Arm designs the brains inside nearly every phone chip
Chip designs
licensed to the companies that make the chips
Phones to servers
phones, cars, laptops and data centres
Its own chip
a server chip of its own since March 2026
Arm does not make chips. It draws the blueprint a processor follows and sells the right to use it to Apple, Qualcomm, NVIDIA and many others.
The company · 2 · What exactly does it sell?
Two kinds of fee, about half each
Each $1 of sales in fiscal 2026, by kind of fee
53%47%
A fee on every chipFees to use the designs
Chip makers pay a licence fee to use a design, then a royalty on each chip they ship. Its own server chip will add a third kind of sale.
The company · 3 · How does it make money?
Paid once to design with it, then again per chip
53 centsof each $1 from royalties on chips shipped
47 centsof each $1 from licence fees and services
23 centsof each $1 paid to staff in shares
A royalty keeps coming for as long as a chip sells, often for years. Licence fees jump around with the timing of large deals.
Is it a good business? · 4 · What gives it an edge?
Phone software is written for Arm, so chip makers stay
Over 99%
of smartphone processors use its designs
Over 10 billion
chips shipped by partners in one quarter
Over 30 million
developers write software for it
Moving a product line away means rewriting years of software. The US competition regulator called Arm's designs a critical input rival chip makers rely on.
Is it a good business? · 5 · What is the catch?
A third of sales comes from friends, not strangers
16%of revenue from Arm China, which Arm does not run
$704.4Mfrom a SoftBank affiliate in fiscal 2026, up from $145.5M
Licence fees from outside customers fell 9% in fiscal 2026; the growth came from Arm China and SoftBank's affiliate, whose payments are slow to arrive.
Is it a good business? · 6 · What does its record show?
Sales nearly doubled, but official profit shrank as a share
Sales, $B
2.7
2.7
3.2
4.0
4.9
5.5
Mar 2022Mar 2023Mar 2024Mar 2025Mar 2026last 12 mo
Sales$2,703M → $4,920Mrevenue, fiscal 2022 to 2026
Profit23.4% → 18.3%official profit per $1, fiscal 2022 to 2026
Share pay23.2%of revenue paid to staff in shares, past year
Revenue grew in four of the last five years. Research and pay in shares grew faster, so the official profit on each dollar fell.
Is it a good business? · 7 · How much room is there to grow?
Phones are full; servers are where Arm can grow
Just over 1 billionphones forecast to ship in 2026, down 16.7%
Just over 10%of server processors were Arm-based in 2025
About 30%forecast for 2027 by TrendForce
Arm is already in almost every phone, so phone growth must come from higher fees per chip. In servers Intel and AMD still sell most processors.
Is it a good business? · 8 · Who is in charge?
SoftBank owns most of it and picks the board
86.4%of the shares owned by SoftBank Group
Rene Haaschief executive since 2022, also on SoftBank's board
$60.6MHaas's pay for fiscal 2026, mostly in shares
Bonuses pay on revenue and on a profit measure that leaves out pay in shares, so pay rewards growth more than profit per share.
The story right now · 9 · How much do AI data centres matter here?
Data centres are moving to Arm as AI spending booms
Over 50%Arm's share at the top cloud providers in 2025, says management
Above 20%Arm's share of all server processors forecast for 2026
AI needs many more servers, and each one needs main processors as well as AI chips. Arm earns a royalty on each, and now sells its own chip too.
The story right now · 10 · Why count pay in shares as a cost?
Management adds share pay back; we count it
Revenue grows fastUp 27% in the September 2026 quarter.
Staff are paid in shares23.2% of revenue over a year.
Management's profit leaves it out44 cents per $1 of sales.
Counted, profit is thin14 cents per $1 that quarter.
New shares handed to staff shrink each owner's slice as surely as a cash wage. Counted as a cost, the profit owners keep is a third of management's figure.
The story right now · 11 · What changed lately?
Licence fees jumped, royalties slowed, and the shares fell
$1,443Mquarterly revenue, up 27%
8%royalty growth, as phone sales weaken
13.1%share fall the day after the report
Solid quarter
Management says data-centre royalties doubled. But cash from operations fell to $207M from $567M, as fees owed by related parties built up unpaid.
