Two Quality-of-Earnings Bears Died This Quarter. Arm Retired the Metric Behind the Third.

Results vs. Consensus

Arm reported its fiscal first quarter, ended June 30, 2026, after the close on July 29. It was a clean double beat and the highest first-quarter revenue in the company's history. Both revenue lines cleared the roughly 20% year-over-year growth management had guided to in May, and non-GAAP EPS of $0.45 landed above the $0.44 top of the guided range rather than merely above the midpoint. The Street had done no independent work: consensus EPS of $0.40 was the company's guide midpoint to the cent.

Q1 FYE27 Scorecard

Year-over-Year Comparison

Sequential Comparison

The sequential table needs a seasonal caveat before it is read as deterioration. Arm's fiscal fourth quarter is structurally its heaviest licensing quarter and its first is the lightest, so a double-digit sequential revenue decline into Q1 is the normal shape of the year: total revenue fell 15.1% sequentially into Q1 FYE26 against 13.5% this year, so the step down was actually shallower. What is not seasonal is the operating expense line holding flat at $733 million while revenue fell $201 million sequentially, which is why non-GAAP operating margin gave back 790 basis points. That is arithmetic, not deterioration.

Revenue

Total revenue of $1,289 million grew 22% and set a first-quarter record. For the first time in our coverage both halves of the model grew for the same reason rather than opposite ones: royalties on rate escalation and data center share, licensing on what management described as multiple high-value renewals from hyperscalers, automotive and robotics companies and handset makers securing access to the forward roadmap. That is a better composition than February, when licensing growth was entirely a single related-party agreement. It is not yet a clean composition, because the agreement is still 33.6% of the licence line and related parties are still 30.1% of total revenue.

Assessment: The gap between the reported 22% and the ex-SoftBank 4.1% is the single most important number in this report and it has not narrowed since February. Both figures are true; they describe different companies. Until the agreement's terms are disclosed, an investor cannot know which one to capitalise.

Margins

Non-GAAP operating margin of 41.2%, up 210 basis points, is the datapoint that most directly contradicts our February bear case. We wrote then that operating expense growing 37% against revenue up 26% was "investment, not deleverage, but the payback sits in a period nobody has guided yet." The payback arrived one quarter later than the earliest date we would have guessed. Non-GAAP R&D grew 20% against revenue up 22%, taking R&D from 41.8% to 41.1% of revenue, the first decline in that ratio in the coverage period.

The GAAP picture is the mirror image and deserves its own sentence. GAAP operating income of $91 million on $1,289 million of revenue is a 7.1% margin, down from 10.8%, because equity-settled share-based compensation of $343 million is now 26.6% of revenue and growing at twice the rate of the top line. Arm is paying for its engineering build in stock, and the stock has been expensive, which makes the accounting cost of that build rise with the share price.

Assessment: The non-GAAP operating leverage is real and the guided path extends it. The GAAP dilution cost is also real and is not going away while the share price does the compensating. Both belong in a model; only one of them is in the multiple most investors are quoting.

Earnings and Cash

Free cash flow of $665 million in a single quarter is 75% of the $882 million Arm generated across all of fiscal 2026. Trailing twelve-month free cash flow of $1,397 million is up 134%. The mechanism is straightforward and it is the reverse of last quarter's: operating cash flow of $902 million against $332 million, driven by a $389 million working capital release where a year ago working capital consumed $89 million. Capital expenditure kept climbing, with purchases of property, plant and equipment of $197 million against $154 million, and property and equipment on the balance sheet reaching $1,391 million from $714 million fifteen months earlier.

Assessment: The cash conversion bear point is resolved, and resolved in the way we specified: the contract assets converted to cash rather than compounding. What replaces it as the cash-line watch item is capital intensity. Net property and equipment has grown 95% in five quarters, and Arm has just told us it is a silicon company from the end of this fiscal year. Those two facts belong together.

Revenue Line Performance

Arm does not report profit by segment. It reports two revenue lines and organises its narrative around three AI domains, to which it attaches no revenue. What follows uses the disclosed revenue lines, with end-market colour drawn from management commentary.

Royalty

Royalties of $715 million grew 22% and reaccelerated from Q4's 11%. The composition is the point. Management named cloud AI as the largest single driver for a third consecutive quarter, with data center royalty revenue more than doubling year over year on hyperscaler server ramps plus networking silicon where, in the CFO's words, Arm is "deployed in nearly all leading products." Underneath that, the edge held up against a genuinely deteriorating end market.

"Edge AI royalty revenue continued to grow despite a soft end market in smartphones. We continue to benefit from higher royalty rates as Armv9 and compute subsystems continue to increase their penetration in smartphones, tablets and other consumer electronic devices. These drivers more than offset the decline in smartphone sales due to higher memory prices." — Jason Child, CFO

The scale datapoint management chose to lead with is the Neoverse installed base: cumulative shipments surpassed 1.5 billion cores, with the most recent 500 million shipping in nine months against six years for the first billion. Third-party corroboration came in the form of an IDC observation, cited on the call, that spending on Arm-based accelerated server platforms has nearly doubled in the past two quarters and now exceeds x86.

