Qualcomm

IThe business

Qualcomm earns its money in two ways. It sells chips, and it licenses patents. The chip business is QCT, which made $37.2B of revenue and $10.1B of profit before tax in fiscal 2026. The licensing business is QTL, which made $5.6B of revenue and $4.0B of profit before tax.³ A third segment, QSI, invests in start-ups and earned $794M, mostly from gains. A set of small segments that includes the new Data Center business, the government unit and Modular lost $607M on $476M of revenue.³ Qualcomm’s fiscal year ends on the last Sunday of September, so fiscal 2026 ran to 27 September 2026.

It is based in San Diego, was founded in 1985 and has about 57,000 workers in 38 countries. Its chips are made by other companies; only some radio-frequency parts are built in its own plants.³ Customers are named by where they are headquartered: China including Hong Kong was 42% of fiscal 2026 revenue, the United States 25% and South Korea 22% R (reported).³ Apple and Samsung each paid it 10% or more of consolidated revenue.³

The chip business: three markets, one in decline

QCT sells to three markets. Handsets are the original business: the Snapdragon system chips and modems in Android phones, plus radio parts. Handset revenue was $24.1B in fiscal 2026, 56% of the company, and it fell 13% from $27.8B. In the fourth quarter it fell 26% to $5.1B.³² Two reasons are given. Customers cut back build plans because of the memory shortage and higher memory prices. And Apple now uses its own modem in some phones; the 10-K expects Apple to use mainly its own modem from here, with a significant negative effect on QCT revenue.³ The June 10-Q adds that Apple buys the thin modem alone, a lower-revenue and lower-margin product than the combined modem and processor Qualcomm sells to Android makers.⁹ It does not say how much of handset revenue Apple is. I can’t size the loss from the filings.

Automotive is the bright line. Revenue was $6.0B in fiscal 2026, up 52%, and $2.0B in the fourth quarter alone, up 89%; the company calls it its first $2B quarter.³² Three years ago it was $1.9B a year. The 10-K attributes the rise to higher prices per unit from mix and to more shipments from new car launches using its cockpit and driver-assistance chips.³ IoT, the third market, grew 8% to $7.1B. It covers PCs built on Snapdragon X, smart glasses and other personal devices, and industrial, networking and robotics chips.³

The licensing business

QTL licenses the patents Qualcomm holds on 3G, 4G and 5G. Phone makers pay a royalty on each phone, usually a percentage of the wholesale price with a cap or minimum on some products.³ The costs are small. QTL reported $1.5B of costs on $5.6B of revenue, and its margin was 72% in fiscal 2026 for the third year in a row.³ QTL revenue has hardly moved: $5.3B in fiscal 2023, $5.6B in 2024, 2025 and 2026, even as chip revenue swung by $8B.^1012113 The 4 November 2026 release says Qualcomm renewed long-term licences with Apple, Huawei and Transsion. I did not find the terms in the filings I opened, so I can’t say whether the royalty rate went up, down or stayed the same.²

Data center: the new line

The newest line is the data center. In fiscal 2026 Qualcomm bought Alphawave, which makes high-speed wired-connection technology, for $2.3B, paid with 11M shares and $301M in cash.³ On 28 July 2026 it closed Modular, an AI software company, for an accounting price of $2.6B, paid with 18M shares.³¹³ It also bought eight smaller businesses for $1.1B.³ Its products are called Dragonfly, and it says shipments will begin to ramp from the first quarter of fiscal 2027 under custom silicon agreements with large cloud buyers.³ The 4 November release announced a multi-generational product collaboration with Amazon for AI data centers, and in September Qualcomm gave an Amazon affiliate a warrant for up to 25M shares, described in Sections IV and VI.²⁴ Management’s target is $5B of Data Center revenue in fiscal 2027, which it reiterated on 4 November.² For scale, my Broadcom page records that company’s AI chip revenue at $16.7B in its latest quarter.¹⁴ The fiscal 2026 base for Qualcomm’s data center is small: all the segments outside the big three had $476M of revenue.³

Where the profit comes from

The segment table answers a question the headline numbers hide. QCT is 86% of revenue and 68% of segment profit. QTL is 13% of revenue and 27%.³ But the company does not charge stock pay to any segment, and stock pay is the largest cost outside them. In fiscal 2026 it was $3.5B, or 8.0% of revenue.³ If I charge all unallocated costs, stock pay included, to the chip side where the engineers work, QCT’s share of operating profit falls to $4.6B, and QTL’s $4.0B is 46% of the $8.7B total E (estimate). If I charge the unallocated costs in proportion to each segment’s own costs, QTL’s share is 44% E (estimate). Either way, about 45% of what is left after stock pay comes from a business that is 13% of revenue and runs on licence terms, not product cycles. That is why the licence renewals matter more than their headline.

