--- title: "ARM (Trans): Core biz red-hot; new biz demand surging" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/43054762.md" description: "Below is Dolphin Research's compiled Trans of $Arm(ARM.US) FY27Q1 earnings call. Part I: Core financial highlights review.1) Results recap: FY27Q1 revenue was $1.29bn (+22% YoY). It was the highest quarterly revenue in the company's history.Royalty revenue was $715mn (+22% YoY), while licensing and other came in at $574mn (+23% YoY). Both set Q1 records. Non-GAAP EPS was $0.45 (+29% YoY)..." datetime: "2026-07-29T23:49:15.000Z" locales: - [en](https://longbridge.com/en/topics/43054762.md) - [zh-CN](https://longbridge.com/zh-CN/topics/43054762.md) - [zh-HK](https://longbridge.com/zh-HK/topics/43054762.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- # ARM (Trans): Core biz red-hot; new biz demand surging **Below is Dolphin Research's compiled transcript of** $Arm(ARM.US) **FY27Q1 earnings call** **I. Key takeaways from the print** 1\. **Results**: FY27Q1 revenue was $1.29bn (+22% YoY), a record high for any quarter. Royalty revenue reached $715mn (+22% YoY) and license & other was $574mn (+23% YoY), both a Q1 record; Non-GAAP EPS was $0.45 (+29% YoY), above the high end of guidance. Cloud AI remained the biggest driver for royalties, with data center royalties again more than doubling YoY. 2\. **Margins and FCF**: Non-GAAP Opex was $733mn (+18% YoY), driven by ongoing R&D, about $27mn below guidance due to timing, with full-year spend plans broadly unchanged. Non-GAAP OP was $531mn, implying ~41% OPM (+200bps YoY). FCF was $665mn for the quarter and $1.4bn over the last 12 months, providing ample flexibility for long-term investment. 3\. **Guidance and royalty outlook**: FY27Q2 revenue guidance is $1.38bn ± $50mn, midpoint +22% YoY; license & other is guided ~+30% YoY, while royalties are guided low-teens YoY (+13% on guide basis). Non-GAAP Opex is ~$780mn; Non-GAAP EPS $0.47 ± $0.04. Hurt by mobile memory price hikes, full-year royalty growth is trimmed from ~20% to high-teens, but stronger licensing should offset, making the combined outlook better than initially expected with no overall slowdown; 20%+ royalty growth view for FY28–FY29 is unchanged. 4\. **License mix and ACV**: Of the $574mn license revenue, $193mn came from SoftBank tech license and design services, with a run-rate of ~$200mn per quarter for the rest of the year. Given the lumpiness of high-value deals, annual contract value (ACV) is the core trend metric; ACV rose 13% YoY this quarter, still above Arm’s long-term license growth expectation. This supports a healthy licensing trajectory despite deal timing variability. 5\. **Arm AGI CPU reporting**: Manufacturing capacity to support a $1bn opportunity (FY27–FY28) has been secured, with demand exceeding $2bn. Confidence in surpassing $1bn has increased over the past 90 days, with more detailed revenue and margin updates to come with FY27Q3 results. First-gen GPM remains high-30s to low-40s, with a 50% target over the next two years by insourcing previously external work. Silicon revenue will be disclosed as a third line item, separate from licenses and royalties, once it reaches at least 10% of total revenue, expected from FY28. **II. Details from the call** **2.1 Management highlights** **1) Arm AGI CPU (new in-house silicon)** a. Launched in Mar, offering another route for customers to deploy the Arm compute platform, with strong progress since: first units have shipped to multiple customers. This marks the transition from design wins to initial deliveries. b. Capacity to support a $1bn opportunity across FY27–FY28 is secured; with customer additions in the U.S. and China, demand has exceeded $2bn and the overall pipeline continues to strengthen. Customer mix and commercial terms are being optimized alongside supply expansion. c. Arm is working closely with manufacturing and supply chain partners to expand capacity; beyond securing incremental supply, it is refining customer mix and terms. Confidence in delivering above $1bn has improved over the last 90 days and will be updated with FY27Q3, when visibility into FY27Q4 and FY28 is better. Further detail on revenue and margins will follow. **2) Data center and Neoverse (Cloud AI)** a. Data center royalties more than doubled YoY again, and Arm is gaining share as the CPU foundation for AI infrastructure. Drivers are twofold: hyperscalers’ Arm-based server CPUs continue to ramp, and data center networking silicon deployments are rising, especially DPUs and SmartNICs, where nearly all mainstream products use Arm. This underscores Arm’s