Assessment: This is the cleanest evidence yet that the royalty mechanism is rate-driven rather than volume-driven, because it is the first quarter in which the volume input was unambiguously negative and the line grew anyway. Bull pillar one is not just on track; it has now been stress-tested.

Licence and other

Licence and other revenue of $574 million grew 23% and set a first-quarter record, with $193 million from the SoftBank agreement. Management pointed to multiple long-term renewals in the quarter and, as it has every quarter, redirected attention from the reported line to annualised contract value as the better indicator of underlying trend.

"As always, licensing revenue varies quarter-to-quarter due to timing and size of high-value deals. So we continue to focus on annualized contract value, or ACV, as a key indicator of the underlying licensing trend. ACV grew 13% year-on-year, maintaining strong momentum. This continues to be above our long-term expectation for license revenue growth." — Jason Child, CFO

That framing deserves scrutiny, because the metric management nominates as the truth-teller has decelerated sharply. ACV grew 28% in Q3 FYE26, 22% in Q4 FYE26 and 13% this quarter. The characterisation attached to it has not changed across those three prints: "maintaining strong momentum" was the phrase used in Q4 as well, when the growth rate was 22%. A growth rate that has more than halved in six months is being described in language calibrated for one that has not.

Assessment: If ACV is the right lens, and we agree with management that it is, then the underlying licensing trend is decelerating hard while the reported line is being held up by a related-party agreement and by renewal timing. That combination is the licence line's core problem restated, one quarter on, with worse disclosure attached.

Key Operating Metrics

Key Topics & Management Commentary

Overall Management Tone: Confident and notably less defensive than at Q3, when the posture was built around justifying a related-party licence line and a compressing margin. This call led with proof points rather than explanations, and the one place management was pressed hardest, the supply ceiling on the new silicon business, it answered with specifics rather than framework. Two topics were handled by omission rather than argument: the withdrawal of remaining performance obligations went unmentioned in prepared remarks and unasked in Q&A, and the halving of annualised contract value growth was narrated with the same language used when it was double.

1. Data center royalties double for a third consecutive quarter

The Cloud AI story is no longer a claim about the future. It is now the largest single contributor to royalty growth for the third quarter running, and management supported it with named deployments across every hyperscaler rather than aggregate assertions.

"Arm Neoverse shipments have now surpassed 1.5 billion cores with most recent 500 million shipping in just the last 9 months, where the first 1 billion took 6 years." — Rene Haas, CEO

The specific proof points: NVIDIA has brought Vera into production, delivering "up to 50% higher CPU performance and 2x greater energy efficiency than comparable x86 systems"; Google has said its Axion CPU is the host CPU for its latest TPU systems; AWS announced plans to deploy "tens of millions of Graviton5 cores"; Microsoft expanded Azure Cobalt 200 virtual machines on Neoverse compute subsystems; and Qualcomm announced entry into the AI data center CPU market with the Arm-based Dragonfly C1000. That last one matters most, because it is a new entrant choosing Arm rather than an incumbent expanding on it.

Assessment: The competitive question for this pillar was never whether Arm wins the AI data center CPU socket. It was whether anyone would try to displace it. Five hyperscalers and a new merchant entrant all building on Arm is the answer, and it is the strongest single element of the bull case.

2. The rate escalator proves itself against a falling smartphone market

Our February note argued that royalty growth was rate-driven rather than unit-driven, on management's word. This quarter tested that claim under adverse conditions: memory prices have inflated handset bills of materials and the smartphone market is contracting.

"Generally speaking, we have been somewhat isolated from the negative growth of the smartphone market because of the fact that we have moved the vast majority of the customers to V9, in some cases CSS and in some cases, the second version of CSS. And what that's all contributed to has been an increase in royalty growth." — Rene Haas, CEO

The mechanism is Armv9 and compute subsystem penetration raising the royalty captured per device, which is a mix effect that compounds independently of units. The CEO put the resulting divergence plainly: the smartphone market is "projected to be down by some level of double digits" while Arm is "projecting double-digit growth in royalties in the smartphone market." Management also disclosed that the memory impact has broadened in a way it had not anticipated: the initial expectation was that only the low end would be affected, and "now we are seeing all parts of the market, even some upper and mid-tier being affected."

Assessment: A double-digit royalty gain against a double-digit unit decline is roughly 20 points of rate and mix in a single year, and it validates the pillar more convincingly than three quarters of growth in a rising market did. The offsetting disclosure is that the headwind is spreading up the price stack, which is why the full-year royalty frame came down.

