The story told about Arm Holdings after 29 July is that a great quarter got sold for no reason. That is wrong, and the mistake matters if you are trying to price the stock. Arm beat its own revenue guidance, beat its own earnings guidance, and guided the next quarter above consensus — and the shares still closed at $224.89 that day and spent the following session recovering. The market was not disputing the print. It was repricing what Arm intends to become. ARM closed at $268.93 on 11 August 2026, 40.6% below its $452.70 52-week high and 168.9% above its $100.02 low, according to stockanalysis.com. Our bull case is $412. Our bear case is $158. Both numbers fall out of one piece of arithmetic that almost nobody has run in public.
Here it is. Arm’s GAAP gross margin in the June quarter was 97.2%, per the Form 6-K it filed with the SEC on 29 July. Its own fiscal-2031 ambition is $25 billion of revenue — but $15 billion of that, 60% of the total, comes from the Arm AGI CPU, a physical silicon business management has guided to gross margins in the high-30s to low-40s initially, working toward roughly 50%. Blend $10 billion of IP at 97% with $15 billion of silicon at 50% and you get a company earning $17.2 billion of gross profit on $25 billion of revenue: a 68.8% blended gross margin. At the low end of the silicon guide, it is closer to 63%. The plan roughly quadruples revenue and cuts the gross margin by nearly 30 percentage points. Arm is not a royalty company growing into a bigger royalty company. It is a royalty company buying its way into the chip business, and on 29 July the market started charging it a chip-company multiple on the margin it is trading away.
Key facts
- Total Q1 FYE27 revenue $1,289m, up 22% year on year, against guidance of $1.26bn ± $50m — Arm Holdings Form 6-K, SEC, 29 July 2026
- Royalty revenue $715m (+22%); licence and other revenue $574m (+23%); GAAP gross margin 97.2%, non-GAAP 98.1% — same filing
- Non-GAAP free cash flow $665m, up 343% year on year; cash and short-term investments $3,888m — same filing
- Q2 FYE27 guidance: revenue $1.38bn ± $50m and non-GAAP EPS $0.47 ± $0.04, both above consensus — same filing
- Trailing P/E 274.5, forward P/E 112.7, price/sales 55.7, EV/FCF 188.5 at the 11 August close — stockanalysis.com, 12 August 2026
- Free float of just 140.53 million shares against 1.07 billion outstanding — roughly 13% of the company — with 17.46 million shares sold short — same source
- FY2031 targets of $25bn revenue and more than $9 adjusted EPS, of which $15bn is silicon — Arm Q1 FY27 investor slides, via Investing.com, 29 July 2026
What actually happened on 29 July
Arm reported the first quarter of its fiscal 2027 — the three months to 30 June 2026 — after the close on 29 July. Every headline number was a beat. Revenue of $1,289 million was a record and 22% above the year-ago quarter, clear of the $1.26 billion guidance midpoint. Non-GAAP EPS of $0.45 came in above the top end of the $0.40 ± $0.04 range and grew 29%. Operating cash flow was $902 million, up 172%, and non-GAAP free cash flow of $665 million was up 343%. Then came the guide, and it was also a beat: $1.38 billion ± $50 million of revenue for the September quarter with non-GAAP EPS of $0.47 ± $0.04, both above where the street sat. On the raw scorecard, three for three.
The stock had already fallen 8.14% on 24 July and 8.11% on 28 July before the numbers landed, then fell another 8.11% on 29 July to $224.89, dropped again after hours, and reversed to close up 7.4% on 30 July. Six days later it printed a 17.4% single-session gain. From the record close of $439.46 on 18 June to the 29 July low, Arm lost 48.8% of its value in 27 trading sessions, and fell 32.4% in the month of July alone.
None of that is the behaviour of a market digesting a beat. It is the behaviour of a market arguing with itself about a five-year plan.
The number that actually moved was smartphones — and the culprit is memory
Strip the call back to what changed and one thing did. Chief financial officer Jason Child walked down the full-year royalty growth expectation. “I think we said last quarter that we’re expecting somewhere around 20% year-on-year,” he told analysts on the Q1 FY2027 earnings call. “Right now, if I had to guess, that’s probably somewhere closer to the high teens right now.” For the September quarter specifically, he said, “we guided to kind of the low to mid-teens.”