The latest quarter

The fourth quarter of fiscal 2026 was the weakest of the year. Revenue was $10.3B, down 8%, and operating income was $1.4B, down 54% from $2.9B.² Operating income by quarter went $3.4B, $2.3B, $1.6B and $1.4B, a margin of 27% in the first quarter and 13% in the fourth C (calculated).⁸² The fourth quarter carried $235M of restructuring. Handsets fell 26%. Automotive rose 89%. The company guided the first quarter of fiscal 2027 to revenue of $10.4B to $11.2B, with $9.0B to $9.6B from QCT and $1.3B to $1.5B from QTL, and earnings per share of $1.03 to $1.23, or $2.15 to $2.35 on the company’s own adjusted basis.² It also said it is raising prices to offset higher costs for wafers, packaging, memory and other inputs, and expects the full benefit in gross margin only in the second half of fiscal 2027.² The stock closed at $173.19 on 4 November, the day of the release, $156.75 on 5 November, down 9.5%, and $171.47 on 9 November C (calculated).¹⁵

The 2017 to 2019 detour

A reader who looks only at ten years of earnings will see a hole in fiscal 2018. The cause was a set of fights and one deal. In fiscal 2017 the company says actions by Apple and its contract manufacturers cut its revenue, and Korea fined it about $927M.¹⁷ In fiscal 2018 the European Commission fined it $1.2B, a decision the EU General Court annulled in fiscal 2022, and it paid NXP a $2.0B termination fee after China’s regulator had not approved the planned purchase by the deadline.¹⁸¹⁰ In April 2019 it settled with Apple, booked $4.7B of licensing revenue, and signed a six-year licence with an option for two more years and a chip supply agreement.¹⁵ That licence is the one the November 2026 release says was renewed.² Operating income went from $2.6B in fiscal 2017 to $0.6B in 2018 and $7.7B in 2019.⁸

IIThe moat

Qualcomm has two kinds of protection, and they are not alike. One is a pile of patents that hundreds of companies license. The other is a position in chips that customers are free to leave.

The patents

The 10-K says Qualcomm holds a leading portfolio of patents on 4G and 5G and that the portfolio is the most widely and extensively licensed in the industry. It has licensed or otherwise provided rights to hundreds of companies, and it has told standards bodies it will offer licences on the essential ones.³ The vast majority of QTL revenue comes from sales of 4G and 5G devices, mostly smartphones.³ The margin on the licences, 72% in each of the last three years, shows what the market will bear.³ This is the part of the business a competitor can’t copy by spending money.

The protection is a legal one, and the law has been tested. Korea fined Qualcomm about $927M in fiscal 2017, Taiwan fined it $778M, and the European Commission fined it $1.2B in fiscal 2018.¹⁷¹⁸ The fines did not all stand. Taiwan settled in fiscal 2018 and Qualcomm reversed $676M of the $778M, the EU General Court annulled the Commission’s decision in fiscal 2022 and Qualcomm booked a $1.1B benefit, and a Korean court in December 2019 upheld part of that decision and cancelled the rest, with appeals then pending.¹⁸¹⁶¹⁰ Apple and its contract manufacturers then fought it for two years, until the April 2019 settlement.¹⁵ The risk factors still list efforts by some phone makers to avoid paying what Qualcomm calls fair royalties, proposals at standards bodies to reduce the value of essential patents, and the chance that governments change its licensing practices.³ They also say a drop in licensing would hit the company more than a similar drop in chips, because licensing is where the margin is.³ That is true of my model too: a $1B a year loss of QTL profit costs me about $9 a share E (estimate).

The licences also end. The Apple agreement signed in April 2019 ran six years with an option for two more, to March 2027 at the latest.¹⁵ The 4 November 2026 release says it is renewed, along with Huawei and Transsion.² I could not find the terms. Until a filing gives the royalty rate or the length, I treat QTL revenue as flat and I treat the renewals as a strength of the position and not as a number.