3. Arm AGI CPU: demand doubles, and supply becomes the binding constraint

The March 2026 event we flagged as an unquantifiable risk at initiation produced the Arm AGI CPU, the company's first self-designed data center chip. Ninety days on, the constraint has flipped from demand to supply.

"Initial product has now been delivered to multiple customers, and we have secured the manufacturing capacity needed to support the $1 billion opportunity we outlined last quarter across fiscal 2027 and fiscal 2028. Demand now exceeds $2 billion as we continue to add new customers, including multiple customers in the U.S. and China, while the overall value of our pipeline has continued to strengthen." — Rene Haas, CEO

Initial product has been delivered to multiple customers, first shipments are guided to "the end of this year," and the CEO said confidence in exceeding the $1 billion figure has increased over the past 90 days across wafers, substrates, test capacity and memory. Management will update the revenue and margin profile at the Q3 results, when it has visibility into Q4 FYE27 and fiscal 2028.

Assessment: Demand above $2 billion against $1 billion of secured supply is a genuinely good problem, and the supply commentary is specific enough to be gradeable. The risk has moved from whether anyone will buy it to how much can be built, which is the better of the two risks to carry.

4. The gross margin arithmetic of becoming a silicon vendor

This is where the AGI CPU stops being pure upside. Arm's intellectual property business runs at a 98.1% non-GAAP gross margin. The silicon business does not.

"I think last quarter, we said there was probably going to be somewhere in the high 30% range, maybe low 40s percent for the first generation for — really in Q4 of this year as well as for next year. As I'd say over the next couple of years, we do expect to get to 50%, but that's going to basically just entail us actually bringing more of some of the work that maybe an ASIC has helped us in the past and bring some of that work in-house." — Jason Child, CFO

Management characterised this as unchanged from the prior quarter, while acknowledging that "with some of the price increases and things that have happened, we're still kind of working through how to digest those". The revenue will sit inside the existing lines until it exceeds 10% of total revenue, at which point it becomes a third reported line, expected in fiscal 2028.

Assessment: Take the arithmetic seriously. On our fiscal 2027 estimate of roughly $6.0 billion of revenue with about $0.5 billion of silicon at 40%, blended non-GAAP gross margin lands near 93%, against 98% today. Scale the silicon business to the $2 billion of demand already booked and the blend moves toward the high 80s. Arm is currently valued on a software-like gross margin it is deliberately diluting, and the dilution will not be visible as a separate line until fiscal 2028.

5. Operating leverage finally arrives

For three quarters our central bear point was an expense base outgrowing a record revenue line. This quarter reversed it, and did so against management's own guide.

"Non-GAAP operating expense was $733 million, up 18% year-on-year due to the ongoing R&D investment. This was about $27 million below our guidance due to timing as our spending plan for the year remains largely unchanged." — Jason Child, CFO

The sequence is worth laying out, because it shows a pattern rather than a single beat. At the Q3 call management promised a low-double-digit sequential step up from Q4 to Q1. By the Q4 call that guide had been cut to roughly $760 million, a 3.5% sequential step. The print came in at $733 million, flat sequentially against Q4's $734 million. Guidance for the remainder of the year is mid-single-digit percentage sequential growth, which puts fiscal 2027 non-GAAP operating expense near $3.19 billion, or 17% to 18% growth against fiscal 2026's 33%.

Assessment: Management committed in February that expense growth would "moderate more so than it did this year", and on the guided path it roughly halves. Two consecutive underruns against its own guide is the more useful signal: this is a company that has stopped surprising to the upside on spend.

6. The cash flow reversal

February's weakest line is now the strongest. Free cash flow of $665 million against $150 million a year ago converted at 139% of non-GAAP net income, against 40%.

"More importantly, even at these elevated levels of investments, we generated free cash flow of $665 million for the quarter and $1.4 billion over the trailing 12 months, giving us the flexibility to continue to invest long-term growth." — Jason Child, CFO

The bridge is a $389 million working capital release, including $231 million from receivables. Accounts receivable fell to $1,119 million from $1,300 million at the fiscal year end, and related-party contract assets fell to $577 million from $646 million. Those are the same balances whose build consumed $231 million of cash in Q3 FYE26 and prompted our concern about a receivable-like asset owed by the controlling shareholder.

Assessment: The contract assets converted to cash. That is the specific test we set in February and it passed. A single quarter of 139% conversion is not a run rate, but the trailing twelve-month figure of $1.40 billion against $597 million a year ago is not a timing artifact either.

7. The SoftBank agreement at $193 million, now routine

The related-party licence agreement has settled into exactly the run rate the CFO described in February, and into exactly the same disclosure vacuum.