expanding footprint across compute and networking. b. Cumulative Neoverse shipments have surpassed 1.5bn cores; the most recent 500mn took only nine months vs. six years for the first 1bn. Adoption is accelerating materially. c. Customer updates: NVIDIA’s Vera has entered mass production on Arm, delivering up to 50% higher CPU performance vs. comparable x86 and 2x energy efficiency, serving as the CPU foundation of its next-gen AI infrastructure; Google said Arm-based Axion is a core component of its AI infrastructure strategy and the host CPU for its latest TPU systems; AWS plans to deploy tens of millions of Graviton5 cores for agentic AI workloads; Microsoft expanded Azure Cobalt 200 VMs based on Arm Neoverse CSS; Qualcomm announced entry into AI data center CPUs with the Arm-based Dragonfly C1000. Paths differ, but the direction is the same: Arm-based CPUs are becoming central to next-gen AI infrastructure. d. IDC data show spend on Arm-based accelerated server platforms nearly doubled over the past two quarters, surpassing x86 platforms. This reflects rapid mix shift in AI-oriented deployments. **3) Edge AI and personal computing** a. Smartphone end-demand is soft, but edge AI royalties continue to grow. Rising penetration of Armv9 and compute subsystems (CSS) in phones, tablets, and other consumer devices lifts royalty rates, offsetting unit pressure from memory-driven price hikes. Mobile structural mix gains are mitigating cyclical headwinds. b. As AI scales, customers focus more on the economics of broad deployment. Whether AI runs in the cloud, at the edge, or in the physical world, efficient compute is as important as model capability; moving AI closer to where data is generated improves performance, reduces latency, enhances privacy, and lowers infra cost, which is Arm’s core differentiation. This underpins Arm’s strategy across cloud-to-edge. c. Two device classes are emerging: power-efficient AI PCs for mobile scenarios, and higher-performance agentic systems that run complex models locally. NVIDIA’s RTX Spark is the first agentic PC based on an Arm compute subsystem, capable of running complex AI agents and larger models on-device; on mobile AI PCs, OEMs are expanding the Windows on Arm ecosystem with new Snapdragon models, and Google’s continued investment in AI Chromebooks broadens on-device AI coverage. Arm is positioned across both segments. **4) Physical AI** a. Autos, robotics, industrial systems, and autonomous machines increasingly rely on efficient, safe, real-time Arm compute for perception, inference, and actuation. Arm is embedded across the stack. b. ADAS and autonomy expansion contributed meaningfully to royalties this quarter. Auto remains a multi-year driver. c. NVIDIA expanded its Physical AI stack with Cosmos 3 and the Isaac GR00T humanoid platform, powered by Jetson Thor combining Arm CPUs with NVIDIA Blackwell GPUs. This highlights hybrid CPU-GPU architectures. **5) Software ecosystem and developers** a. Supported developers globally have topped 22mn. The ecosystem continues to scale. b. Performix launched this quarter with support from Microsoft, MongoDB, Redis, and SAP, helping developers and AI agents analyze and optimize workloads on Arm infrastructure. Tooling is focused on performance and efficiency. c. AI dev tools keep expanding; Docker downloads for Arm MCP server have exceeded 10k, and it has integrated with mainstream AI dev environments. Developer adoption is broadening. **6) Licensing and customer relationships** a. License growth is driven by strong demand for next-gen architectures and deeper strategic collaborations with key customers. This includes multi-year, roadmap-level engagements. b. Multiple high-value agreements were signed this quarter; existing customers renewed long-term licenses, including several of the world’s largest hyperscalers, auto and robotics companies, and mobile OEMs, securing access to Arm’s future roadmap. Visibility on long-term adoption continues to improve. **7) Platform strategy and overall view** a. Customers can deploy Arm as IP, as a compute subsystem, or as silicon, but all rest on the same compute platform, software ecosystem, and developer community. As AI permeates every cloud, device, and industry, the market is converging on a common compute platform, which will shape the next decade of computing. Arm aims to be that converged foundation. b. This quarter centers on two themes: core IP performed very strongly with record Q1 results and better-than-expected guidance, driven by data center with rising share and multiple key customers now shipping Neoverse cores (cumulative 1.5bn+); and growing confidence in the Arm AGI CPU business launched in Mar, with demand now above the level disclosed 90 days ago and rising confidence in meeting that demand. Both engines are powering growth. **2.2 Q&A** **Q: What strengthened confidence in exceeding $1bn for Arm AGI CPU — better wafer availability, customers securing memory, or other factors?