Financial disclaimer: This article is for informational purposes only and is not investment advice. Stock prices and guidance can change quickly; investors should review company filings and consult a qualified adviser before making decisions.
Arm did not lose investors because the quarter was weak. In its May 6, 2026 shareholder letter, Arm reported record Q4 FY26 revenue of $1.49 billion, licensing revenue of $819 million, royalty revenue of $671 million and non-GAAP EPS of $0.60. The selloff came from a different problem: Reuters reported that ARM shares reversed after executives discussed supply not yet being secured for all demand tied to the new Arm AGI CPU.
That makes the next ARM stock debate cleaner than the headline move suggests. The question is not whether Arm can attach itself to AI infrastructure. It already has. The question is whether Arm can turn agentic-AI CPU demand into real silicon revenue without weakening the royalty model that made the company valuable in the first place.
The line that changes the ARM stock debate
The most important number in Arm's earnings materials was not the EPS beat. Arm said data centers may need more than four times current CPU capacity per gigawatt as agentic AI scales, creating a CPU market opportunity of more than $100 billion by 2030. That is a different framing from the usual AI-chip story, because it puts the CPU back at the center of power, density and workload coordination.
Agentic AI is not just a larger chatbot query load. Arm argues that continuous agent-driven workloads require CPUs to coordinate tasks, move data, manage memory, enforce security and orchestrate work around accelerators, all of which are spelled out in the company's May 2026 shareholder letter. If that view is right, the AI data center does not only need more GPUs. It needs a stronger host CPU layer that can keep accelerators, memory, networking and software agents synchronized.
That is where Arm's pitch becomes investable rather than just technical. TECHi previously covered Arm's AI royalty story through Amazon Graviton. The new test is whether the same architecture can become a larger economic claim on AI data-center capacity.
Why the AGI CPU matters more after earnings
Arm describes the Arm AGI CPU as its first production silicon product for AI infrastructure at scale, built on Neoverse CSS V3. The product page lists up to 136 Neoverse V3 cores, a TSMC 3 nm process, 300W TDP, up to DDR5-8800 support and 6 GB/s of memory bandwidth per core, which makes the product a concrete CPU platform rather than a generic licensing announcement.
The ecosystem evidence also moved beyond one hyperscaler. Arm's AGI CPU partner page lists Meta, OpenAI, Cloudflare, SAP, SK Telecom, Cerebras, Lenovo, Microsoft Azure, Nvidia, TSMC and others in the ecosystem around the product, while Meta says it worked with Arm on a multi-generation roadmap for AI infrastructure. Siemens separately said its hardware-assisted verification tools were used to support full-system verification of the Arm AGI CPU for hyperscale deployment targets.
For investors comparing AI infrastructure names, this puts Arm in a different lane from AMD's accelerator platform test or Nvidia's AI network and accelerator cycle. Arm is trying to monetize the CPU coordination layer that sits beside those accelerators. That is not as dramatic as a GPU backlog, but it can be more durable if every new AI rack needs more efficient general-purpose compute.
Q4 gave Arm room to tell the story
The operating numbers gave Arm credibility. For fiscal 2026, revenue reached $4.92 billion, royalty revenue rose 21% to $2.61 billion, licensing revenue rose 25% to $2.31 billion and non-GAAP EPS reached $1.77. Arm also said data-center royalty revenue more than doubled year over year, which matters because royalties are the part of the model investors usually treat as higher quality than one-time chip sales.
The Q1 FY27 guide was not weak either. Reuters reported that Arm guided for about $1.26 billion of Q1 revenue versus LSEG consensus of about $1.25 billion. Licensing can still be lumpy, but management is telling investors that royalty growth and AI infrastructure demand remain intact after a record fiscal year.
That is why the stock reaction is useful. Google's TPU push and the broader AI stocks trade have trained investors to reward visible accelerator demand. ARM stock is now being judged on something harder: whether a new AI CPU product can create silicon economics without making the company's royalty business look less clean.
The supply issue is the real bear case
Arm said it had more than $2 billion of customer demand across fiscal 2027 and fiscal 2028 for the Arm AGI CPU, more than double what it stated at launch. The problem is that demand is not revenue until supply is locked, parts are validated, racks are deployed and customers scale workloads.
That is why the stock fell after a strong print. Investing.com reported that the drop followed comments indicating Arm had not secured the supply needed to meet an additional $1 billion of AGI CPU demand. The same report said management maintained a revenue forecast while it works through supply capacity, which makes wafer access, memory, packaging and launch timing the next gating factors.
Analysts are already splitting around that tension. Needham raised its ARM price target to $255 and kept a Buy rating after earnings, citing upside to Arm's CPU revenue forecast, according to StreetInsider's May 7 analyst-target report. That bullish read is plausible only if demand turns into supplied product on a schedule investors can trust.
What investors should watch next
First, watch supply conversion. The demand headline is large, but the stock needs evidence that Arm can secure the wafers, advanced memory, packaging capacity and server partner ramp required to ship meaningful AGI CPU volume. Without that, the AGI CPU remains a valuation argument rather than a revenue line.
Second, watch royalty quality. Arm's full-year royalty growth and more-than-doubled data-center royalty revenue show that the older model is not broken, based on the company's shareholder letter. The best ARM stock setup is not silicon replacing royalties. It is silicon expanding Arm's data-center share while IP and CSS continue to compound.
Third, watch whether customers treat Arm as neutral infrastructure or as a supplier that now competes with parts of its own ecosystem. Arm's 2025 Form 20-F warned that moving into compute subsystems, chiplets and complete end-chip solutions could add competitive, technological, regulatory and financial risks, and it also noted that top-five customers including Arm China accounted for 56% of fiscal 2025 revenue. That is the risk investors should keep attached to any AI-stock rerating.
Bottom line for ARM stock
The new ARM stock test is not simply whether AI data centers need more chips; Arm has already argued CPU capacity requirements rise sharply as agentic workloads scale. The test is whether Arm captures a larger share of AI rack economics as CPUs become the orchestration layer for agentic workloads. Arm has a strong quarter, a clear product, named partners and more than $2 billion of indicated AGI CPU demand. It also has a supply-chain proof point still in front of it.
That makes ARM a more interesting AI stock after the post-earnings selloff, but not a simpler one. The bullish case is now measurable: more data-center royalties, supplied AGI CPU demand, and no damage to Arm's neutral IP role. If those three pieces move together, the market will have to value Arm as more than a smartphone royalty company.