"Of the $574 million of license revenue, our agreement with SoftBank for technology licensing and design services contributed $193 million. We expect the quarterly run rate for the rest of the year to be around $200 million." — Jason Child, CFO

The contribution history is now four quarters long: none in Q1 FYE26, $200 million in Q3 FYE26, $200 million in Q4 FYE26 and $193 million this quarter, with roughly $200 million guided for the balance of fiscal 2027. What has not appeared in any of those four quarters is the agreement's term, scope, pricing basis, renewal mechanics or the recognition pattern. At initiation we named that disclosure as one of two specific upgrade triggers. It has not arrived.

Assessment: The run-rate stability makes the revenue more predictable without making it better. On the guided run rate, roughly $790 million of fiscal 2027 revenue comes from the controlling shareholder on terms no minority holder can inspect, which is a permanent 15% haircut to the quality of the reported growth rate.

8. Remaining performance obligations retired, and ACV growth halved

This is the most consequential disclosure event of the quarter and it received no airtime at all. Buried in the shareholder letter is the statement that "beginning with Q1 FYE27, we are no longer reporting the remaining performance obligations and the number of extant Arm Total Access and Arm Flexible Access licenses."

The context makes the timing hard to read charitably. Remaining performance obligations fell to $2,148 million in Q3 FYE26 from $2,246 million, which we flagged in February as a divergence to watch because annualised contract value was compounding at 28% over the same period. The metric then fell again in Q4 FYE26, to $2,071 million, down 7% year over year. One quarter after its second consecutive decline, it was withdrawn. Over the same three quarters, annualised contract value growth went from 28% to 22% to 13%.

Assessment: Removing a backlog metric while the forward indicator that replaced it decelerates by more than half is the single worst governance signal of our coverage period. Neither of the two remaining ways to see contracted future revenue is now intact: one is gone and the other is slowing while being narrated as stable.

9. GAAP net income doubles on items that are not the business

GAAP net income of $270 million against $130 million will be widely reported as a doubling of profit. It is not one.

GAAP operating income fell 20% to $91 million, and operating margin fell 370 basis points to 7.1%, because share-based compensation grew 42% to $343 million. The entire increase in GAAP net income came from below the operating line: $128 million of income from equity investments and a $17 million income tax benefit, against a $16 million tax expense a year ago. Together those two items are worth more than the whole of GAAP operating income.

Assessment: Neither item is repeatable and neither is operational. Equity investment gains in particular are a mark-to-market on a portfolio, and this is the same line that swung from $39 million to $6 million between Q3 FYE25 and Q3 FYE26. Any model built off the GAAP net income growth rate is building off a portfolio revaluation.

10. The royalty growth frame comes down for fiscal 2027

The clearest guidance change on the call was not in the Q2 numbers but in the full-year royalty framing, which management walked down in response to a question rather than in prepared remarks.

"I think we said last quarter that we're expecting somewhere around 20% year-on-year for the next few years in royalties, including this year. Right now, if I had to guess, that's probably somewhere closer to the high teens right now." — Jason Child, CFO

Q2 royalty growth is guided to low teens, which the CFO later specified as 13%. Licence growth was raised to roughly 30% to compensate, and on the combined basis management said Arm is "actually ahead of where we'd expected to be." Fiscal 2028 and 2029 royalty growth of 20% or better was described as "very much still intact."

Assessment: The mix of the growth is shifting from the line we assign a premium multiple to the line we discount. Royalty is the annuity; licence is lumpy, renewal-timed and 34% related-party. Holding total growth constant while that swap happens is not a neutral event for valuation, even though it is neutral for the revenue line.

11. TAM inflation and the fiscal 2031 frame

Arm put a "$100 billion plus" CPU total addressable market on the table in March against a Street consensus of $50 billion to $60 billion, and was challenged on it at the time. The debate has since moved past them.

"So we're not changing the numbers at all, but there are a lot of indicators that the numbers that we talked about back in March relative to our view of the TAM may have been conservative." — Rene Haas, CEO

The CFO put the current external estimate at $220 billion. The logic offered for the inflation is specific and worth recording: inference demand is growing, agentic workloads dominate inference, and agentic throughput is constrained by CPU core count rather than accelerator count. The $15 billion fiscal 2031 silicon revenue frame from the March event was referenced by an analyst and not restated by management on this call.

Assessment: The TAM logic is sound and the refusal to raise the published number is the right posture. But a TAM is not a forecast, and the gap between a market that may be twice the size management argued for in March and any committed Arm revenue figure beyond fiscal 2028 remains total. Investors capitalising the $15 billion figure are capitalising a slide, not a guide.

Guidance & Outlook

The Q2 guide is above consensus on both lines and internally consistent: applying the guided +30% to Q2 FYE26 licence revenue of $515 million gives roughly $670 million, and the guided +13% on royalties of $620 million gives roughly $701 million, summing to $1,370 million against the $1,380 million midpoint. The composition, however, is the opposite of this quarter's. Royalty decelerates from 22% to 13% and licence accelerates from 23% to 30%, which means the beat-quality mix inverts.

