The Architecture Layer of AI Computing - But Priced Like the Future Already Arrived?

Rating: High-quality hold / speculative buy on pullbacks

Style: Semiconductor IP compounder with AI data-center optionality

Core debate: Is Arm becoming the central architecture layer for the AI computing era, or has the market already priced in the full success of its move from neutral IP licensor to direct AI CPU vendor?

Executive View

Arm is one of the highest-quality business models in semiconductors. It does not traditionally manufacture chips. It designs the CPU architecture and core IP that chipmakers use, then earns licensing revenue upfront and royalties when Arm-based chips ship. That model has historically produced exceptional gross margins, recurring royalty economics, and deep ecosystem lock-in across smartphones, consumer devices, automotive, cloud, and increasingly AI data centers.

The latest earnings were strong. Arm reported record Q4 FY2026 revenue of $1.49B, up 20% year over year, and adjusted EPS of $0.60, ahead of expectations. It guided Q1 FY2027 revenue to around $1.26B, also slightly above consensus. The important point is that this was not just a smartphone story: data-center royalties more than doubled, while customer demand for Arm’s new AGI CPU was stronger than management originally expected.

The catch is valuation. Arm is now being valued less like a normal semiconductor company and more like a future monopoly-like infrastructure layer for AI compute. The stock recently traded above $400 after a huge run, though it pulled back with the broader semiconductor selloff on July 13, when Arm fell 7.6% alongside other chip leaders.

So the thesis is not “Arm is cheap.” It is: Arm may be one of the most strategically important companies in computing, but the stock already assumes a very large part of that future.

Why Now - The AGI CPU Changes the Story

For most of its history, Arm was a neutral IP licensor. Apple, Qualcomm, Nvidia, MediaTek, AWS, Google, Microsoft, and many others could use Arm’s architecture without worrying that Arm itself would compete directly with them in finished silicon.

That changed in March 2026, when Arm announced the AGI CPU, a data-center processor for agentic AI workloads. Reuters reported that the chip is being manufactured by TSMC on 3nm, with Meta as the lead partner, and Arm expects it to generate roughly $15B in annual revenue within about five years. Management also said it expects Arm overall to reach about $25B of annual revenue and $9 of EPS in that timeframe.

That is the entire re-rating story. Arm is no longer only monetizing a small royalty on someone else’s chip. If AGI CPU works, Arm can capture full chip-level revenue in the AI data-center market. That is a massive TAM expansion.

The near-term demand signal is strong. Arm said the new product is expected to generate more than $2B across fiscal 2027 and fiscal 2028, but CEO Rene Haas also noted that supply is currently available for only the first $1B of demand, with the company working to secure more wafers, memory, packaging, and test equipment.

That is a good problem - but still a problem.

What Arm Does

Arm is the architecture layer behind much of modern computing. The company designs CPU architectures, processor cores, and related system IP that other chipmakers use to build their own semiconductors. Arm usually does not manufacture the chips itself. Instead, it licenses its technology to companies such as Apple, Qualcomm, Nvidia, Amazon, Google, Microsoft, MediaTek, and Samsung, then earns royalties when Arm-based chips are shipped.

This model has made Arm one of the most widely used technology platforms in the world. Its architecture powers almost every smartphone, Apple’s Mac chips, Nvidia’s Grace CPUs, AWS Graviton, Google Axion, Microsoft Cobalt, automotive systems, embedded devices, and a growing share of AI infrastructure.

The simplest way to understand Arm is this: Arm does not need to own every chip market directly. It wants its architecture to become the default foundation on which the next generation of computing is built.

TAM -How Big Is The Opportunity?

Arm’s TAM has three layers.

• Existing IP TAM This is the traditional Arm royalty and licensing business across smartphones, consumer, automotive, IoT, PCs, and infrastructure. Reuters reported that Arm expects its IP business to roughly double over about five years, implying roughly $10B of annual IP revenue if management’s long-term framework plays out.
• AI Data-Center CPU TAM This is the new AGI CPU opportunity. Arm is targeting roughly $15B of annual AGI CPU revenue in about five years. That alone would be larger than the entire current company.
• Broader server CPU / AI infrastructure TAM MarketWatch cited analyst expectations for the server CPU market to reach roughly $170B by 2030, with Arm positioned to benefit from AI-driven data-center demand and energy-efficiency needs.

So the simple TAM framing is:

• Current Arm revenue base: about $4.9B
• Management’s FY2031 revenue ambition: about $25B
• Implied revenue expansion: roughly 5x
• Primary TAM unlock: AGI CPU + data-center infrastructure

This is why the market is so excited. Arm is not just growing inside existing categories; it is trying to move from royalty participation to product-level economics in AI compute.

How They Win

Arm wins through four structural advantages.

• First, ecosystem lock-in. Once developers, operating systems, compilers, and chip designers optimize around an architecture, switching becomes hard. Arm’s dominance in smartphones proves how powerful that network effect can become.
• Second, energy efficiency. Power is one of the biggest bottlenecks in AI data centers. Arm’s designs are valued because they offer strong performance per watt, which matters when hyperscalers are fighting for power capacity, cooling, and data-center economics. Reuters noted that Arm’s designs are prized for power efficiency, especially as data-center operators manage energy costs and heat from AI workloads.
• Third, royalty uplift. Arm benefits when customers migrate to newer architecture generations and higher-value licenses. The move toward Armv9, Neoverse, and Compute Subsystems can increase royalty dollars per chip even if end-market units are not exploding.
• Fourth, direct silicon optionality. The AGI CPU gives Arm a new revenue stream that is potentially much larger per unit than a royalty. That is the upside case, but also the source of controversy.