The cause he named was not competition, not China, and not the data centre. It was the bill of materials. “We have seen some incremental slowdown versus what was expected at the beginning of the year,” Child said. “Higher memory prices are affecting demand.”
That sentence is the most interesting thing in the whole release, because it makes Arm and Micron the same trade viewed from opposite ends of a phone. DRAM and NAND inflation is compressing handset build economics; handset builds are the volume base on which Arm collects a per-chip royalty; so Arm’s royalty guidance goes down for precisely the reason Micron’s earnings power goes up. Our Micron MU price prediction put the memory maker at $868.52 with a $1,550 bull case on the strength of exactly this pricing cycle. An investor who owns both names is, at the smartphone layer, hedged without knowing it.
It also means the royalty downgrade is cyclical, not structural. Memory pricing is a supply cycle, and supply cycles turn. That distinction is doing a lot of work in the bull case below, and it deserves to be stated plainly rather than assumed.
The data-centre case is real, and it is genuinely good
Everything Arm said about servers was strong, and the third-party corroboration is unusually specific. Data-centre royalty revenue “continues to more than double year-on-year,” Child said. Arm’s share of cloud compute measured by chip value has gone from 9% in fiscal 2022 to 23% in fiscal 2026. Management cited IDC data showing that spending on Arm-based accelerated server platforms has nearly doubled in two quarters and has now surpassed x86 platforms.
The named design wins are not vapour. Chief executive Rene Haas told the call that Nvidia has brought its Vera CPU into production — a part Arm says delivers up to 50% higher CPU performance and twice the energy efficiency of comparable x86 — that Google’s Axion is the host CPU for its latest TPU systems, that Microsoft has expanded Azure Cobalt 200 virtual machines built on Arm Neoverse CSS, and that AWS “announced plans to deploy tens of millions of Graviton5 cores to power agentic AI workloads.”
The efficiency argument underneath all of this is a live debate rather than a settled one. A Hacker News thread on 8 August 2026 titled “Can Intel finally beat ARM on performance per Watt?” drew 213 points and 199 comments — engagement that only lands on questions people think are genuinely open. Intel is contesting the ground, and has just raised $20 billion to fund the attempt, as covered in our report on Intel’s overnight stock sale at $95 a share. Perf-per-watt is Arm’s entire structural moat in the rack, and it is narrowing at the edges even as Arm’s share compounds.
Note the asymmetry in the royalty model, though, because it is the most underrated feature of the business. Arm gets paid on volume regardless of which vendor wins a given socket. When a hyperscaler picks Nvidia’s Vera, Arm collects. When it builds Graviton instead, Arm collects. The model is indifferent to the outcome of the fight it is sitting inside — the property that made Nvidia’s $500 billion infrastructure financing push a tailwind rather than a threat.
So why sell a company like that? The margin, not the growth
Because Arm has decided that collecting a royalty is not enough. In March 2026 it launched the Arm AGI CPU, a data-centre part co-developed with Meta, and the stock jumped 16% on what CNBC described at the time as “a significant shift.” By the July call, Haas said demand “now exceeds $2 billion as we continue to add new customers, including multiple customers in the U.S. and China,” against an initial $1 billion of secured manufacturing capacity.
That is a genuinely impressive commercial start. It is also, at the gross-margin line, a completely different company. Every dollar of AGI CPU revenue arrives with wafers, substrates, packaging and test attached to it. Management has guided first-generation gross margins to the high-30s and low-40s, improving toward 50% as more of the work moves in-house, and the chip business to roughly 35% operating margin at fiscal-2031 scale against about 65% for the IP business.
Run the blend again slowly, because this is the whole argument. Today: $1,289 million of revenue, $36 million of cost of sales, 97.2% gross margin. In fiscal 2031 on Arm’s own plan: $10 billion of IP and CSS at roughly 97%, plus $15 billion of silicon at 50% in the good case. That is $9.7 billion plus $7.5 billion, or $17.2 billion of gross profit on $25 billion of revenue — 68.8%. If the silicon business lands at 40% instead, it is $15.7 billion, or 62.8%.