Implied Q-over-Q ramp: Arm guides one quarter at a time and gave no fiscal 2027 revenue guidance. On management's framing (royalty growth in the high teens for the full year, licence weighted roughly 60/40 to the second half as described at the Q4 call), the implied second half requires licence revenue near $1.87 billion against $1.24 billion in the first half. Part of that step is the first Arm AGI CPU shipments, guided to begin at the end of the fiscal year and reported inside existing lines until they exceed 10% of revenue.

Street at: Consensus took the Q1 guide at face value, with EPS consensus matching the guide midpoint to the cent. The Q2 guide sits roughly 3% above consensus on revenue and 9% above on EPS, so estimates move up.

Guidance style: Conservative and getting more so on the expense line. Management has now beaten its own operating expense guide in consecutive quarters, by $11 million in Q4 and $27 million in Q1, and beat the EPS guide's high end this quarter rather than its midpoint. On revenue the pattern is a modest, reliable beat: +1.4% versus the midpoint in Q4 and +2.3% in Q1.

Analyst Q&A Highlights

Why supply, not demand, is the ceiling on the new silicon business

The sharpest question of the call went directly at the credibility of the supply constraint. If the addressable market is as large as management says and multiple foundries exist, an incremental billion dollars of capacity should not be hard to find. The answer was unusually concrete: the shortage is not in any one input but across all of them simultaneously, and it is binding even for vertically integrated competitors.

Q: "I had a near and a longer-term question on your AGI CPU. Near term, I'm very curious what is preventing Arm from securing supply for just one more billion? I mean it's a $50 billion market, right? And there are a number of foundries who can make it. So I'm just curious what is preventing Arm from getting that extra $1 billion a year from now?"
— Vivek Arya, Bank of America

A: "I'm glad you think $1 billion is not a big number. There are a lot of folks just trying to secure extra supply up to $50 million. It is a very, very, very tight market across everything, whether it's memory, whether it's test equipment, whether it's substrates, whether it's TSMC wafers, it's a very, very, very tight world. I think even my compatriots who are in the CPU business for living and have their own fabs have not been able to supply the demand."
— Rene Haas, CEO

Assessment: A satisfying answer to a hostile question. The reference to integrated competitors failing to meet demand is the strongest available evidence that the constraint is industry-wide rather than a reflection of Arm's newness as a silicon buyer. It also caps fiscal 2027 and 2028 silicon revenue at something closer to the secured $1 billion than the demanded $2 billion.

The gross margin structure of the incremental silicon revenue

The follow-up asked what the extra volume does to margin, and the answer confirmed that scaling the silicon business scales a 40%-gross-margin revenue stream into a 98%-gross-margin company. Management held the prior framing while conceding that input cost inflation has not yet been passed through to pricing.

Q: "I mean, on top of the extra above $1 billion of AGI revenue that you now think you can do, how do we think about the gross margin structure relative to kind of what you were thinking about for the original $1 billion?"
— Joseph Quatrochi, Wells Fargo

A: "No change at this time. We said that — I think last quarter, we said there was probably going to be somewhere in the high 30% range, maybe low 40s percent for the first generation for — really in Q4 of this year as well as for next year. […] Obviously, with some of the price increases and things that have happened, we're still kind of working through how to digest those and what needs to be kind of — what needs to affect our pricing."
— Jason Child, CFO

Assessment: "No change at this time" is doing more work than it appears to. Input costs have risen across wafers, substrates, test and memory, by management's own account in the prior answer, and pricing has not yet been adjusted. A first-generation margin held constant through that is a margin at risk, and the 50% target depends on in-sourcing work that takes "a couple of years."

Whether memory-driven handset inflation breaks the royalty base

A recurring line of questioning tested the resilience of the largest royalty pool as handset makers absorb bill-of-materials inflation and raise prices. The answer separated units from rate cleanly, and produced the quarter's most important disclosure about how the headwind has changed shape.

Q: "My question is on the smartphone market because of the much higher memory handset costs, handset OEMs are absorbing meaningfully more BOM inflation. They're starting to raise prices. I think given that smartphone royalties remain a significant portion of your overall royalty base, can you maybe just talk a little bit about what you're seeing in terms of unit demand and mix and how this impacts your royalty revenue outlook for the fiscal year?"
— Sebastien Naji, William Blair

A: "And I think the new piece was initially the expectation was just going to affect the lower end of the market, and now we are seeing all parts of the market, even some upper and mid-tier being affected. And so that's going to be an incremental for us."
— Jason Child, CFO

Assessment: This exchange is where the full-year royalty guide came down, and it came down in Q&A rather than in prepared remarks. The broadening of the memory effect into mid- and high-tier handsets is new information and it is the reason the number moved. The offset offered was cloud AI overperformance, which is credible but not quantified.

