ARM Semiconductors · AI infrastructure · IP royalties · High valuation · Thesis updated July 16, 2026

Royalty step-ups meet a rich stock price

Arm makes most of its money from chip design licenses and per-chip royalties.
The bull case is that Armv9 and CSS raise royalty rates even when device units are weak.
The new Arm AGI CPU adds direct silicon sales, with more than $2 billion of demand cited for FY27 and FY28.
Supply limits matter now because management kept near-term AGI CPU revenue capacity at $1 billion.
The stock already prices in a lot of future growth, so execution has little room for error.

01 Running thesis

Higher royalties, harder execution

Arm is trying to make revenue less tied to the number of chips sold. The key is mix. Armv9 carries higher royalties than older v8 designs, and CSS, or Compute Subsystems, packages more Arm technology into a ready-made block. Management has said v9 is roughly double v8, and CSS is roughly double v9. New CSS deals have also moved above the old 10% royalty idea.

That matters because phone units are no longer a clean growth engine. Management said mobile unit growth flipped negative in the latest quarter. The bull case says premium phones, cloud servers, AI chips, and more complete Arm designs can still lift revenue per chip even if unit growth stays slow.

The second leg is new and riskier. Arm announced the 136-core Arm AGI CPU for agentic AI systems, where many AI agents need CPUs to coordinate work. Management said it has line of sight to more than $2 billion of demand across FY27 and FY28, but is holding the near-term outlook at $1 billion while it seeks more wafer, memory, and supply chain capacity.

This is a strong company, but the stock is not cheap. A lot has to go right: v9 and CSS must keep lifting royalties, data center royalties must keep more than doubling, direct silicon must scale without hurting margins too much, and legal noise with Qualcomm must not damage the ecosystem.

May 2026▲Arm formalized the DreamBig acquisition with a definitive agreement for about $265 million in cash. The deal adds networking technology that could help more complete AI and data center systems.

May 2026▲Management made direct silicon a clear second growth vector with the Arm AGI CPU. Demand is above $2 billion across FY27 and FY28, but near-term supply capacity is still capped at $1 billion.

Feb 2026▲The data center story strengthened as management said it expects data center to become larger than mobile in a few years. Phone volume risk stayed real, but premium mix helped limit the royalty hit.

Nov 2025▲Neoverse royalties more than doubled year over year, and Lumex CSS started producing royalties earlier than expected. DreamBig was also announced as a way to add networking IP.

Jul 2025▲Management said new CSS royalty rates are above 10%, raising the ceiling on Arm's value per chip. The same update also brought more execution risk as Arm explored chiplets and fuller solutions.

May 2025→The Qualcomm and Nuvia legal dispute remained unresolved after an incomplete jury verdict and pending post-trial motions. The core AI and royalty thesis did not change.

May 2025▲Armv9 crossed more than 30% of royalty revenue, and smartphone royalties grew about 30% year over year despite flat unit growth. That supported the idea that Arm can grow through richer mix, not only more units.

Feb 2025▲Management reinforced the royalty ladder, saying v9 is roughly double v8 and CSS is roughly double v9. AI infrastructure wins and edge AI trends also supported the long-term growth case.

02 Business model

IP margins, silicon ambitions

Arm's core business is simple to explain. It designs CPU and related chip technology, then lets other companies use that technology in their own chips. Customers pay license fees up front or over time. When chips ship, Arm also collects royalties.

In fiscal 2026, total revenue was $4.920 billion. License and other revenue was $2.307 billion, or about 47% of revenue. Royalty revenue was $2.613 billion, or about 53%. Royalty revenue grew 21% year over year, helped by a better mix of higher royalty products such as Armv9.

The model has very high gross profit because Arm usually sells designs, not physical chips. Fiscal 2026 gross profit was 98% of revenue. The catch is heavy spending. Research and development was 56% of revenue in fiscal 2026, and it rose 34% year over year as Arm invested in next-generation products such as the Arm AGI CPU.

Direct silicon changes the risk profile. Internal targets call for about a 35% operating margin profile for direct silicon by 2031, compared with about 65% for IP. The upside is a much larger revenue pool. The risk is that Arm now has to compete for foundry, packaging, and memory supply instead of only selling blueprints.

03 Product portfolio

From blueprints to AI CPUs

Growth engine

Armv9 architecture

Armv9 is the newer instruction set, meaning the language software uses to talk to Arm CPUs. It is already more than 30% of royalty revenue and carries higher royalties than older v8 designs.

