The Royalty Engine Is Real, the Licence Line Is Rented: Initiating Arm at Hold

Results vs. Consensus

Arm reported its fiscal third quarter, ended December 31, 2025, after the close on February 4. It was a beat on both lines, but a narrow one on revenue: the Street sat essentially on the company's own guide midpoint, and the print cleared it by less than a percent. The earnings beat was wider and came from two places, higher revenue and operating expense that landed $4 million below plan.

Q3 FYE26 Scorecard

Year-over-Year Comparison

Sequential Comparison

The sequential table is where the quarter's real shape shows up. Royalties added $117 million quarter over quarter, a 19% step. Licensing went backwards. Operating expense grew faster than revenue for the third consecutive quarter. And remaining performance obligations, which had been flat to slightly up through the first half, turned down.

Revenue

Total revenue of $1,242 million grew 26% and marked a fourth consecutive quarter above $1 billion. Both halves of the model grew, but for opposite reasons. Royalties compounded on rate: management attributes the growth to Armv9 and compute subsystem adoption raising the royalty captured per chip, plus data center share gains, rather than to unit growth. Licensing grew on a single related-party agreement. The nine-month picture makes the point sharply. Revenue from related parties over the first nine months of FY26 was $1,088 million against $601 million in the prior-year period, growth of 81% against total company growth of 24%. Related parties now supply 32% of nine-month revenue, up from 22%.

Assessment: Two revenue lines with genuinely different quality are being reported as one growth rate. The royalty line deserves a premium multiple. The licence line, at current mix, does not.

Margins

Non-GAAP operating margin of 40.7% is down from 45.0% and, on a sequential basis, down 40 basis points from Q2 despite revenue growing 9.4%. That is the signature of an expense base growing faster than a record revenue quarter. The CFO tied it directly to engineering headcount and to the company's move up the stack, describing investment in "next-generation architectures, compute subsystems, and into our exploration into chiplets and complete SoCs." That last phrase is the most consequential five words in the release, and it went almost entirely unexamined in Q&A.

Assessment: Arm is spending like a company that intends to sell silicon, while being valued like a company that only collects rent on it. Those are different businesses with different margin structures, and the market has not yet been asked to price the second one.

Earnings and Cash

The cash statement is where this quarter is weakest and where the least call time was spent. Operating cash flow fell 14% to $365 million while non-GAAP net income rose 10%. Free cash flow fell 52% to $169 million, a conversion rate of 37% against 84% a year ago. Two items bridge it. Contract assets, revenue recognised but not yet invoiced, consumed $231 million of cash in the quarter, and related-party contract assets specifically rose to $499 million at December 31 from $152 million at March 31. Capital expenditure was $179 million against $63 million, taking property and equipment on the balance sheet to $732 million from $354 million over nine months.

Assessment: Trailing twelve-month free cash flow of $893 million is still up 56%, so this is a deterioration in conversion rather than a broken cash model. But a $347 million nine-month build in receivable-like assets owed by the controlling shareholder, arriving in the same quarter that the controlling shareholder became 40% of the licence line, is a combination that deserves more disclosure than it received.

Revenue Line Performance

Arm does not report profit by segment. It reports two revenue lines and, once a year, an end-market split. The company reorganised this quarter into three stated AI domains, Edge AI (smartphones and IoT), Physical AI (automotive and robotics) and Cloud AI (data center and networking), but did not attach revenue to them. What follows uses the disclosed revenue lines, with end-market colour drawn from management commentary.

End-market colour (management-sized on the call; not a reported segment disclosure)

Royalty

Royalties of $737 million grew 27% and set a record, with management identifying two drivers: smartphones at higher royalty rates per chip, and data center, where revenue continues to roughly double. The mechanism matters more than the number. Arm is not being paid for more chips; it is being paid more for each one, through the Armv9 architecture and through compute subsystems, where a new subsystem is delivered each smartphone cycle at a contractually higher rate.

"The biggest growth contributors were smartphones with higher royalty rates per chip, and in the data center where our revenues continue to grow triple digits year on year as we see ongoing share gains from custom hyperscaler chips."
— Jason Child, CFO

The company disclosed that Neoverse deployments have passed one billion cores and that its share among top hyperscalers is expected to reach nearly 50%. The supporting product evidence is unusually concrete for a quarter of this kind: AWS Graviton5 at 192 cores, double Graviton4; NVIDIA's Vera at 88 Arm cores against 72 in Grace; Microsoft Cobalt 200 at 132 cores on Neoverse CSS V3; and Google's second Axion server processor, with more than 30,000 applications migrated to the Arm instruction set.