What offsets the royalty deceleration, and whether 20% returns

The natural next question was whether the licence line can carry the walked-down royalty number, and whether the multi-year 20% royalty framework survives. Management answered both, and volunteered that the combined outlook has improved even as its better-quality component weakened.

Q: "Jason, I think, if I hear you correctly, you are talking about maybe royalty revenue growth this year instead of 20%, probably going to be high teens because of all the things happening in the smartphone incremental weakness, et cetera. Anything we should be looking at, at the moment, let's say, on the licensing side that can show up with a little bit of upside to offset the incremental weakness on the royalty side?"
— Charles Shi, Needham & Company

A: "We're now guiding to 13% royalty growth in Q2. We also increased our license growth. Overall, on a combined basis, we're actually ahead of where we'd expected to be. I would say, at least as much license revenue to overcome that. So I don't really expect overall to be any sort of a slowdown."
— Jason Child, CFO

Assessment: The substitution is presented as a wash and it is not one. Royalty revenue is a recurring, unit-linked annuity on an installed base; licence revenue is lumpy, renewal-timed and one third related-party. Trading the first for the second at constant total growth lowers the quality of the growth even when it preserves the rate.

Why research and development landed below plan

An operational question probed the mechanics of the operating expense underrun, which matters because a miss driven by delayed engineering hiring would be bearish while one driven by tooling efficiency would not. The answer distinguished them explicitly and gave a usable forward run rate.

Q: "Just talking about the timing and the sequential quarter-to-quarter cadence of R&D, you talked about how it had come in below your expectations. And I just wonder maybe if you could talk about how that kind of comes to pass. Is it kind of a milestone-based thing and that maybe they're not met and so that the cost slips a quarter or 2? Is it a capitalization issue?"
— Timm Schulze-Melander, Rothschild & Co

A: "And so in this last quarter, we did come in a little bit below, and that would be more not on the people's side because that was pretty much right in line. I would say it was really more on kind of the tools that were utilized and necessary versus what our forecast was. So I think for the most part, you should think about kind of banking that savings that we delivered in Q1. And then going forward, our estimates, I think, are pretty similar, maybe a little bit lower."
— Jason Child, CFO

Assessment: The best answer on the call for modelling purposes. Headcount spend was on plan and the underrun was tooling, which means the savings are structural rather than deferred, and management said so explicitly by telling analysts to bank them. Combined with the mid-single-digit sequential growth guide, this converts our most active bear point into a modelling input.

When silicon revenue becomes a separate reported line

A question on product roadmap turned into the quarter's clearest disclosure commitment, and set a date after which the margin dilution from the silicon business becomes visible rather than inferred.

Q: "And for Jason, do you expect to like start breaking out the silicon side of revenues, the CPU side because that seems to be becoming bigger and bigger, I guess."
— Vijay Rakesh, Mizuho

A: "On the revenue breakout, well, first, as soon as we start shipping, which will be at the end of this year, that's the first milestone. And then once it becomes at least 10% of revenue, we'll break it out separately as a third line. So separate from license, separate from royalties. Lastly, there will be silicon revenue. So I would expect that based on our forecast, that should be broken out then in FY '28."
— Jason Child, CFO

Assessment: A firm commitment on a specific fiscal year, which is welcome, and also an admission that for the whole of fiscal 2027 a structurally different, 40%-gross-margin business will be reported inside lines investors read as intellectual property revenue. That is a full year of blended reporting during the transition that matters most.

Whether the March TAM was conservative

A question on the out-year pipeline drew management into the total addressable market debate, where the interesting detail is that Arm declined to raise its own number even as external estimates doubled past it.

Q: "But what about the demand pipeline for the out years, fiscal year '29 through '31? And I assume the $15 billion in projected AGI revenue expected for '31 contemplates all the supply chain-related headwinds, that's a supply-served revenue, correct?"
— Gary Mobley, StoneX Group

A: "So when we talked about the $100 billion number in March, there was some surprise around it, and there was a lot of back and forth in terms of justifying that number. Since that time, a number of my peers have talked about numbers quite a bit bigger than that, in some cases, up to $200 billion. […] So as inference demand goes up, which it clearly is, and as agentic demand goes up as it clearly is, that means an increase in CPU demand. So we're not changing the numbers at all, but there are a lot of indicators that the numbers that we talked about back in March relative to our view of the TAM may have been conservative."
— Rene Haas, CEO

Assessment: Restraint is the right call and the CPU-bound-inference argument is the most persuasive structural claim management makes. Note what was not answered: the question asked whether the $15 billion fiscal 2031 figure is supply-served, and no direct confirmation was given. The CFO separately said internal targets are higher than the published numbers, which is the kind of remark that fuels a multiple without committing to anything.