Growth engine

Compute Subsystems

CSS packages CPUs and related parts into a more complete, pre-tested block. It helps customers launch chips faster and lets Arm earn more value per chip.

Growth engine

Neoverse data center IP

Neoverse powers Arm's push into cloud and AI servers. Management said data center royalty revenue continues to more than double year over year, with no break in momentum.

Option

Arm AGI CPU

The Arm AGI CPU is a 136-core production silicon product for agentic AI orchestration. It could become a large new business, but near-term supply caps are the main bottleneck.

Cash cow

Mobile and Lumex CSS

Mobile remains Arm's best-known market. Lumex CSS shows the strategy of selling more complete mobile compute platforms, even while phone unit growth is soft.

Steady

Automotive, IoT, and networking IP

Automotive is growing at a solid double-digit pace, while IoT has bottomed but has not clearly recovered. DreamBig is meant to add advanced networking technology for scale-up and scale-out systems.

04 Business segments

Revenue mix today

Royalty revenue53%growing fast

License and other revenue47%modest

Arm reports revenue mainly as license and other revenue versus royalty revenue. The mix below uses fiscal 2026 revenue from the Form 20-F, not end-market revenue, because the filing does not give a full phone, cloud, auto, and IoT revenue split.

05 Risk factors

What could break the thesis

AGI CPU supply bottleneck

High impact · Medium odds

Arm has more than $2 billion of stated AGI CPU demand across FY27 and FY28, but management kept the near-term outlook at $1 billion while it seeks more capacity. This is a new kind of risk for Arm because finished silicon needs wafers, advanced packaging, and memory. If capacity does not open up, demand may not turn into revenue on time.

We watchManagement updates on AGI CPU supply capacity, customer demand, and FY27 or FY28 revenue outlook.

Smartphone unit weakness

Medium impact · High odds

Mobile unit growth flipped negative in the latest quarter. Arm can offset some of this with higher v9 and CSS royalties, especially at the high end. But if weakness spreads into premium phones, the offset gets harder.

We watchPremium smartphone shipment trends and Arm commentary on mobile royalty growth versus unit growth.

Direct silicon margin drag

Medium impact · Medium odds

Arm's IP business has a much higher margin profile than direct silicon. Internal targets point to about 35% operating margin for direct silicon by 2031 versus about 65% for IP. If AGI CPU revenue grows but margins disappoint, investors may question whether the new business deserves the same valuation as IP royalties.

We watchGross margin, operating margin, and any separate disclosure for production silicon products.

Qualcomm litigation

Medium impact · Medium odds

The internal thesis still treats Qualcomm litigation as a legal overhang. The latest 20-F adds that Arm appealed part of the Nuvia outcome, and a separate Qualcomm case is expected to go to trial in the fourth calendar quarter of 2026. The main risk is not one quarter of legal cost. It is whether a major customer relationship or license structure changes.

We watchThird Circuit appeal updates, Delaware trial timing, and any settlement or license change with Qualcomm.

China and trade controls

Medium impact · Medium odds

Arm depends on Arm China for access to the PRC market, and the 20-F says export rules could limit some direct or indirect business. New tariffs, entity list rules, or advanced chip controls can hurt customers that build Arm-based chips. This could slow royalties or delay license deals.

We watchNew U.S. or PRC export rules, Arm China revenue trends, and management comments on restricted customers.

Valuation reset

High impact · Medium odds

Arm has a strong royalty engine, but the market already expects years of growth from CSS, data center, and AGI CPU. If growth slows or silicon execution costs rise, the stock could fall even if the business remains healthy. This is the main reason the valuation setup is less forgiving than the operating story.

We watchRevenue growth, royalty growth, guidance changes, and any sign that data center royalties stop doubling.

06 Quick answers

In one breath

How does Arm make money?

Arm licenses chip designs and collects royalties when customers ship chips using Arm technology. In fiscal 2026, license and other revenue was about 47% of total revenue, and royalty revenue was about 53%.

Why is Arm important for AI?

AI systems still need CPUs to coordinate work, even when GPUs or NPUs do the heavy math. Arm is pushing into AI from phones to data centers, and the Arm AGI CPU is meant for agentic AI systems with many agents running at once.

Is Arm mainly a smartphone company?

Phones remain important, but the thesis is shifting toward cloud, AI data centers, automotive, and more complete CSS designs. Management has said data center could become larger than mobile in a few years.

What is the biggest risk for Arm right now?

The biggest near-term risk is execution in direct silicon. Arm has demand for the AGI CPU, but supply capacity is capping near-term revenue potential.