Assessment: This is the part of the business that justifies a premium multiple, and it is compounding on a mechanism that does not depend on semiconductor unit cycles. The rate escalator is contractual, the core-count trend is documented in shipping silicon from four hyperscalers, and none of it requires Arm to win a new socket. The caution is that Q4 royalty growth is guided down to low teens, which management attributed to seasonality and to lapping an unusually timed prior-year MediaTek launch rather than to any change in the rate story.

Licence and other

Licence and other revenue of $505 million grew 25% year over year but fell 1.9% sequentially. Two Arm Total Access agreements and two compute subsystem licences were signed in the quarter, taking total access licences to 50 and cumulative CSS licences to 21 across 12 companies. Annualised contract value, which management prefers as the underlying licensing indicator because reported licence revenue is lumpy, grew 28% for the third consecutive quarter to $1,620 million.

The composition, however, is the disclosure that matters. Under questioning, the CFO confirmed that $200 million of the $505 million came from the SoftBank technology licensing and design services agreement, up from $178 million in Q2 purely because Q3 was the first full quarter, and that $200 million is the go-forward run rate.

"The full quarter has the impact now is about $200. So nothing new. It's just a full quarter impact. I would expect that $200 going forward is the right run rate going forward."
— Jason Child, CFO

Assessment: A quarter of the company's total revenue and 40% of its licence line now comes from an agreement with its controlling shareholder that had no equivalent a year ago. Excluding it, licence revenue declined roughly 24% year over year. Management's answer to the durability question is that the work relates to current-generation designs that continue as SoftBank executes its roadmap, which is a reasonable answer and also an unfalsifiable one from outside the company. Until Arm discloses the agreement's term, its scope, or the associated payment schedule, the market is being asked to capitalise a related-party revenue stream at the same multiple as third-party royalties.

Key Operating Metrics

Assessment: The ACV and RPO lines point in opposite directions and Arm reports both. ACV annualises committed fees across assumed contract lives and is the metric management foregrounds; RPO is the accounting measure of contracted revenue not yet recognised, and it fell 8%. The reconciliation is almost certainly that Arm has been recognising licence revenue faster than it has been signing replacement backlog, which is exactly what a large upfront related-party recognition would do to the ratio. Nobody asked about it.

Key Topics & Management Commentary

Overall Management Tone: Confident and unusually specific on the demand side, noticeably less forthcoming on composition. The prepared remarks were dense with third-party product evidence rather than adjectives, which is a stronger form of argument than this company has typically made, and the CFO volunteered a quantified downside sensitivity on the memory shortage before being asked twice. Where the call thinned out was on anything that would let an outsider decompose the revenue: data center dollars, CSS royalty dollars, the SoftBank agreement's terms, and next year's guide were all deferred, three of them to a March event and one to an annual disclosure cadence.

Data center royalties and the claim that they pass mobile

The single largest thesis claim of the call was not about this quarter. Management stated that data center royalty revenue has grown more than 100% year over year and that within a few years the data center business will be larger than mobile. Sized against the CFO's own framing, that means moving from the teens toward 20% of revenue today past a smartphone business at 40% to 45%, without smartphone shrinking.

"Our data center royalty revenue has grown more than 100% year on year, and we expect in a few years our data center business to be our largest business, larger than mobile."
— Rene Haas, CEO

The mechanism management describes is agentic inference. Because agent workloads are persistent, orchestration-heavy and power-constrained, they consume CPU cycles rather than accelerator cycles, and they scale with core count per socket rather than with socket count. The shareholder letter puts it more bluntly than the call did: "This type of work can only be done by CPUs, which is why this evolution is making the CPU more essential than ever in AI datacenters."

Assessment: This is the most credible version of the Arm data center argument we have heard, because it is grounded in shipping core counts from AWS, NVIDIA, Microsoft and Google rather than in design wins. It is also the claim with the least disclosure attached to it. Arm reports data center revenue once a year, so investors are being asked to underwrite the largest single driver of the thesis on an annual data point and a growth adjective.

Compute subsystems: from a licensing product to the royalty rate escalator

CSS has quietly become the most important product line Arm sells, because it is the vehicle through which royalty rate per chip rises without Arm winning anything new. Two more licences were signed, taking the total to 21 across 12 companies. Five customers are shipping CSS-based silicon, two of them on a second generation, and the top four Android smartphone vendors are all shipping CSS-powered devices. Asked to size it, the CFO gave the most useful number of the call.