The story right now · 12 · What is everyone arguing about?
Can its own server chip earn what the price expects?
Many investors think
The blueprint of AI data centres, with its own chip about to add billions
We think
A rare franchise whose pay in shares leaves owners thin profit
Against us: if the chip reaches management's plan and pay in shares falls fast, profit could grow into the price sooner than we expect.
The price and our view · 13 · What is a share worth?
The price sits far above even our sunny future
$49.01what we think a share is worth · the last close is of that
$49.01our value
Attractive below $39.21
Neutral
Unattractive above $58.81
$16.18gloomy: growth half what analysts expect
$49.01our estimate: growth as analysts forecast, then slowing
$74.20sunny: growth a quarter faster than forecast
Value per share in three futures. None is a prediction of the share price.
We estimate what a share is worth from the cash the business should earn over many years, counting pay in shares as a cost. Our rule: buy only at least a fifth below our estimate.
The price and our view · 14 · What if it all works?
As a bet on its own chip, it is priced for success
Priced for successthe kind of bet this is
0.8×if it works
74%kept if it fails
Multiples of the company's value when the bet was last struck, 16 Nov 2026.
Even if the server chip reaches management's plan, the shares would be worth less than today; if it fails, more is lost.
The price and our view · 15 · How could we be wrong?
Related parties, its own customers and a free rival
Arm China and SoftBankBiggest
A third of revenue, slow to pay.
Competing with customersWatch
Its own chip may push licensees away.
Qualcomm's lawsuitShowing now
About 9% of revenue at stake.
What the chip earnsUnknown
No profit shown yet.
RISC-V, free to use and backed by many of Arm's customers, could also press on what Arm can charge.
The price and our view · 16 · How will we know?
3 February shows whether royalties grow again
Revenue of $1.54B or more for the quarterGood sign
Royalties growing above 10% againGood sign
The server chip slipping past March 2027Bad sign
FEB 20273
Results for the quarter to December 2026, about 3 February. Companies report every three months; each report is a test of the story.
Our view also changes on price alone: if the shares fall to within a fifth above our estimate, we move to Neutral.
The price and our view · 17 · So, what do we think?
A rare franchise at a price far above its worth
$49.01our value
Attractive below $39.21
Neutral
Unattractive above $58.81
UnattractiveHigh conviction
Chip makers cannot easily leave, and data centres are moving to Arm. But pay in shares eats most of the profit, and the price is about 357% of our value, above even our sunny future.
↑ ↓ to move · C for contents
What it does
More about the company
Who it is
Every phone, laptop and data-centre server has a processor that runs its programs. Arm does not make those chips. It designs the blueprint they follow and the building blocks inside them, and sells the right to use them to chip makers such as Apple, Qualcomm and NVIDIA.
Partners shipped over 10 billion chips built on its designs in one quarter, and almost every smartphone uses one. In March 2026 Arm also began selling a server chip of its own, the Arm AGI CPU.
How it makes money
Arm is paid twice. A chip maker pays a licence fee up front to use a design, then a royalty on each chip it ships, often for decades. In the year to March 2026 licence fees were $2,307M and royalties $2,613M, and phones were about 43% of royalties. Its own server chip would add a third kind of revenue: the whole price of a chip.
Growth and profit
What growth costs
Growth has made Arm less profitable on the official measure. From fiscal 2022 to 2026 revenue rose 82%, but research grew from about a third to over half of revenue and stock pay from almost nothing to about a fifth, so the official operating margin fell from 23.4% to 18.3%. Only selling and admin costs shrank as a share.
Its record
The numbers
Is it growing?Steady
Sales grew 24% over the twelve months to Sep 2026, about the same pace as the 23% in the year to Mar 2026.
Is it profitable?Healthy
Operating margin is 17%, up from 3.4% in the year to Mar 2024.
Does growth pay off?Costly
Running costs take 81¢ of each sales dollar, up from 72¢ in the year to Mar 2022.
Does it earn good returns?Low
Each dollar put into the business earns 8.9% a year after tax, down from 15% in the year to Mar 2022.
Is it making cash?Strong
It turned 22% of sales into free cash, $1.21B over the last twelve months.