A software-like business trading on a software-like multiple has announced it is becoming a two-thirds-gross-margin hardware-and-IP hybrid. That is the disclosure the market repriced. It has nothing to do with the June quarter, which is precisely why every article that examined the June quarter concluded the sell-off was irrational.
The FTC probe is not a side risk — it is the same story
On 15 May 2026, Bloomberg reported that the Federal Trade Commission had opened an antitrust investigation into Arm’s licensing practices. The question the FTC is reportedly asking is whether Arm might degrade or deny the CPU architecture licences that Apple, Qualcomm, Nvidia and hundreds of other firms depend on, while selling its own competing silicon into the same market. The trigger was the March AGI CPU launch.
Most coverage files this under legal overhang. It is better understood as the price of the bull case. Arm cannot simultaneously own a $15 billion merchant silicon line and a near-monopoly architecture licence over its new competitors without a regulator asking whether the second is being used to protect the first. The strategy that creates the upside creates the enquiry. You do not get to underwrite one and discount the other, and the arrival of a former licensor-turned-competitor into a concentrated market is a pattern US antitrust has form on.
Now the mechanical problem: there is barely any stock
Before any target is credible, the microstructure has to be stated. Arm has 1.07 billion shares outstanding and a free float of 140.53 million — about 13% of the company. SoftBank owns essentially all of the rest. Against that float, 17.46 million shares are sold short. The five-year beta is 3.91. Using the same daily closes that produced the chart above, Arm’s realised annualised volatility over the last 60 sessions is roughly 108%.
This explains behaviour that otherwise looks unhinged: three consecutive sessions down more than 8%, then a single session up 17.4%. Those are not repricings on new information. They are a very small quantity of stock being moved by a very large quantity of conviction. It is the same float dynamic that produces the violent moves in names like Palantir, only more extreme, because Arm’s float is smaller.
The honest consequence is that every price target on this name — including both of ours — carries a wider error bar than the analysis behind it. Sell-side targets currently span $125 to $450 with an average of $274.02, per TradingKey. That is a 3.6x spread between high and low, which is itself a confession that nobody has a tight view.
Bull case: $412
The bull case does not require the market to pay a valuation it has never paid. It requires Arm to get back to roughly where it traded on 18 June 2026, nine weeks ago. That is the whole ask, and it is worth saying out loud because it makes the target auditable rather than aspirational.
The arithmetic: royalty revenue compounding in the high teens once the memory-driven smartphone drag rolls off, data-centre royalties continuing to more than double off a small base, and the AGI CPU line beginning to recognise against the $2 billion of stated demand, gets fiscal-2029 revenue to roughly $8.8 billion. Apply 50 times sales — a discount to the 55.7 times the market pays today — and you get a $440 billion market capitalisation. Across 1.07 billion shares that is $411, which we round to $412, or 53% above the current price.
Cross-check it against earnings and it holds up on the same logic. Arm’s own fiscal-2031 target is more than $9 of adjusted EPS. At $412, an investor is paying about 46 times a number the company expects to earn five years out. That is expensive. It is not absurd for a business with a 97% gross margin on its legacy line, $3.9 billion of net cash, free cash flow that just grew 343%, and a structural share gain in the fastest-growing compute market on earth. The bull case is not that Arm is cheap. It is that the June repricing was directionally right and the July one overshot.
Bear case: $158
The bear case does not need the business to break. It needs the market to stop underwriting fiscal 2031.
Assume the memory-driven smartphone weakness runs longer than a quarter or two, FY27 royalty growth slips from “high teens” toward the mid-teens, and the AGI CPU ramps into revenue on schedule — which is the point at which reported gross margin starts visibly falling. Fiscal-2028 revenue lands around $6.8 billion and investors apply 25 times sales, a multiple appropriate to a high-quality semiconductor franchise with a hardware mix rather than a pure-play IP licensor. That is $170 billion of market capitalisation, or $159 a share, which we round to $158 — 41% below the current price.
Note what the bear case is not. At $158, Arm still trades on roughly 27 times trailing sales and about 18 times the company’s own fiscal-2031 EPS target. It is still one of the most expensive large-cap semiconductor businesses in the world. The bear case here is a de-rating, not a collapse, and it sits well above the $100.02 low the stock printed on 5 February 2026. Anyone modelling a return to that low is modelling a different thesis — that the data-centre share gain reverses — and nothing in the current data supports it.