Q: "I'm curious to know what percentage of the royalty mix is CSS today, and what proportion of the royalty revenue could it become over the next two to three years?"
— Harlan Sur, JPMorgan

A: "Think of CSS last year. I think it was just kind of approaching double digit. And this year, it well into double digit. Think of it as being into the teens. And then I would say, you know, over the next couple of years, I expect it to probably it could be upwards of 50%."
— Jason Child, CFO

The CEO added the mechanical detail that explains why this compounds: a new CSS ships each smartphone cycle, and each generation carries a higher contractual royalty rate.

Assessment: If CSS goes from the teens to roughly half of royalties over two to three years at structurally higher rates, royalty revenue can grow at a healthy clip on flat industry units. That is the strongest argument for owning this stock and it is the reason the memory-driven handset unit risk matters less than headlines suggest. The hedge in the language is real, though. "Could be upwards of 50%" was offered with an explicit "we'll have to see" attached.

The SoftBank agreement, now declared permanent

The clarification that produced the quarter's most important disclosure came as a throwaway second half of a two-part question. The CFO confirmed $200 million of the $505 million licence line was SoftBank, that the sequential increase from $178 million was a full-quarter effect rather than new business, and that $200 million is the ongoing run rate. Asked separately whether that flow is durable, management defended it on roadmap grounds.

"We believe that the revenues we are receiving from SoftBank are durable as they relate to current generations that will continue as SoftBank executes on its roadmap."
— Jason Child, CFO

What was not disclosed: the contract term, the total contract value, the payment schedule, whether the $200 million is recognised ratably or on delivery, and how it relates to the $347 million nine-month increase in related-party contract assets sitting on the balance sheet.

Assessment: The characterisation of this revenue as durable is doing enormous work. If it is durable, it is $800 million of annualised revenue at very high incremental margin and the multiple is defensible. If it is a multi-year design-services engagement tied to a specific SoftBank silicon programme, it is a project that ends. Management has told us which one it believes it is; it has provided no disclosure that would let anyone check.

The memory shortage, quantified twice with two different answers

The dominant macro question in Q&A was whether DRAM scarcity destroys handset demand and therefore Arm's royalty base. Management engaged directly and, unusually, put numbers on it, citing a peer's disclosure of an approximately 15% unit-volume reduction as the industry marker.

"So I think, if you were to say, what if there's a 20% reduction in volumes next year? For us, that would translate to probably somewhere around a 2% or 4% at worst. Impact on smartphone royalties. If you then project that across the whole business, it'd be a 1%, maybe 2% negative impact on total royalties."
— Jason Child, CFO

The logic behind the insulation is sound: constraints bite hardest at the low end, which runs Armv8 and older at materially lower royalty rates, while Arm's CSS and Armv9 economics are concentrated in premium and flagship devices that OEMs are prioritising. Two questions later, however, the same sensitivity was restated as "at most a kind of four to 6% revenue impact just specifically within smartphones", roughly double the top of the first answer.

Assessment: Even taking the wider of the two figures, a 20% industry unit decline costing Arm one to two points of total royalty growth is a genuinely defensive setup, and it is the clearest evidence available that the rate-over-units model works. The discrepancy between the two answers within a single call is not a rounding matter, and it went unchallenged.

Operating expense: R&D growing 20 points faster than revenue

Non-GAAP operating expense rose 37% against revenue up 26%, with non-GAAP R&D up 46%. Pressed on whether that persists, the CFO gave a two-part answer: the Q4-to-Q1 step-up will look like last year's low double-digit sequential increase, and growth should moderate after that.

"I can tell you right now, our expectation is that the Q4 to Q1 step up will be similar to last year. Think last year, it was you know, low double-digit sequential growth, and you should see the same kind of sequential growth as a year ago. I right now, I would say the growth after Q1 is probably gonna to moderate more so than it did this year."
— Jason Child, CFO

Assessment: This is the only forward commitment on the call that can be graded cleanly next quarter, and it is a soft one. "Moderate more so than it did this year" against a year in which non-GAAP operating expense grew 37% is a low bar. With FY27 revenue framed at roughly 20% growth, operating expense growing in the twenties would compress margin again.

Chiplets and complete SoCs, mentioned once and never revisited

Inside the explanation for elevated R&D, the CFO listed where the money is going: "next-generation architectures, compute subsystems, and into our exploration into chiplets and complete SoCs." A complete SoC is a product Arm would sell, not license, and it puts Arm into direct competition with the customers who currently pay it royalties. No analyst followed up. The topic was effectively fenced off by the CFO's pre-emptive request that questions about the March 24 event be deferred.