How steady is it?Uneven
Yearly sales growth has ranged from −0.9% to +24% over the last 4 years, with sales falling in 1 of them.
Are shares being diluted?Creeping
The share count rose 5.2% in 4.5 years, and pay in shares costs 23% of sales.
Why it wins
The evidence
Software for phones and servers is written for Arm's instruction set, the language a processor understands. Moving a product line to another means rewriting and retesting years of software, so chip makers stay. Arm has held over 99% of smartphone processors for years.
The US competition regulator called Arm's technology a critical input that rival chip makers rely on, because it is licensed to all of them on neutral terms.
Who could beat it
The rivals in full
RISC-V is a free, open alternative to Arm's designs, and many of Arm's own customers back it. In servers Intel and AMD still sell most processors. Qualcomm, about 9% of revenue, is suing Arm over its licences. And now that Arm sells its own chip, it competes with the customers who license its designs, who could move elsewhere.
Each rival
RISC-V, the open instruction set, and core designers such as SiFiveprivate
a chip maker adopts free RISC-V cores for controllers and then larger processors and stops paying royalties
a data centre buys AMD's processors rather than Arm's own AGI CPU
Cloud providers' own Arm chips: Graviton, Axion, Cobalt and NVIDIA's VeraAMZNGOOGMSFTNVDA
a buyer uses its cloud provider's or NVIDIA's Arm-based CPU instead of buying Arm's own
How big it could get
The market and who has signed
The market
Arm is already in almost every phone, and shipments are forecast to fall by about a sixth in 2026, so growth there comes from higher royalty rates. Servers are where units can still rise: a researcher puts Arm's share of server processors at just over a tenth in 2025 and above a fifth in 2026.
Nobody publishes how much chip makers spend on processor designs in total, so the market Arm sells into cannot be counted from outside.
Who runs it
Leaders, pay and the cash
Who runs it
Rene Haas has run Arm since February 2022, after leading its product group from 2017; he also sits on SoftBank's board and runs a SoftBank unit. Jason Child has been finance chief since November 2022.
Directors and managers own well under 1% of the shares. Haas was paid $60.6M for fiscal 2026, mostly in shares, and bonuses pay on revenue and on a profit measure that leaves stock pay out, so pay rewards growth more than profit per share.
The last report
The latest report
Arm's revenue rose 27% in its September quarter, but royalties grew only 8%
What it did to the thesis Solid quarter
It supported the business: licence fees and data-centre royalties grew fast and the next guide is higher. It did not ease our worries: royalties slowed, cash fell, and related parties still drive the fees. Nothing in it closes the gap to the price.
The report
In the quarter to September 2026 revenue rose 27% to $1,443M. Licence fees rose 50% and royalties 8%; management says data-centre royalties doubled. Official operating profit was $201M, 13.9% of revenue, because stock pay is counted; management's own measure, which leaves it out, was 44.2%.
Cash from operations fell to $207M from $567M a year earlier, as fees owed by Arm China and SoftBank's affiliate built up unpaid. The guide for the next quarter is $1.54B. The shares fell 13.1% the next day.
What's happening now
The background
The shares are 61% below their high of the past year: they fell 13.1% after the September-quarter report and 8.1% more on 16 November in a chip sell-off. Revenue per share kept rising, so the price per dollar of sales has fallen, but from a very high level.
The next test is the December-quarter report, due about 3 February 2027, then the first deliveries of Arm's own server chip, promised by the end of March 2027.
Where we disagree
Both sides in full
The market
The blueprint of AI data centres, with its own chip about to add billions: most analysts rate it a buy.
We believe
A rare franchise whose stock pay and related-party revenue leave owners thin profit, priced far above our best case.
The argument
Most analysts value Arm on management's profit, which leaves out stock pay, and expect revenue to nearly triple by fiscal 2030 as its own chip adds billions. Counted as a cost, stock pay leaves too little profit for that path to reach today's price.
The evidence against us: if the server chip reaches management's plan and stock pay falls fast, profit could grow into the price sooner than we model.
What it's worth
How we value it
Valued on the analysts' revenue path for three years, which already carries the AGI CPU plan, each later year closing 30% of the gap to 4%, on official operating profit with stock pay a cost at today's share of sales. Arm is not a young company, so stock pay is not faded; the thesis's own evidence is the check on that path.