The faster route to $158 is the FTC. A consent order constraining how Arm prices or conditions architecture licences would attack the 97% margin line directly, at the same moment the silicon line is diluting it. That combination, not a soft quarter, is the genuine tail risk.
What we are watching next
Three things, in order of how much they will move the stock.
First, the FY27 royalty language at the Q2 print in late October. The revenue and EPS numbers will almost certainly clear the $1.38 billion and $0.47 midpoints; Arm has beaten its own guidance consistently. What matters is whether “high teens” full-year royalty growth holds. If it slips to mid-teens, the bear case activates, because it converts a memory-cycle story into a demand story.
Second, segment disclosure. As AGI CPU revenue starts being recognised, consolidated gross margin has to fall. Watch for Arm to begin reporting IP and silicon separately. That disclosure change would be a tell in both directions: management confident enough to show the mix, or management pre-empting a headline margin decline. Either way, the first quarter in which reported gross margin prints below 95% will be a genuine event for a stock that has always been valued on the 97% figure.
Third, SoftBank. With 13% of the company in public hands, any secondary offering is the single largest supply event available to this stock, and a rational thing for a majority holder to do after a 146% year-to-date gain. Arm’s annual general meeting is on 9 September 2026 in Cambridge. Nothing in the filings signals a sale — but with a float this thin, the absence of a signal is not the absence of a risk.
The scorecard: a very good business, a genuinely accelerating end market, a strategy that is correct and margin-destructive at once, and a share count so small that the price keeps telling you about positioning rather than about the company. Where the margin structure is stable and the float is not the story, our AMD stock forecast and Texas Instruments price prediction make useful contrasts.
Frequently asked questions
Why did Arm stock fall after beating earnings in July 2026?
Arm beat on revenue, EPS and forward guidance, but cut its full-year royalty growth expectation from around 20% to the high teens, blaming higher memory prices hurting smartphone demand. More importantly, its fiscal-2031 plan puts $15 billion of $25 billion in revenue into a silicon business with gross margins near 40-50%, against 97.2% on its IP today. The market repriced revenue quality, not the quarter.
What is Arm’s current share price and valuation?
ARM closed at $268.93 on 11 August 2026, 40.6% below its 52-week high of $452.70. It trades on roughly 274 times trailing earnings, 113 times forward earnings and 55.7 times sales, per stockanalysis.com. Enterprise value is about $284 billion.
How much of Arm does SoftBank own?
SoftBank retains roughly 87-90% of Arm following the 2023 IPO. The public free float is about 140.53 million shares against 1.07 billion outstanding, or roughly 13% of the company — one of the thinnest floats among large-cap US-listed technology stocks, and a direct cause of the stock’s volatility.
What is the Arm AGI CPU and why does it matter?
It is Arm’s first own-brand data-centre processor, co-developed with Meta and launched in March 2026. Management says demand now exceeds $2 billion against $1 billion of secured capacity, and targets $15 billion of revenue from it by fiscal 2031. It matters because it turns Arm from a licensor into a competitor of its own customers — which is both the growth story and the reason the FTC opened an antitrust probe in May 2026.
Is the FTC investigation a serious risk to Arm?
It is the risk most directly attached to the valuation. The FTC is reportedly examining whether Arm might degrade or deny architecture licences to firms it now competes with. Any remedy constraining how Arm prices those licences would hit the 97% gross-margin line at the same time the silicon business is diluting it. Probes of this kind typically take 12-24 months, so a resolution before mid-2027 is unlikely.
What would make the $412 bull case work?
Smartphone royalty weakness proving to be a memory-price cycle rather than structural demand loss, data-centre royalties continuing to more than double, and the AGI CPU converting its stated $2 billion of demand into recognised revenue. That path supports roughly $8.8 billion of fiscal-2029 revenue at 50 times sales — a $440 billion market capitalisation, or $412 a share.
This article is analysis, not investment advice. Figures are sourced and dated; prices are as of the close on 11 August 2026 and will have moved. Do your own research before making any investment decision.