"There'll be a million ways of asking what we may or may not be announcing. Please be patient as we won't be providing any details ahead of the event."
— Jason Child, CFO

Assessment: The single largest unpriced variable in this story is whether Arm becomes a chip company. It would change the gross margin structure, the capital intensity, which the 184% capex increase may already be signalling, and the customer relationship. Management has confirmed it is spending on the exploration and has explicitly declined to discuss it until March.

Physical AI: Rivian, Optimus and the first Armv9 production car

The physical AI evidence was more concrete than in prior quarters. Rivian announced a third-generation Autonomy Computer built on a custom Arm-based processor, described as its first production vehicle on a custom Arm chip and the first to deploy Armv9 in a production car. Tesla's forthcoming Optimus humanoid robot is powered by a custom Arm-based AI processor. NVIDIA's Jetson Thor and Qualcomm's Dragonwing platforms carry Arm into the broader robotics market. Automotive royalties grew double digits.

Assessment: Real, early and not yet financially material. Automotive is the only one of these disclosed as a growth contributor and it grew double digits against total royalties up 27%, which means it diluted the growth rate rather than driving it. The value here is optionality on a long-dated attach rate, not near-term revenue.

Remaining performance obligations fell while ACV rose

RPO of $2,148 million was down 8% year over year and 4.4% sequentially, the first meaningful decline of the fiscal year after $2,232 million in Q1 and $2,246 million in Q2. Over the same period ACV rose 28%. Management discussed ACV at length on the call as the preferred indicator of licensing health and did not mention RPO once.

Assessment: These two metrics can diverge legitimately, because ACV annualises committed fees over assumed contract lives while RPO measures contracted revenue not yet recognised. Recognising a large licence upfront raises revenue and lowers RPO while leaving ACV untouched. That is consistent with what appears to have happened. It is also the mechanism by which a licensing business can print growth while its contracted backlog shrinks, and it deserved a question.

Free cash flow conversion and the contract-asset build

Free cash flow of $169 million against non-GAAP net income of $457 million is 37% conversion, versus 84% a year ago. The drivers are a $231 million contract-asset consumption in the quarter and capital expenditure of $179 million against $63 million. Related-party contract assets specifically stand at $499 million against $152 million at the March 31 fiscal year end. Trailing twelve-month free cash flow of $893 million is nonetheless up 56%.

Assessment: Nothing here is alarming in isolation, and the TTM figure argues this is a timing effect rather than a deterioration in economics. But the composition is uncomfortable: the quarter in which the controlling shareholder became 40% of the licence line is also the quarter in which unbilled receivables from related parties tripled since year-end. Neither the letter nor the call connected those facts.

SoftBank's stake and the overhang question

Asked whether SoftBank might need to sell Arm shares to fund its AI investments, the CEO gave the most emphatic answer of the call, citing direct conversations with SoftBank's founder.

"I can tell you from talking to Masa about this, and I would quote him directly, he is not interested in selling one share of ... stock. And that doesn't mean two shares or three shares. That means any shares. He's very long on the company. He's very, very bullish."
— Rene Haas, CEO

Assessment: As categorical as a CEO can be about a shareholder's intentions, and it plainly helped sentiment. It is also a statement of present intent by a third party that Arm does not control, offered without any standstill or lock-up to give it force. Treat it as a removed near-term catalyst, not as a structural change to the float.

The FY27 frame and the FY28 non-answer

Management declined to guide FY27 but let the existing framework stand. Asked whether the roughly 20% growth rate previously signalled for FY27 remains reasonable, the CFO endorsed it. Asked to extend to FY28, he declined and pointed at optionality.

"For '27, not guiding on full year, in terms of kind of at a high level, the 20% growth rate I think, certainly is very reasonable. And not anything that we back away from. In terms of '28, we haven't thrown anything out there yet. I'd say maybe stay tuned. You know, there are opportunities as we contemplate, you know, other possible offerings and what that could do to our numbers is still something we're working through."
— Jason Child, CFO

Assessment: The "other possible offerings" line is the chiplet and SoC thread surfacing again, this time attached to a revenue line. The FY27 endorsement is genuinely useful and de-risks the near term. The FY28 deferral tells you the company expects its own revenue model to look different by then.