What They're NOT Saying

1. Remaining performance obligations, now withdrawn: The metric was declining for two consecutive quarters and was removed in the third, with the announcement confined to the shareholder letter. It was neither explained in prepared remarks nor raised in Q&A. There is now no disclosed measure of contracted-but-unrecognised revenue.
2. The Total Access and Flexible Access licence counts, also withdrawn: Removed in the same sentence. These were the only quantitative window into how many customers are adopting the subscription licensing model that underpins the ACV figure management now asks investors to rely on exclusively.
3. Any term of the SoftBank agreement, four quarters in: No duration, no scope, no pricing basis, no renewal mechanics, no recognition pattern. On the guided run rate roughly $790 million of fiscal 2027 revenue from the controlling shareholder remains uninspectable, and this was an explicit upgrade trigger we set at initiation.
4. Data center royalty revenue in dollars: Management has now said "more than doubled" for three consecutive quarters without ever sizing the base. The single most important growth driver in the company cannot be modelled from disclosure, only from the aggregate royalty line.
5. Any full-year fiscal 2027 revenue guidance: Arm guides one quarter forward. The full-year royalty framing was walked down from 20% to high teens in a Q&A answer, and the licence offset was described qualitatively as "at least as much" without a figure.
6. Where Arm AGI CPU revenue will be reported during fiscal 2027: Confirmed only that it is not a separate line until fiscal 2028. Given the guided second-half licence acceleration to roughly 30% growth, investors cannot separate silicon revenue from intellectual property revenue in the quarters where the distinction first starts to matter.
7. The capital intensity plan: Net property and equipment has grown 95% in five quarters to $1,391 million, and Arm has just become a chip vendor. No capital expenditure guidance, no target for capital intensity, and no discussion of what a silicon business needs on the balance sheet.
8. The reconciliation of February's inconsistent smartphone sensitivity: Two different figures were given on the same prior call for the royalty impact of a handset unit decline. No restated sensitivity was provided. Instead the exposure was described as having broadened up the price stack.
9. The $15 billion fiscal 2031 figure: Raised by an analyst, not restated by management, and the question of whether it is supply-constrained went unanswered.

Market Reaction

• Pre-print setup: Shares closed at $224.89 on July 29, having fallen 8.1% in that session alone and roughly 13% over the prior week. The stock was down 34.5% over the trailing 30 days and 48.8% below its record close of $439.46 on June 18. Despite that drawdown it entered the print up 105.7% year to date and 37.6% over trailing twelve months, against a 52-week closing range of $104.55 to $439.46.
• Reaction session (July 30): Opened at $258.00, a 14.7% gap up, traded as high as $268.48 (+19.4%) and as low as $232.67 (+3.5%), and closed at $241.54, up 7.4% or $16.65. The close surrendered more than half the opening gap, ending 6.4% below the open. Volume of 12.8 million shares ran 1.6 times the 30-day average of 8.0 million. The S&P 500 rose 1.7% that session.
• Subsequent sessions: $239.69 on July 31 and $239.06 on August 3, then a sharp recovery on a broad semiconductor rally, with the August 4 session alone adding 17.4%. The last completed session before publication closed at $278.65 on August 13, which is 23.9% above the pre-print close, 15.4% above the reaction-session close, and 36.6% below the June 18 peak.

The setup explains more of the price action than the print does. A stock that has halved from its June peak in five weeks, on a sector-wide de-rating of AI infrastructure names rather than anything company-specific, carries a very different reflex to a double beat than one making new highs. The 14.7% opening gap was that reflex, and it was mechanical: revenue above consensus, EPS above the guided high end, a Q2 guide above the Street on both lines, and $2 billion of demand against $1 billion of secured supply on the new silicon product.

The all-day fade from that gap is the more informative half of the session. Three things were available to sellers by mid-morning that were not in the headline. The full-year royalty growth frame had been cut from roughly 20% to high teens, and it was cut in Q&A rather than disclosed in prepared remarks. Q2 royalty growth is guided to 13%, less than a third of the licence line's guided 30%, so the higher-quality revenue line is the decelerating one. And the operating expense guide steps back up to roughly $780 million after a quarter of restraint, which caps the margin expansion the print had just demonstrated. A stock at this multiple has no room to absorb a downgraded annuity, and the market took the gap back accordingly.

The subsequent recovery to $278.65 is not a verdict on the quarter. The August 4 session added 17.4% on a sector move, and Arm's beta to AI infrastructure sentiment has been the dominant driver of its price for two months in both directions. Positioning in this name is currently a bet on sector flows at least as much as on the franchise.

Street Perspective

Debate: Does the AGI CPU create value or destroy the multiple?

Bull view: The bull case on the Street holds that a $2 billion demand book in the first two quarters of a brand-new product line, in a CPU market management now believes could be $200 billion, is the largest optionality in semiconductors, and that a 40% gross margin on incremental revenue Arm would otherwise never have touched is accretive to dollars of profit even if dilutive to percentages.