** A: It is a combination of improvements across the board. Ninety days ago we had supply secured for a $1bn opportunity and a $2bn demand pipeline, and we were working on the gap between $1bn and $2bn. Since then, demand has improved further to above $2bn, and our confidence in securing supply beyond $1bn has risen across wafers, substrates, test capacity, and memory. **Q: For AGI revenue above the initial $1bn, has your gross margin view changed vs. the original plan?** A: No change at this time. Last quarter we said first-gen GPM would be in the high-30s, possibly low-40s, for FY27 Q4 and next fiscal year. We still expect to reach ~50% over the next two years by insourcing work previously done externally, which takes about two years; we will update revenue and margin details with Q3 results, factoring in recent cost increases and pricing where appropriate. **Q: Memory inflation is pushing handset OEMs to raise prices and absorb more BOM pressure. Given mobile royalties remain a key base, what are you seeing on units and mix, and how does this affect the royalty outlook?** A: With most customers migrating to Armv9, some to CSS and some to CSS Gen2, we are partially insulated from the handset market’s decline, as these transitions lift royalties. Thus, even as smartphones are expected to fall double-digits, we still expect mobile royalties to grow double-digits. For forecasting, we use broad industry sources such as IDC and adjust by customer mix — royalty rates differ across CSS, Armv9, and Armv8 adopters. Compared with early-year expectations, there is some marginal slowdown, now impacting all price tiers rather than just low-to-mid, with mid and even parts of upper-mid affected. Accordingly, our multi-year royalty growth view including this year moves from ~20% to high-teens, with Q2 guided to low-to-mid-teens; Cloud AI’s outperformance continues to offset mobile pressure, supporting our confidence this year and next. **Q: How does the demand pipeline look for FY29–FY31, and does the prior FY31 ~$15bn AGI revenue view already reflect supply constraints, i.e., supplyable revenue?** A: Since Mar, supply has indeed been a constraint in multiple links, and our outlook reflected that. We also stated a long-term CPU TAM of ~$100bn at the Mar event, above prior consensus of $50–60bn, which drew debate; since then, peers have published higher numbers, some up to ~$200bn. The driver is total compute growth, with inference — largely agentic workloads — consuming CPU-bound throughput; we are not changing numbers today, but indicators suggest our $100bn TAM may be conservative. As a supplement: our $100bn TAM was framed as ‘$100bn+’; some recent estimates are ~$220bn. Holding the same share assumption, a $200bn+ market implies much higher outcomes. The near-term constraint is tight supply through FY27–FY28 (CY2026–2027), and whether we can grow beyond $2bn will be updated in coming quarters. From CY2028–2029 (FY29–FY30), supply should ease materially, with wafer and memory capacity up an estimated 70–100% depending on partner assumptions, and even larger on memory; external numbers could be exceeded, but near term we remain focused on delivering ~$2bn next year. **Q: AMD’s Analyst Day cited a ~$220bn TAM by decade-end, led by agentic AI across traditional CPUs, head nodes, and agentic apps. Where does Arm’s silicon business fit, and are early orders skewed to any sub-segment?** A: Arm AGI CPU spans all three. Among customers announced in Mar, Cerebras and OpenAI map to head nodes; Meta and Cloudflare map to agentic apps and general servers; plus Oracle OCI. Agentic workloads are broad within the cloud — agents run on head nodes and on general server racks — so our design is well suited across categories, which underpins our optimistic demand view. **Q: Near term, if this is a $50bn market with multiple foundries available, why is securing another $1bn of supply so hard?** A: It is positive that $1bn looks small, but in reality many companies struggle to secure even $50mn of incremental supply. The entire chain is extremely tight — memory, test gear, substrates, and TSMC wafers. Even long-standing CPU vendors with internal fabs are short of demand; we are more optimistic than 90 days ago as demand is higher and our supply position has improved. **Q: Long term, NVIDIA, Amazon, and Google pair Arm CPUs with in-house accelerators. Does offering only CPUs limit Arm, and will you add beyond CPUs?