Guidance & Outlook

The Q4 guide is a big number and its internal construction is more demanding than the headline suggests. Revenue of $1,470 million at the midpoint requires an 18.4% sequential increase. Operating expense is guided up only 4.1%, to roughly $745 million. Hold gross margin at the Q3 non-GAAP level of 98.3% and the guide implies non-GAAP operating income near $700 million, an operating margin of roughly 47.6%. That is about seven points above the 40.7% just delivered and above any quarter in the fiscal year to date.

Implied quarter-over-quarter ramp: With royalties guided to low-teens year-over-year growth, decelerating from 27%, virtually all of the $228 million sequential revenue increase has to come from licensing. Hold royalties flat sequentially at $737 million and the licence line has to reach roughly $733 million, a 45% sequential increase in one quarter. Management separately describes the January-to-March quarter as seasonally one of its slower royalty periods, which would push the required licence figure higher still. Licensing is lumpy by construction, and management says so every quarter, so this is not implausible. It does mean the guide rests on deal timing rather than on the royalty engine that carried Q3.

Street at: Consensus was sitting on the company's Q3 guide midpoint before the print, which is the normal pattern here. The Q4 guide came in above where the Street had been modelling and is the reason the initial after-hours reaction did not hold.

Guidance style: Arm guides tightly and beats narrowly. Q3 revenue landed 1.4% above the midpoint and non-GAAP EPS $0.02 above it, which is close to the pattern of the two prior quarters. The full year is now tracking to roughly 22% growth against an "at least 20%" framework set a year ago, so the company has beaten its own annual frame by about two points. Nothing in the Q4 construction looks sandbagged, which is itself worth noting: the sequential margin expansion embedded in the guide leaves less room than usual.

Analyst Q&A Highlights

Whether the memory shortage destroys handset demand

The dominant line of questioning on the call, raised at the top of Q&A and returned to three separate times. The framing throughout was that DRAM scarcity forces handset OEMs to cut build plans, and that Arm's royalty base moves with units. Management's answer was that it moves with royalty-weighted units, and that the units at risk carry the lowest rates.

Q: "I know you're not giving fiscal '27 commentary today, but just how do we think about the puts and takes of this royalty revenue growth and the risks that are associated with the potential like demand destruction that we're seeing, you know, in consumer electronics potentially from memory?"
— Joe Quatrochi, Wells Fargo

A: "What's interesting is we're hearing from our various partners that they're really trying to make sure that they protect the high end of the market, so the premium and flagship portion of the market. Which is great for us because that's where all of our CSS and v9 royalties are. So the highest, by a significant margin. And then on the very bottom end of the segment, that's where most of the supply chain constraints will probably be felt. For us, that's V8 and even older generations that are dramatically smaller royalties."
— Jason Child, CFO

Assessment: The strongest answer of the call, because it converts a macro fear into a mix question that favours Arm and then sizes it. It is also the answer that was restated inconsistently later, at roughly double the impact. Management engaged rather than deflected, which is the right instinct, but the two figures cannot both be the sensitivity.

Why Q4 royalty growth decelerates to low teens

Having just beaten a 20% royalty plan with 27%, the guide down to low teens invited the obvious question of whether something had changed in the demand picture. Management attributed it to comparison effects rather than to the memory constraint, isolating an unusually timed prior-year product launch.

Q: "You've provided a forecast for some deceleration in the royalty revenue growth. I'm just wondering if you could elaborate on the trend. Is it more difficult comps? Or is there something else shifting that we should be considering?"
— Simon Leopold, Raymond James

A: "Absolute growth in royalties has much more to do with typically seasonality, our Q4 or calendar Q1 is always one of the slower quarters. And the one thing that happened a year ago is we did have Mediatek chip come out in Q4 of a year ago ... Which was unusual timing. So we are lapping that."
— Jason Child, CFO

Assessment: A specific, checkable explanation rather than a hand-wave, and it is consistent with the separate statement that full-year royalties should still land north of 20%. The deceleration is arithmetic, not deterioration. Worth grading next quarter.

Sizing the data center business

A recurring frustration in Q&A: management describes data center as the fastest-growing part of the company and the future largest business, but discloses the dollars only once a year. The question was asked directly and answered with a range rather than a figure.

Q: "I was hoping you could quantify the exact amount of data center revenue. I know you said that it doubled, but how much is it so we can get a sense for, like, what the magnitude is versus the overall company sales."
— Vivek Arya, Bank of America

A: "Data center revenue we provide the details on that once a year. I think at the beginning of this year, we said it had hit double digit. And because it's growing so much faster than the rest, assume it's gonna be, you know, somewhere in kind of the teens to probably getting closer to 20%."
— Jason Child, CFO

Assessment: The declined disclosure is the point. The company is asking the market to underwrite a business it will describe but not size, at a moment when that business is the primary justification for the multiple. The range given is useful and probably honest; it is not a substitute for a reported number.