The chips

The chip business has no law behind it. Customers choose Qualcomm chips because they are good and because the modem, processor, radio parts and AI engine come together and work with the standards. They can also stop. The 10-K names Apple, Samsung and Xiaomi as customers that have built their own chips, and says it expects them to use their own more over time.³ It warns that supply shortages make that more likely, because customers want control of their supply.³ Apple and Samsung each accounted for 10% or more of revenue in fiscal 2026. Xiaomi was above 10% in fiscal 2025 and is not named in 2026.¹¹³ A business whose two largest buyers are also its competitors has a lead that it has to keep paying for, and the R&D line shows it: $10.2B in fiscal 2026, up 13% while revenue fell 3%.³

The dispute with Arm

Qualcomm licenses the Arm instruction set for its products, and the dispute with Arm is the main legal matter in the 10-K.³ In August 2022 Arm sued, saying that after Qualcomm bought Nuvia it breached Nuvia’s licence and had to stop using and destroy the technology, which Arm says includes Qualcomm’s custom Oryon processor cores. A jury sided with Qualcomm in December 2024, the court entered final judgment for Qualcomm on 30 September 2025, and Arm appealed the next day. The Third Circuit heard the appeal on 21 October 2026 and has not ruled.³

Qualcomm also sued Arm. That trial began on 5 October 2026, and on 13 October the jury found that Arm breached Qualcomm’s licence by failing to provide certain deliverables and improperly interfered with Qualcomm’s relationship with a customer, but did not breach the technology licence as Qualcomm had claimed. The judge has not yet decided the claims and remedies reserved to the court.³ In October 2024 Arm gave notice alleging a breach of Qualcomm’s licence and asserting a right to terminate it; it withdrew the notice in January 2025 and said it has no current plan to terminate, while reserving its rights.³ Qualcomm has recorded no loss for any of this.³ Arm and QTL earn the same kind of money, a royalty on products other companies make and ship. The difference is that the 10-K says Arm has begun to supply chips itself and so competes with its licensees, and that Qualcomm’s ability to get Arm’s latest technology on acceptable terms and in time has been and may continue to be affected.³ I can’t forecast the Third Circuit or the judge. I show what it would cost in the risk section.

IIIThe competitive advantage

The moat section says where Qualcomm is protected. This section is about what it does better than the people it competes with.

The first thing is breadth. It builds the modem, the application processor, the radio parts, Wi-Fi and Bluetooth, the AI engine and power management, and it ships them as one platform with software.³ A phone maker who buys the platform gets chips that were designed together for the same radio standards, which cuts the work to bring a phone to market. The second is the scale of the spending behind it: research and development was $10.2B in fiscal 2026, 24% of revenue, against 20% a year earlier.³ The third is the licensing business, which gets paid again on the same phone. A handset with a Qualcomm chip may carry two Qualcomm revenue streams, and the royalty is not hurt by a chip price war.

The advantage is shown in the margins, and the margins have slipped. Gross margin was 53% in fiscal 2026 against 55% in 2025.³ For QCT alone, cost of revenues was $19.5B on $37.2B of revenue, a gross margin of 47.7% against 49.7% in fiscal 2025 and 49.9% in 2024 C (calculated).³ QCT profit before tax was 27.1% of revenue against 30.4% a year earlier.³ The company says part of the squeeze is input cost: wafers, packaging, advanced test, memory and other materials are all up, and it is raising prices to recover them, with the full effect due in the second half of fiscal 2027.³² I can’t tell from the filings whether price rises will stick. That depends on whether customers have other places to go, and for the biggest customers that question is open.

Evidence the advantage is widening

• Automotive revenue went from $1.9B to $6.0B in three years and was $2.0B in the fourth quarter alone.¹⁰³²
• QTL held its revenue and margin for three years while handset chips fell, and the 4 November release reports renewals with Apple, Huawei and Transsion.³²
• Qualcomm sells Snapdragon X chips for PCs and builds chips for smart glasses, XR and robotics.³
• Amazon made initial purchase commitments for its server chips, enough to vest 3,750,000 of the 25M warrant shares on the day the warrant was issued.⁴
• A jury sided with it in Arm’s case in December 2024, and sided with it on two of its three claims in its own case in October 2026.³

Evidence it is narrowing

• Handset revenue fell 26% in the fourth quarter, and the 10-K expects memory shortages and higher memory prices to hurt several handset makers in the coming quarters.³²
• The 10-K expects Apple to use mainly its own modem from here, and says Samsung and Xiaomi build their own chips.³
• Gross margin fell two points in a year, and QCT’s EBT margin fell from 30.4% to 27.1%.³ The fourth quarter QCT margin was 23.8% C (calculated).²
• Inventory rose $2.1B to $8.6B and purchase commitments are $18.6B, so a drop in demand would land on the balance sheet.³
• Arm competes with licensees and has made its latest technology harder to get on acceptable terms.³
• The data-center business is about 1% of fiscal 2026 revenue and has no profit yet: the small segments lost $607M.³

Where Qualcomm wins and where it loses

Qualcomm wins where a customer needs a leading Android platform quickly and cannot afford to do it alone: the premium Android phone, the cockpit in a new car, the smart device that must run AI on a battery. It loses where the customer is large enough to make the chip itself, which is Apple and increasingly Samsung. It will have to prove itself in the data center, where the buyers are few, large, and used to making their own chips. The licensing side wins by default. It is the one place where the advantage does not depend on this year’s chip.