Business Model

Historically, Arm makes money in two ways:

• Licensing revenue: customers pay upfront for access to Arm IP.
• Royalty revenue: Arm earns money when chips using its technology ship.

This is a beautiful model because Arm does not carry the same manufacturing risk as traditional semiconductor companies. Gross margins are structurally high because the company sells design IP, not physical inventory.

The AGI CPU changes that. Arm is now entering a business that may offer far more revenue per unit, but also comes with supply-chain, manufacturing, customer-conflict, and regulatory risk. Arm says TSMC is fabricating the AGI CPU on 3nm, and it is working with server makers such as Lenovo and Quanta to offer complete systems.

That is why the company is becoming more exciting and less simple.

By The Numbers

The latest reported figures show a business still growing strongly:

• Q4 FY2026 revenue: $1.49B, up 20% YoY
• Q4 FY2026 adjusted EPS: $0.60
• Q1 FY2027 revenue guide: about $1.26B
• AGI CPU demand through FY2027–FY2028: over $2B
• AGI CPU long-term revenue ambition: roughly $15B annually
• Total long-term revenue ambition: roughly $25B annually
• Long-term EPS ambition: about $9

The issue is not operating quality. The issue is that the market is capitalizing a large part of the FY2031 dream today.

Key Drivers

• The first driver is AGI CPU execution. This is the single most important catalyst. If Arm ships successfully, secures enough supply, and converts early customer demand into real revenue, the stock can continue to command a premium.
• The second driver is data-center royalty growth. Even before AGI CPU scales, Neoverse-based designs from Nvidia, AWS, Google, Microsoft, and others can keep lifting royalty revenue.
• The third driver is royalty-rate expansion. Armv9, CSS, and higher-value architecture deals can drive revenue per chip higher even if smartphone unit growth is mediocre.
• The fourth driver is AI at the edge. Smartphones, PCs, cars, robots, and IoT devices all need more on-device compute. Arm is already embedded in many of those categories, so AI can lift content per device.
• The fifth driver is TAM expansion from IP to silicon. This is the big one. If AGI CPU works, Arm is no longer only collecting a small royalty on AI infrastructure. It becomes a direct product vendor in one of the fastest-growing markets in semiconductors.

Risks

• The biggest risk is valuation. Arm trades like a company that will almost certainly execute its long-term plan. If growth slows, the AGI CPU is delayed, or regulatory risk increases, the multiple can compress sharply.
• The second risk is customer conflict. Arm’s neutrality has always been part of its moat. By launching its own data-center CPU, Arm risks competing with customers and licensees that helped build the ecosystem.
• The third risk is FTC / antitrust scrutiny. Tom’s Hardware reported that the U.S. FTC is investigating whether Arm is leveraging its architecture position to advantage its own AGI CPU and potentially limit access or degrade designs for rivals. That may not destroy the thesis, but it can create a long regulatory overhang.
• The fourth risk is supply constraints. Arm already said demand for its new chip is stronger than the supply it has secured. That means the near-term revenue ramp may be limited by manufacturing capacity, packaging, test equipment, and component availability.
• The fifth risk is RISC-V and architecture competition. RISC-V is still far behind Arm in ecosystem maturity, but it is open-source and strategically attractive to some hyperscalers and governments. If RISC-V gains real server traction, Arm’s long-term royalty power could be challenged.
• The sixth risk is SoftBank control. SoftBank remains the dominant owner. That creates limited float, governance concerns, and potential selling overhang if SoftBank monetizes part of its stake.

Bottom Line

Bull case: Arm is one of the most important companies in the AI compute stack. It already dominates smartphones, is gaining share in data centers, benefits from royalty-rate expansion, and now has a potentially transformational AGI CPU opportunity. If management’s FY2031 framework proves even broadly accurate, Arm can grow from a ~$5B revenue company into a ~$25B platform business.

Bear case: the stock has already priced in a lot of that future. The AGI CPU is unproven at commercial scale, regulatory risk is real, SoftBank control remains a governance overhang, and customers may become more cautious if Arm’s neutrality weakens.

Investment conclusion: I would frame Arm as a high-quality hold and speculative buy only on meaningful pullbacks. It is one of the best semiconductor businesses in the world, but the current valuation requires nearly flawless execution. The business is exceptional. The entry point is the debate.

Fair Value

Current Fair Value: $340 per share. My optimistic fair value estimate for Arm is $430 per share. (This said, you must believe in Arm’s 2031 plan, and you are willing to pay for that future almost fully today.)

This is not a conservative valuation. It assumes:

• AGI CPU launches successfully,
• supply constraints are solved,
• early hyperscaler demand converts into real revenue,
• IP revenue continues compounding toward management’s long-term target,
• and the market remains willing to pay a premium for Arm as a strategic AI architecture platform.

Why $430? Because if Arm can credibly earn around $9 of EPS by FY2031, a premium long-term multiple of roughly 45–50x would support a future value in the $405–450 range. That is an optimistic framework, not a margin-of-safety framework.

So the clean takeaway is:

Current setup: exceptional company, stretched valuation

Optimistic fair value: ~$430/share

Best entry logic: buy only if you believe AGI CPU materially expands the TAM and Arm keeps its ecosystem intact despite direct silicon and regulatory risk.

Slidedeck: https://cdn.prod.nntech.io/company-events/reports/1a5470d7-8959-3458-be22-008846f304b7/presentation.pdf