The SoftBank contribution and its run rate

The most consequential number to emerge from Q&A arrived as a clarification appended to a question about something else. The exchange established the $200 million figure, that the sequential rise from $178 million was mechanical rather than new business, and that it should be modelled forward indefinitely.

Q: "The other clarification I had was, I think you mentioned software contributed $200 million. I somehow recall the original expectation was about $178, $180 million. And if you could clarify that and what what are you embedding for March and onwards, from that contribution?"
— Vivek Arya, Bank of America

A: "The $178 last quarter, it was it was so no new deals were signed. It's just the deals from last quarter. It was $178 for the quarter. The full quarter has the impact now is about $200. So nothing new. It's just a full quarter impact. I would expect that $200 going forward is the right run rate going forward."
— Jason Child, CFO

Assessment: A clean, direct answer to a question that should have generated ten more. Nobody asked for the contract term, the total value, the recognition pattern, or how it relates to the tripling of related-party contract assets on the balance sheet. The single most important composition fact about this quarter was disclosed and then dropped.

Whether CSS pricing pressures customer bills of materials

With component costs rising across the handset supply chain, the question was whether Arm's higher CSS royalty rates become a target for cost engineering. Management rejected the premise on time-to-market grounds, arguing the value of a shortened design cycle dominates the royalty delta.

Q: "Given the pressure on the bill of materials do you anticipate or are you seeing any impact in terms of the adoption of CSS and V9 I guess, look into the next few quarters given the bill of materials challenges?"
— Srini Pajjuri, RBC

A: "No. We're not seeing any of that at all. What we are seeing is that the value gained by accelerating time to market outweighs anything that customers are considering. Given the complexity of building these chips. The increased cycle times, through the fabs going from five nanometer three nanometer to two nanometer means that the design windows are really short and missing the first few months of shipment or having any kind of delay would be critical to profits."
— Rene Haas, CEO

Assessment: The pricing-power answer investors want, and it is corroborated by the separate disclosure that every CSS customer eligible to renew for the next generation has done so. The caveat is that node-migration complexity is a cyclical tailwind to this argument, not a permanent one. If design windows ever lengthen, the value proposition compresses.

Whether R&D keeps outgrowing revenue into FY27

The only sustained pushback on the cost base. With non-GAAP operating expense up 37% against revenue up 26%, and FY27 revenue framed at roughly 20%, the question was whether the investment cycle continues to consume margin.

Q: "Utility R&D has been growing at a faster rate than revenue in the current period. Is that something we can expect to continue into fiscal '27 given everything that you guys have got in front of you, or will we actually start to see R&D growth slow relative to the revenue?"
— Andrew Gardiner, Citi

A: "So a little early to talk full year. I can tell you right now, our expectation is that the Q4 to Q1 step up will be similar to last year ... low double-digit sequential growth ... I right now, I would say the growth after Q1 is probably gonna to moderate more so than it did this year. We did see pretty significant step-ups throughout the year. I don't expect there to be quite significant step-ups for next year."
— Jason Child, CFO

Assessment: The answer commits to a direction without a number, and the direction is only that the rate of increase slows. Set against roughly 20% revenue growth, that is not a margin-expansion commitment. This is the cleanest gradeable commitment from the call and the one most likely to matter to FY27 estimates.

Whether SoftBank needs to sell stock to fund its AI ambitions

The overhang question that has followed the shares since the IPO, sharpened by the parent's escalating AI capital commitments. It drew the most personally sourced answer of the call.

Q: "Wondering what your thoughts on are whether or not SoftBank will potentially need to sell some of the ... stock that it holds to finance some of the investment you've talked about making and how we should think about the implications for your shares?"
— Simon Leopold, Raymond James

A: "You know, that's one that we read a lot about, and there's a lot of speculation on chat boards and whatnot about that. I can tell you from talking to Masa about this, and I would quote him directly, he is not interested in selling one share of ... stock. And that doesn't mean two shares or three shares. That means any shares."
— Rene Haas, CEO

Assessment: About as forceful a denial as a chief executive can offer regarding a shareholder's plans, and it plainly contributed to the reversal in the shares. It remains a characterisation of a third party's present intent rather than a contractual restriction. The overhang is deferred, not removed.