IVManagement

Who runs it

Cristiano Amon, 56, is President and Chief Executive. He joined Qualcomm as an engineer in 1995, ran the Snapdragon product roadmap, became head of the chip business, and has been chief executive since June 2021.³ Akash Palkhiwala is chief financial officer and, since January 2024, chief operating officer. Alex Rogers, 69, has led QTL since 2016 and joined as a litigator in 2001. Baaziz Achour, who joined in 1993, became chief technology officer in February 2025.³ The team is long-serving and comes from the chip and licensing businesses. That is a strength where the business is the same. It is a question where the plan is a data-center business the company has never run.

How they are paid, and what that pays them to do

Mr Amon’s pay for fiscal 2025 was $29.7M: $1.35M of salary, $24.2M of stock awards, $3.2M of cash bonus and $1.0M of other pay. It was $25.9M in fiscal 2024 and $23.5M in 2023.⁶ The proxy says 61% of his target pay depends on company performance, and that at least 90% of target pay for the named executives is tied to financial or stock results.⁶ The annual cash bonus is paid on one-year adjusted revenue and adjusted operating income. The performance shares, 60% of the annual award, are split evenly between three-year adjusted earnings per share and the company’s share return against the Nasdaq-100. The remaining 40% is time-based stock that vests over three years.⁶

Three things follow. First, “adjusted” means the company’s own measure, which leaves out stock pay and the amortization of acquired technology.² Stock pay was $3.5B in fiscal 2026, the largest cost the executives are not paid on. Second, the same measure leaves out the amortization of technology bought in acquisitions, $423M in fiscal 2026, so a company that buys its growth is not charged for the purchase in the pay measure. Third, the share-return half of the performance award is a real test: if the share return is negative over three years, that half pays no more than target.⁶ This is my reading of the definitions, not a statement by the board.

The record shows the effect is not one-sided. The performance shares for fiscal 2023 to 2025 paid 66% of target, because the share return of 37% ranked in the 44th percentile of the Nasdaq-100 and adjusted earnings per share averaged $9.98.⁶ In fiscal 2026 the company also replaced the annual cash bonus for its non-executive leadership team with a two-year stock award for fiscal 2026 and 2027, part of a $669M rise in stock pay: $428M in research, $229M in selling and administration and $12M in cost of revenues.³ That keeps cash in the company and puts the cost in the line I deduct.

How much of their own money is in

Mr Amon owned 198,063 shares on 31 October 2026, worth $34.0M at $171.47, about 25 times his salary against a requirement of 10 times.⁷⁶ Only shares owned outright count toward that rule. All 18 executive officers and directors together owned 573,776 shares in December 2025, under 1% of the company.⁶ The proxy puts Vanguard at 10.5% and BlackRock at 8.7%, but those figures come from Schedule 13G filings dated May 2024 and December 2023.⁶ Newer filings show Vanguard Capital Management at 7.5% on 31 March 2026, after Vanguard began reporting its units separately, and State Street at 5.0% on 30 June 2026; I found no BlackRock filing since January 2024.²³²⁴

Alphabet’s 13F filing for the quarter ended 30 September 2026 lists 2,976,982 Qualcomm shares worth $547.9M on that day, held through GV, its venture arm. None of its 51 earlier 13F filings, back to 2013, lists Qualcomm.²⁵ That does not mean Alphabet chose to buy. Qualcomm’s July prospectus lists two GV funds, both former Modular stockholders, as holding 3,196,261 shares that were part of the payment for Modular. The 13F count is 219,279 shares lower, and the filing does not say why.²⁶ I read it as deal consideration and take no view from it.

The Forms 4 show a different pattern from the ownership rule. Seven officers and directors sold 126,960 shares for $22.4M from early November 2025 to 9 November 2026, and I found no purchase on the market by any of them.⁷ Mr Amon sold 40,000 shares for $7.6M in four lots of 10,000 at $180, $185, $195 and $200, under a Rule 10b5-1 plan he adopted on 12 December 2025.⁷ A planned sale at a set price says little about his view. But I found no insider purchase at $124, the low close of April 2026, or since.

What they did with capital over a full cycle

Over fiscal 2017 to 2026 Qualcomm made $57.1B of owner cash, which is cash from operations less capital spending and less stock pay C (calculated).⁸ It paid out $92.0B: $58.4B in buybacks and $33.6B in dividends C (calculated).⁸ That is 161% of owner cash. Before stock pay it generated $76.2B of free cash flow, so it still paid out 121% of that C (calculated). The gap came from the cash pile it started with, $38.6B of cash and securities against $21.9B of debt at the end of fiscal 2017, and it ended fiscal 2026 with $7.7B against $15.1B C (calculated).⁸³ A company that pays out more than it earns for ten years has used up its cushion once, and the next ten years have to be paid for out of what it earns.