The three cases
Bear case$16.18
Growth at half the analysts' path, 16%, 17% and 21%, the margin a fifth lower at 13.2%, and new capital earning only the 8% rate: the downturn test, phones falling and the AGI CPU slipping.
Our value$49.01
Revenue grows 32%, 35% and 43% over three years, the analysts' path with the AGI CPU in it (C20), then fades toward 4%; the margin moves to 16.5%, the analysts' fiscal 2029 margin less stock pay at today's 23% of sales and their adjusted-basis gap.
Bull case$74.20
Growth a quarter above the analysts' path, 40%, 43% and 53%, the margin a tenth higher at 18.2%, and new capital earning 15%.
The bet
Seen as a bet
This bet is on Arm's own server chip, the Arm AGI CPU, which Meta and other buyers have ordered. Management plans up to $2B of sales over two years, then about $2B and $5B a year.
Valued alone, the licence and royalty business does not cover today's price, so part of it already rests on the chip. Even if the chip reaches management's plan, the shares would be worth less than today, and if it fails more is lost: a bet priced for success, and expensive.
What could go wrong
Every risk
The risks
Related-party revenue
About a third of revenue comes from Arm China, which Arm does not control, and SoftBank's affiliate, and their fees are slow to arrive as cash.
Own chip, own customers
Selling its own server chip puts Arm against the companies that license its designs, who may move away.
A free rival
RISC-V costs nothing to use, and many of Arm's largest customers back it.
Qualcomm's lawsuit
A customer bringing about 9% of revenue is suing Arm over its licences.
SoftBank's control
SoftBank owns most of the shares, has pledged most of them for a loan, and elects the board.
What we still do not know
How much profit the server chip will earn once it ships, and whether management's plan holds.
Whether Arm China and SoftBank's affiliate keep paying on time, and how long their fees last.
How fast stock pay falls as a share of revenue.
What changes our view
Every route to a new view
Every route to a new view
→ Neutral
The shares fall back within 20% above what we think the business is worth.
→ Attractive
The shares fall to 80% of what we think the business is worth.
→ Neutral
Official operating margin reaches 35% over four quarters. Revenue growth over four quarters stays at 25% or more.
The next test
It supports the case if
Revenue of $1.54B or more, royalty growth back above 10%, outside-customer licence revenue growing, and first AGI CPU deliveries confirmed for the quarter.
It pressures the case if
Revenue below $1.49B, royalties flat or falling, or the AGI CPU slipping past March 2027.
We would also be wrong if
Royalties grow below 10% a year in fiscal 2027 and 2028.
Arm China falls below a tenth of revenue, or stops paying.
A major phone or cloud customer moves its main processors to RISC-V.
The verdict
Scores and how we weighed it
The scores
Growth
3 of 4
Business quality
4 of 4
Financial position
4 of 4
Execution record
3 of 4
Per share capture
2 of 4
Potential
3 of 4
Risk profile
2 of 4
Valuation
1 of 4
Evidence quality
3 of 4
How we weighed it
ValueNegative
The price sits far above our value and above our bull case.
BusinessPositive
Software built for Arm keeps phone and device makers, and cloud providers are moving their servers to it.
GrowthMixed
Data centre royalties are growing fast, but phones are shrinking this year and the price needs more than even the analysts expect.
RisksNegative
Arm China and SoftBank's affiliate are large customers, the largest phone licensee is suing, and stock pay keeps owners' profit thin.
RivalsMixed
No one licenses CPU designs at its scale, but free RISC-V is spreading and its own chip competes with licensees.
The groupNegative
It ranks with the AI chip names, which have sold off together on doubts about AI spending.
Latest newsMixed
The last report beat its guide on licences, but royalties grew slowly and the shares fell on the day.
In plain words
Our estimate of value
We forecast how much cash the business can earn over the next ten years and beyond, then ask what that stream is worth today to an investor who wants an 8% yearly return. We do it three times: a gloomy future, an expected one and a sunny one. We call a share Attractive only when its price is at least 20% below the expected estimate, and Unattractive when it is 20% above.