What They're NOT Saying

1. Remaining performance obligations went unmentioned. RPO fell 8% year over year and 4.4% sequentially, the first material decline of the fiscal year, while management discussed ACV at length as the licensing health indicator. The letter disclosed it; the call never raised it and no analyst asked. A licensing business whose contracted backlog shrinks while reported licence revenue grows 25% is worth a question.
2. The SoftBank agreement's terms. The dollar figure and the run-rate expectation were given. The contract length, total value, recognition pattern and renewal mechanics were not. For a stream representing 16% of total revenue and 40% of the licence line, from the controlling shareholder, that is a thin disclosure set.
3. The link between related-party revenue and related-party contract assets. Contract assets from related parties rose to $499 million from $152 million at fiscal year end, a $347 million build over nine months, in the same period the SoftBank agreement began contributing. Neither document connects the two, and free cash flow conversion falling to 37% is the visible consequence.
4. What "complete SoCs" means for the business model. Named once as a destination for R&D spend, never explained. A company that sells complete SoCs competes with its licensees and carries a different gross margin and capital profile. Capex up 184% may already be the leading indicator, and capex received no commentary at all.
5. Data center revenue in dollars. Declared the fastest-growing business, the reason the CPU thesis works, and the future largest segment. Disclosed annually and, when asked directly, given as a range rather than a number.
6. Which memory sensitivity is correct. A 20% handset unit decline was sized at "2% or 4% at worst" on smartphone royalties early in the call and "at most a kind of four to 6%" later. Both were offered as the same calculation. Neither was reconciled.
7. Anything about the March 24 event. Pre-emptively fenced off in the prepared remarks before a single question was asked. Given that the FY28 deferral was justified by "other possible offerings", the event appears to carry a business-model announcement rather than a product refresh, and the company has chosen to let the market wait.
8. GAAP profitability. GAAP net income fell 12% and GAAP EPS fell 13% on $285 million of equity-settled share-based compensation equal to 23% of revenue. Neither the prepared remarks nor any question referenced GAAP results.

Market Reaction

• Pre-print setup: ARM closed at $104.90 on February 4, down 4.0% year to date against the S&P 500 up 0.5%, down 9.7% over the trailing 30 days and down 35.3% over the trailing twelve months. The 52-week closing range entering the print was $85.82 to $178.62, so the shares came in nearer the low than the high after a year of sustained de-rating.
• After-hours move: Shares fell roughly 8% immediately following the release, before the call. The decline reflected the composition of the licence line and the read-through from handset supply constraints to the royalty base, not the headline print, which beat on both revenue and earnings.
• Next-day session: The entire after-hours decline was recovered before the opening bell. ARM opened February 5 at $104.48, down just 0.4%, traded a $100.02 to $115.10 range, and closed at $110.88, up 5.7% or $5.98. The S&P 500 fell 1.2% the same session.
• Volume: 24.6 million shares against a 30-day average of 6.1 million, 4.1 times normal.

The print was sold and the call was bought. That sequence is unusually legible here. The release crossed with a headline beat, and the immediate reaction was negative: a 25% increase in licence revenue looks materially worse once the reader reaches the related-party line, and the handset supply narrative was already the sector's dominant worry. Everything that reversed it happened on the call rather than in the letter. The data center royalty doubling, the quantified memory sensitivity that reframed a unit problem as a mix question, and a categorical denial that the controlling shareholder intends to sell stock were all call artifacts.

The 15% intraday range on 4.1 times normal volume is the more telling statistic than the close. A session that opens flat, trades down 4.7% and closes up 5.7% is not a market digesting a modest beat. It is a market with genuinely unsettled views on what this business is, resolving them in real time as the call transcript circulated. That the reversal held into the close against a market down 1.2% makes it a real repricing rather than a squeeze.

Context matters for how much to read into it. The shares entered this print having lost more than a third of their value over twelve months, so positioning was not demanding and the bar was low. A 5.7% day recovers roughly six weeks of the preceding drift and leaves ARM still down 35% from a year ago and below where it started the calendar year.

Street Perspective

Debate: Does data center growth outrun the smartphone cycle?

Bull view: Data center royalties are doubling annually off a base management now puts near 20% of revenue, driven by hyperscaler core-count expansion that is visible in shipping silicon from all four major cloud providers. Even a severe handset unit decline costs one to two points of total royalty growth. The mix shift does the work regardless of the phone cycle.

Bear view: Smartphones remain 40% to 45% of revenue and data center is not yet half that. Doubling a teens-percentage business does not offset a downturn in a business twice its size, and the Q4 royalty guide decelerating to low teens is the first evidence that the offset is not yet automatic.