Most of the buyback money went out in one place. In fiscal 2018 Qualcomm spent $22.6B, on a $5.1B tender offer and $16.0B of accelerated buybacks among other purchases, and in fiscal 2019 it took delivery of another 96M shares. Together that is about 375M shares at an average near $65.¹⁸¹⁶ The count fell from 1,474M shares at the end of fiscal 2017 to 1,072M now, down 27%.⁸³ Seen from today, the large buybacks of 2018 and 2019 paid off: the shares were bought at about $65 and are worth $171. The buybacks of the last two years were not. Qualcomm bought 104M shares for $16.6B at an average near $160, which is 1.6 times my base value of $98 C (calculated).¹¹³ At my base value, each of those dollars bought about 61 cents of value E (estimate).

The count does not tell the whole story, because Qualcomm also issues shares. In fiscal 2026 it repurchased 48M shares for $7.8B and issued 46M, so the outstanding count fell from 1,074M to 1,072M, two million shares.³ About 29M of the issued shares went to the sellers of Alphawave and Modular and the rest to employees. A buyback that is replaced by new shares is a transfer, not a return. Cash buybacks and stock-paid deals have to be read together: in fiscal 2026 the cash leaving was $7.8B for buybacks and $1.6B for acquisitions and other investments, while the shares issued for deals were worth about $4.4B at the time and appear nowhere in the cash flow statement C (calculated).³

The record on deals is mixed. The 2016 plan to acquire NXP ended in fiscal 2018 with a $2.0B fee paid to NXP.¹⁸ The Nuvia purchase, whose designs Qualcomm uses in its Oryon cores, also produced the Arm lawsuit.³ The fiscal 2026 deals are too new to judge. Of the 18M Modular shares, 4M went to certain executives who must stay four years, and $342M of that is charged as stock pay over the period, on top of the $2.6B price.³ They cost $6.0B in price, 1.2 times the year’s owner cash, and they sit in segments that lost $607M last year.³ What I can say is that this management wants the data-center business enough to pay for it in stock before it has shown a profit.

The Amazon warrant

In September 2026 Qualcomm gave an affiliate of Amazon a warrant for up to 25M shares at $161.26 until 3 September 2036.⁴ It vests as Amazon signs commercial arrangements, places binding orders and buys server chips, systems and manufacturing services, up to $60B of payments. 3,750,000 shares vested on issue.⁴ The company valued the warrant at $71.57 a share, $1.8B in all, and records it as a reduction of revenue as the related sales are recognized.³ The price at which it can be exercised is 6% below today’s $171.47, so the warrant is $10.21 a share in the money, or $255M for all 25M shares. The 25M shares are 2.3% of the shares outstanding C (calculated). At the grant-date value, the $1.8B is 3% of the $60B of purchases it is tied to C (calculated). If Amazon buys the full $60B, Qualcomm sells chips at a discount that grows with the share price; if it buys little, the warrant mostly lapses. I count it as a cost in each case, at the company’s grant-date value of $71.57 a share for the shares I assume vest: 4M in my bear, 10M in my base and all 25M in my bull, or $0.3B, $0.7B and $1.8B E (estimate).

VFinancials

Ten years of numbers tell one story: a business that doubled its revenue, made far more in the middle than at either end, and has paid out more than it earned.

What an owner could really take out

Over ten years Qualcomm reported $81.6B of operating income, which is already after stock pay. The cash an owner could have taken, after capital spending and after stock pay, was $57.1B, or 70% of operating income C (calculated). Net income is no guide: it was $65.6B over the decade, but it includes a $5.4B tax charge in 2018, a $7.1B tax expense in 2025 that holds a $5.7B valuation allowance, and a $4.2B tax benefit in 2026 that reverses it.⁸³

In fiscal 2026 owner cash was $5.2B, down from $10.0B in 2025. Operating cash flow was $11.1B. Capital spending doubled to $2.5B, 5.8% of revenue against 2.7%, to build engineering and chip-test capacity. Stock pay took $3.5B.³ Two items made the year look worse than the engine underneath. Inventory took $2.0B of cash as the company bought supply to protect itself from shortages, and the last installment of the one-time repatriation tax, $663M, was paid.³ If neither recurs, owner cash would have been about $7.9B E (estimate). Neither is certain. Inventory may take a year to sell, and the 10-K expects capital spending to rise again.³

The gap between the company’s earnings and owner cash is the whole argument about this stock. The company’s adjusted earnings per share were $10.54. Take out stock pay, take out all capital spending, and count the inventory it paid for, and the cash an owner could take was $4.85 a share. Here is the price on each measure.