Bear view: The bear camp argues that Arm is trading its structural distinction for volume. The 98% gross margin, negative-working-capital, no-inventory licensing model is precisely what earned the multiple, and a business at 40% gross margin with wafer commitments, substrate exposure and rising capital intensity earns a foundry-adjacent multiple, not a royalty one. Selling silicon also puts Arm in competition with the licensees who pay the royalties.

Our take: The bears have the better of the valuation argument and the bulls have the better of the strategic one. Both can be right: incremental profit dollars go up while the multiple those dollars deserve goes down. The reason this is not resolvable yet is disclosure, not analysis. Silicon revenue sits inside the existing lines until fiscal 2028, so for the entire transition year investors will be marking a blended gross margin they cannot decompose. That is an argument for waiting, which is what a Hold is.

Debate: Can data center royalties outgrow structural smartphone weakness?

Bull view: The optimistic camp points out that data center royalties have more than doubled for three consecutive quarters, that the Neoverse base is compounding at an accelerating rate (500 million cores in nine months against six years for the first billion), and that every hyperscaler plus a new merchant entrant is now building on Arm. On that trajectory the smartphone line becomes a rounding error on the growth rate within a few years.

Bear view: The skeptics note that management has never sized the data center royalty base in dollars, so "more than doubled" is unfalsifiable and may be doubling off a small number, while smartphones are the majority of a royalty line that just had its full-year growth guide cut. The memory-driven handset weakness has broadened from the low end into mid- and upper-tier devices, and the Q2 royalty guide of 13% is the arithmetic result.

Our take: This quarter goes to the bulls on evidence and to the bears on disclosure. Growing smartphone royalties by double digits into a double-digit unit decline is a genuinely impressive demonstration of the rate mechanism, and it is the first time the claim has been tested against a falling market. But a company that will not size its most important growth driver after three quarters of citing it is asking to be taken on faith, and the faith premium is already in the price.

Debate: Is 139x forward earnings defensible on the fiscal 2031 frame?

Bull view: The bull framing is that near-term multiples are the wrong lens for a company whose silicon business goes from zero to a $15 billion revenue frame by fiscal 2031 while the royalty annuity compounds at 20% underneath it. On that arc, today's earnings are irrelevant and the only question is whether the terminal share of a $200 billion CPU market is being underestimated.

Bear view: The bear camp counters that Arm has guided exactly one quarter forward, has never guided a full year, walked down its multi-year royalty framework this quarter, and declined to confirm whether the fiscal 2031 figure is even supply-served. Paying 139 times a fiscal 2027 estimate and accepting a 0.47% free cash flow yield for a 2031 outcome that management will not restate is underwriting a slide.

Our take: We are with the bears, with a caveat we have earned the hard way. Our February Hold made a version of this argument at 63 times forward earnings and the stock then rose 151%, so we hold this view knowing the market has recently and repeatedly disagreed with our valuation discipline on this name. What has changed is that the operating case is stronger than we underwrote, and what has not changed is that the disclosure needed to underwrite the 2031 frame has gone backwards rather than forwards this quarter. A multiple that has more than doubled while the company withdrew a backlog metric is not a multiple we will chase.

Model Update Needed

Valuation impact: We continue to publish no price target on ARM, for the same reason as at initiation: the range of outcomes turns on disclosure that does not exist. At the August 13 close of $278.65 the shares carry a market capitalisation near $300 billion against $3.4 billion of net cash, so an enterprise value near $297 billion. That is roughly 139 times our fiscal 2027 non-GAAP EPS estimate of $2.00, about 50 times our fiscal 2027 revenue estimate, and a 0.47% trailing free cash flow yield. At initiation the same framework produced 63 times forward earnings and a 0.75% free cash flow yield. Our estimates have gone up; the multiple has gone up more than twice as fast.

Thesis Scorecard Post-Earnings

Overall: The operating thesis strengthened materially and the disclosure thesis deteriorated. Two of the three quality-of-earnings bear points that drove our February Hold have been answered on the exact tests we set: operating leverage turned positive and the contract assets converted to cash. Both remaining bear points are about what we cannot see rather than what we can, and this quarter the amount we cannot see grew. Net of both, the standing thesis is stronger on fundamentals and weaker on verifiability, which is the same trade at a materially higher price.

Action: Hold. The two specific triggers we published at initiation were disclosure of the SoftBank agreement's terms and clarity on the SoC ambition without a material rise in capital intensity. The second half-arrived: the ambition is now clear and quantified, but capital intensity is rising and unguided. The first did not arrive at all, and a different disclosure was withdrawn instead. For existing holders the operating trajectory justifies staying, and the free cash flow inflection genuinely de-risks the balance sheet. For new capital, 139 times a fiscal 2027 estimate with 15% of revenue uninspectable and no backlog metric is not a defensible entry. We would revisit on any of three events: disclosure of the SoftBank agreement's terms, reinstatement of a contracted-revenue metric, or a de-rating that puts the shares nearer 90 times forward earnings.