Our take: The bulls have the better of this over a three-year horizon and the bears have the better of it over the next two quarters. The insulation argument is arithmetically sound because low-end units carry the lowest rates, and management sized it credibly. But the Q4 deceleration shows royalty growth is not yet decoupled from handset seasonality, and management attributed the deceleration to comps rather than claiming the data center offset absorbed it. The mix shift is real and it is early.

Debate: Is the SoftBank licensing revenue worth the same multiple as the rest?

Bull view: It is contracted, recurring at a declared $200 million quarterly run rate, tied to a controlling shareholder with escalating AI silicon ambitions and the balance sheet to fund them. Management has said the work relates to current-generation designs that continue as the roadmap executes. That is a durable, high-margin annuity from the most committed customer Arm could have.

Bear view: It is 40% of the licence line and 16% of revenue from the entity that controls the board, arriving with no disclosed term, no disclosed value, and a tripling of unbilled related-party receivables alongside it. Excluding it, licence revenue fell 24% and total revenue grew 6%. Related-party revenue should be capitalised at a discount, not at 63 times.

Our take: The bear case is closer to right on disclosure and the bull case may well be right on economics. Those are compatible. The problem is not that the revenue is bad; it is that no outsider can test the durability claim, and the company has given the market no way to. Until Arm discloses the agreement's term and structure, a discount to the royalty multiple is the correct treatment, and the market is not currently applying one.

Debate: Does the R&D build create a new business or just compress margins?

Bull view: R&D up 46% is Arm buying its way up the value stack, from IP blocks to compute subsystems to chiplets and complete SoCs. Each step up multiplies revenue per device. CSS already demonstrates the pattern, going from nothing to the teens of royalty mix with management pointing at half over two to three years. Investment ahead of that curve is exactly right, and the March event likely names the next step.

Bear view: Operating expense growing 37% against revenue up 26% has already cost 430 basis points of operating margin, GAAP operating margin is down to 14.9%, and the only commitment offered is that the rate of increase will "moderate." Moving into complete SoCs means competing with the customers who pay the royalties, at lower gross margin and higher capital intensity. Capex is already up 184%.

Our take: The CSS precedent is genuinely strong evidence that Arm can move up the stack and capture more per device without losing customers. But the market is being asked to fund the next step before being told what it is, and the disclosed facts point somewhere more capital-intensive than CSS. We would rather own this after March 24 than before it, which is a large part of why we start at Hold rather than Outperform.

Model Update Needed

This is initiation of coverage, so the table below sets our opening estimates rather than revising prior ones. Company guidance is used for Q4 FYE26; FY27 reflects management's endorsed framework adjusted for the composition issues discussed above.

Valuation: At the February 5 close of $110.88 and 1,069 million diluted shares, ARM carries a market capitalisation near $118.5 billion. Against our FYE26 non-GAAP EPS estimate that is roughly 63 times, and against trailing twelve-month free cash flow of $893 million the free cash flow yield is 0.75%. Enterprise value, net of $3.5 billion in cash and short-term investments with no disclosed debt, is approximately 23.5 times our FYE26 revenue estimate. On our FYE27 numbers the multiple compresses toward the low fifties, which still requires the data center mix shift to land close to management's description.

Valuation impact: We are not publishing a price target with this initiation. The range of outcomes turns on a March 24 disclosure that management has explicitly declined to characterise, and on the durability of a related-party revenue stream whose terms are undisclosed. Both are resolvable within two quarters, and we would rather set a target on the other side of them than anchor one to a business model that management has signalled is about to change.

Thesis Scorecard Post-Earnings

This is initiation of coverage, so the pillars below are established here rather than carried forward, and their status reflects what this quarter's print and call showed against them.

Overall: Thesis established. The operating case is stronger than we expected going in, and the disclosure case is weaker. Both bull pillars that can be verified from reported numbers were confirmed this quarter, and all three quality-of-earnings risks we would have flagged in advance also confirmed simultaneously. That combination, a genuinely excellent franchise reporting a quarter whose composition is harder to underwrite than its headline, is what a Hold is for.

Action: Hold. We would want to own this business at a price that pays us for the disclosure gap, and 63 times forward non-GAAP earnings at a 0.75% free cash flow yield is not that price, even after a 35% de-rating. The two things that would move us to Outperform are specific and near-dated: disclosure of the SoftBank agreement's term and structure, and a March 24 announcement that clarifies the chiplet and SoC ambition without materially raising capital intensity. A move to Underperform would require the royalty rate escalator to break, and this quarter it did the opposite.