I value the company on the owner-cash lines. The adjusted line is the company’s own, and the one its pay plans use. At 16 times adjusted earnings the multiple looks modest. At 35 times what an owner could take out, or 23 times if I am generous with the two one-offs, it does not. Stock pay is $3.5B and rising: the first-quarter outlook takes $0.99 a share off GAAP earnings for stock pay, about $1.06B after tax and $1.2B before tax, or roughly $4.8B a year E (estimate).² The company’s measures leave that out. I deduct it.

Quality of earnings

• Taxes. The reported tax line swung from a $7.1B expense in 2025 to a $4.2B benefit in 2026 because of one accounting step, the valuation allowance on $5.7B of deferred tax assets. Excluding it, the 2026 rate was about 16.6% against 11.0% in 2025 C (calculated), because the foreign-derived income deduction was worth less after the new tax law.³ Cash taxes paid were $1.6B in 2026 and $3.1B in 2025.⁸ The 10-K expects lower cash taxes ahead from expensing research costs.³
• Stock pay. It was $3.5B in 2026 and $2.8B in 2025. The company also paid $1.1B in cash to cover employees’ tax on vesting shares, which is a financing outflow; I count stock pay once, as an expense, and do not deduct that payment again.³ Unrecognized stock pay on awards already granted is $4.1B, to be expensed over 1.9 years.³ It ran $830M to $888M a quarter through fiscal 2026, and the first-quarter outlook implies about $1.2B; if that proves a one-quarter spike my $4.8B for fiscal 2027 is too high, and each $1B is worth about $9 a share.⁸²
• Investment gains. Pretax income includes $794M from the strategic-investment segment, mostly gains. I leave it out of operating income and carry the investments at their $2.0B book value.³
• One-offs. Restructuring cost $332M in 2026, almost all severance. A payment to a customer under a legal settlement cut reported revenue by $120M, and the adjusted figures leave it out.³²
• Customer incentives. Liabilities for customer incentives and other customer items rose from $1.9B to $2.6B.³ They are paid to customers as a reduction of revenue, and the Amazon warrant adds a new one.⁴
• Interest. Interest expense was $689M. Interest and dividend income fell from $639M to $424M as cash was spent.³

The balance sheet

On 27 September 2026 Qualcomm had $4.6B of cash and $3.0B of marketable securities, $7.7B in all, against $15.1B of debt, so net debt of $7.4B C (calculated). All of the debt is fixed-rate notes, plus $499M of commercial paper. $2.0B of the notes, 3.25% paper, is due in May 2027, and a $4.0B revolving credit line, undrawn, runs to August 2029.³ On 9 November 2026 the company filed a preliminary prospectus for five new series of senior notes, with the size not yet stated. It says the money will repay those 2027 notes, with the rest for general corporate purposes that may include paying down commercial paper, working capital and acquisitions.²⁷ Operating income covers interest 12.6 times C (calculated). Equity is $30.4B, of which goodwill is $16.4B.³ Purchase commitments to suppliers are $18.6B, $13.8B of it due in the next twelve months.³ The company has $19.6B left under its buyback authorization, which has no end date.³ At the dividend rate of $0.92 a quarter the yield is 2.1% C (calculated).² Total cash returned in fiscal 2026 was $11.7B, which is 2.2 times reported owner cash and 1.5 times my normalized figure C (calculated).² That is only possible while the balance sheet supports it, and the cash pile that supported it is mostly gone.

VIValuation

First, how this goes badly wrong

Before the numbers I asked what would most likely hurt an owner here, and worked back. Three things stand out.

Phone chips shrink faster than the rest grows, while stock pay rises. Handsets are 56% of the company and fell 26% in the last quarter. The loss to Apple is not sized in the filings. If handset revenue is $3B a year below my base, value falls about $6 a share. Stock pay is the quiet part: if it runs $1B a year above my base, value falls about $9 a share E (estimate).

The data center costs more than it earns. The company has put $6.0B of price into acquisitions in a year, most of it for this business, and the small segments lost $607M on $476M of revenue.³ On my base the data center is worth about $21 a share. If it earns nothing and the money is lost, I lose that and the $6.0B already spent. A custom-silicon business also depends on a few buyers who can design their own chips or go elsewhere, and the Amazon warrant is a discount that grows with the share price.³⁴

The royalty stream is cut. It is 13% of revenue and about 45% of profit after stock pay. If its profit before tax were $1B a year lower for good, value would fall about $9 a share E (estimate). The paths to that are a renewal on worse terms than I can see, a licensee winning a dispute, a regulator changing the rules, or the Arm case going badly enough to hurt the products the royalty is paid on. I do not put a probability on any of them.