** A: There are clear avenues for us to participate. At Google, Axion replaces x86 as the interface to its TPUs, proving coexistence with custom accelerators; for Arm AGI CPU specifically, NVIDIA has discussed NVLink Fusion, a platform approach enabling mix-and-match — Vera CPUs paired with custom accelerators, or other Arm CPUs paired with Rubin accelerators. Arm can connect to custom accelerators via self-hosted designs or approaches like NVLink Fusion. **Q: AGI CPUs are 128-core today; when will 192-core or 256-core next-gen products arrive, and any ASIC progress?** A: No specifics to share on the roadmap today. Arm AGI CPU is a 128-core base design; the market’s highest-core Arm product today is Graviton5 at 192 cores. Core counts will increase because agentic workloads scale better with more cores — simpler software overhead, per-core VMs or tasks, and higher throughput efficiency — so you can infer the direction even without specific disclosures. **Q: As agentic CPU revenue scales, will you break it out separately?** A: The first milestone is shipping by year-end. Once this revenue reaches at least 10% of total, we will report it as a third line item — silicon revenue, separate from licenses and royalties — expected from FY28. This will enhance transparency. **Q: With royalty growth likely shifting from ~20% to high-teens due to mobile, can licensing upside offset, and does 20% royalty growth still hold for FY29?** A: Q2 royalty growth guidance of 13% is below our internal view from two quarters ago, but we raised license growth, so the combined outlook is actually better than before. For the rest of the year, any softness in royalties should be at least offset by licensing. Overall we do not expect a slowdown. One swing factor is Cloud AI continuing to outperform. We see strong deployments at Google Axion with all new TPU rollouts, Amazon Graviton, NVIDIA Vera, and other hyperscalers accelerating beyond plan; this creates upside risk to royalties this year. Net-net, the only weakness is on mobile; whether Cloud AI fully or partially offsets will be updated through the year. Our 20%+ royalty growth view for FY28–FY29 stands. **Q: Does the spread of low-cost AI (from China or elsewhere) help or hurt AGI CPUs, and what are the implications for IP in data center vs. edge?** A: Interpreting low-cost AI as open models and open weights, the impact is neutral to positive. Neutral because both frontier closed and open models need CPUs and Arm IP for custom silicon, regardless of whether customers choose AGI CPUs, IP, or CSS — agents need CPUs either way. On the positive side, open models may become more differentiated, smaller, and more efficient, enabling more edge deployment; given Arm’s edge footprint, that would be beneficial. Open ecosystems typically expand optionality and innovation, so depending on edge evolution, the outcome ranges from neutral to positive while Arm remains model-agnostic. **Q: Why was R&D lower than expected — milestone slippage, capitalization, or timing? What drives the P&L timing of R&D?** A: Opex was slightly lower this quarter. R&D has two parts: people and development costs, and tools such as emulation and cloud spend. The undershoot was not headcount-related — that was in line — but due to lower-than-forecast tool usage. Treat Q1 savings as flow-through; going forward, estimates are broadly consistent, perhaps slightly lower, with overall Opex growing low single-digit QoQ, and volatility driven more by tool usage than headcount. Innovation and cloud spend flex with development stages, depending on test needs and completion of work. Forecasting has improved but will still see small swings as we have only scaled this for a year or two; the band of variance should narrow over time. Discipline remains on prioritization and efficiency. **Q: Will order visibility for AGI CPUs meaningfully improve only by early CY2027 (Dec quarter results), and is there a ‘golden screw’ risk where a small missing part blocks the whole program despite 99% secured?** A: We are not concerned about a last-minute single-point failure. We have strong confidence in our partners, their capabilities, and their ability to scale. There is no 1% missing piece that jeopardizes the plan. **Risk disclosure and statements:**[**Dolphin Research disclaimer and general disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [ARM.US](https://longbridge.com/en/quote/ARM.US.md)