumption. Amazon, Google, and Microsoft have developed their own Arm-based data center CPUs. We expect Arm to continue to gain market share in this space.

In 2026, Arm announced it would expand into the fabless space, selling its own data center CPUs on top of licensing intellectual property. This means that from now on, Arm will directly compete with some of its own customers like Nvidia, Qualcomm, MediaTek, and the hyperscalers while simultaneously licensing IP to them. This will transform Arm into a hybrid company with a capital-light division that will continue to collect royalties from customers and another division that will sell its own CPUs.

We expect Arm will keep gaining data center market share from x86 architecture, as its chips consume less power and data centers need to minimize energy consumption. We also expect gains in automotive shares, driven by the transition to EVs.

The new data center CPU business allows Arm to capture the whole value of the chip sale, providing meaningful revenue upside for years to come.

Arm's architecture is the most energy-efficient. In a world full of battery-powered devices and data center power constraints, this is a key advantage.

By entering the CPU market, Arm is competing with its own customers, treading in dangerous territory and creating a conflict of interest.

Arm still needs to secure the manufacturing supply to capture CPU upside and compete for such supply with competitors like Nvidia or AMD.

We see downside risk to Arm’s revenue in 2026 and 2027 driven by a weak smartphone market, where demand is expected to compress as handset prices rise due to higher memory costs.