None of these needs anything exotic. They set my bear case, and they are why I start from the cash that reaches an owner and not from the company’s adjusted earnings.

What an owner could really take out

I value cash after stock pay and after the capital needed to stand still. Operating income is already after stock pay. From it I take cash tax, add back depreciation and amortization, and subtract capital spending and the working capital that growth ties up. The tax line is low at first, 12% in the next three years, because the 10-K says expensing research costs lowers cash taxes, and rises to 16% by 2032.³ Capital spending is 6% of chip and licensing revenue in 2027 and 2028, 5% in 2029 and 2030 and 4.5% after that, plus 3% of Data Center revenue E (estimate). Stock pay is a dollar amount: $4.8B in 2027, rising 2% a year in the base. Amortization of acquired technology is a cost in operating income and is added back as the non-cash item it is.

On this basis owner cash is about $4.8B in fiscal 2027, against $5.2B reported for 2026, and $8.5B in 2030 E (estimate). The price needs $14.5B in 2030. That is 1.4 times the best year so far, fiscal 2025, which made $10.0B C (calculated).

What the price implies

Put the other way, what does $171.47 need? At a 10% discount rate and 3% growth after fiscal 2036, the price needs owner cash of about $14.5B in 2030, 1.70 times my base for every year, and $18.3B in 2036. In operating income after stock pay, that means about $17.6B in 2030. The record is $15.9B, in fiscal 2022, on revenue of $44.2B. The alternative way to read it is through the data center alone. Holding everything else at my base, the price needs Data Center revenue of about $43B in fiscal 2030, which is 4.5 times my base and 8.6 times the company’s $5B target for fiscal 2027.² For scale, my Broadcom page records AI chip revenue of $16.7B in a single quarter.¹⁴

What I can’t judge

• How much Amazon will buy, and at what margin. The warrant ties 21.25M of its 25M shares to purchases of up to $60B. I do not know the price, the margin or the timing of those purchases.⁴
• What Apple is now worth. The filings do not say how much of handset revenue, or of QTL, Apple is. They say the modem loss is significant, and the new licence terms are not public.³²
• How long the memory squeeze lasts. It cut handset shipments in fiscal 2026 and the company expects more of it.³
• Where data-center margins settle. The segment lost money in 2026. My 25% before stock pay is a guess, not a measurement.
• The Arm decisions. I have read the filing’s account. I can’t forecast the Third Circuit or the judge.

Because I can’t know these, I show a range and the price’s position in it, and I don’t treat $98 as more exact than it is. A fair reading of my work is that the base is somewhere between $80 and $120.

The discount rate

I ask 10.0%. The capital asset pricing model starts from the 4.88% ten-year Treasury yield on 9 November and Damodaran’s 4.20% implied equity risk premium of 1 October, the figure my other pages use; his 1 November figure is 4.41%.²⁸²⁹ Qualcomm’s betas on month-end price returns against the S&P 500 are 1.52 over ten years, 1.70 over five and 2.50 over three, which give 11.3%, 12.0% and 15.4% C (calculated).¹³⁰ So 10.0% is below what the model gives. At 11.3% my base is $81, my bear $37 and my bull $137, and the price is 111% above the base E (estimate). I keep 10.0% for the reason I give on my NVIDIA page: I would rather use one rate for a large business and put the risk in the cash flows, where the bear case shows it. Qualcomm’s royalty stream is steadier than a chip, and its chip revenue is more cyclical and more concentrated than Apple’s. The rate is the same as I use for NVIDIA, two points under Marvell and Micron, half a point under Broadcom, and above the 9% I use for Apple.^2214213132 At 9% my base is $117 and at 11% it is $85.

Three cases

Bear

Apple’s exit and a long memory squeeze take handset chip revenue to $13.5B by fiscal 2030, from $24.1B. The data center stalls at about $3B and Amazon orders little. The chip margin settles near 21%, lower than any year from fiscal 2021 to 2026. Royalties erode to $5.1B at a 70% margin. Stock pay falls only to $4.4B. Operating income after stock pay is about $5.1B in 2030, against $7.8B in fiscal 2023 and $2.6B in 2017. Value $44.

Base

Handsets fall to $19.5B in fiscal 2027 and $17.5B in 2029, stay there in 2030, then grow about 1% a year. Automotive goes from $6.0B to $8.0B in 2027 and $11.0B in 2030, and IoT grows about 6% a year. Non-handset revenue, my count of automotive, IoT and the data center, is $19.1B in 2027, up 40% from $13.6B, where the company projects more than 60%. The data center reaches $3.5B in 2027, 30% under the company’s target, and $9.5B in 2030, at a 25% margin before stock pay. The chip margin recovers from 22.5% to 27%. Royalties stay near $5.6B to $5.7B. Stock pay is $4.8B and rises 2% a year. Operating income after stock pay is $6.3B in 2027 and $10.8B in 2030. Value $98.

Bull

The company meets its plan and goes on: $5B of data-center revenue in fiscal 2027 and $17B in 2030, a 32% margin before stock pay, and Amazon buys the full $60B. Automotive reaches $13.8B in 2030, IoT $10.6B, and handsets stay near $20.5B for three years. Non-handset revenue is $21.7B in 2027, up 59%, just under what the company projects. The chip margin reaches 29%. Operating income after stock pay is $16.7B in 2030, above the record of $15.9B. Value $168.

Other uses of the same dollar

The closest comparison on this site is Broadcom, which sells the custom AI chips Qualcomm wants to sell. At its price I get 7.5% a year on Broadcom’s base cash flows against the 10.5% I ask, a shortfall of 3.0 points, with the value after year ten struck at that 10.5%.¹⁴ Done the same way, Qualcomm at $171.47 returns 3.0% against 10.0%, a shortfall of 7.0 points. The 7.2% I quote beside it strikes the value after 2036 at the return itself, which is how my Apple page reads its price; it is the kinder reading when a price is far above value. Among the chip pages here the best return at its price is NVIDIA, at 9.3% against 10.0%.²² My AMD page gives −1.6% on the same footing as NVIDIA, so on this basis Qualcomm sits sixth of the seven, ahead only of AMD.²⁰ By my numbers none of them clears its bar at today’s price. By price against value Qualcomm is also further out than most, 75% above its base against 25% for Broadcom and 5% for NVIDIA, though each page is on its own basis. That is not a verdict on Qualcomm. It says that with these chip pages I keep finding prices above my value, and they are not independent: most of them rest on the same AI build-out. Qualcomm’s data-center plan is one more bet on that, with a smaller base under it.

The company’s own alternative is its buyback. At $160, the average paid in the last two fiscal years, against my base value of $98, each dollar bought about 61 cents of value E (estimate).¹¹³ The next-best use of cash inside the company is the data-center build itself, and I have shown that it is worth about $21 a share if it works as I assume.

What the market paid before

The market has paid between 16 and 47 times a year’s owner cash for Qualcomm at the end of its fiscal years. The median was about 23. At 35 times fiscal 2026’s, the price sits near the high end, but owner cash was depressed in 2026 by one-offs. On my normalized figure it is 23 times, the median. That tells me the market is not making an unusual bet on 2026. It is paying a normal multiple for a recovery, and the recovery is the question.

The margin of safety

I set it from the bear case: I want a price at which my $44 bear value is at least three quarters of what I pay, which is $59. The price is 2.9 times that. The last close below $59 was $57.18 on 15 April 2019, the day before the Apple settlement lifted the shares 23%.¹ I am not saying the price will fall to $59. I am saying that is the price at which a buyer is paid for the chance that the data center and the phone chips both disappoint.

Two different questions

The first is how good the business is. By the numbers it is good: a 72% margin on licensing, a car business that grew 52% last year, $10.2B a year of research, and owner cash of $57B over ten years. The second is how good the price is. At $171.47 the price is 75% above my base value and 2% above my bull, and the bull has the data center working. Those are different answers. A good business can be a poor purchase when the price already counts the recovery and the new business. This looks like that case. The strongest form of the market’s side is that I am overcharging for stock pay and undercounting the data center, and Section VII gives that case in full.

VIIThe case for the price

What most people miss

Three facts cut in Qualcomm’s favour and get little notice. First, the royalty business is bigger in profit than its size suggests: $4.0B before tax at a 72% margin, about 45% of operating profit after stock pay by my count, and the 4 November release says Apple, Huawei and Transsion have renewed.³² Second, the car business has already moved: $2.0B in the fourth quarter, an annual rate of $8B against $2.9B in fiscal 2024.¹²² Third, cash taxes: Qualcomm paid $1.6B in fiscal 2026 on $9.4B of pretax income and carries $5.6B of net deferred tax assets that it expects to use.⁸³ I assume low cash taxes for four years, not the whole $5.6B.

The bull case

I found no statement by a named investor or analyst on Qualcomm that I could check against a dated document, so I give the case as the company makes it. Altimeter Capital Management, the fund Brad Gerstner founded, reported 1,882,490 Qualcomm shares worth $347.9M at 30 June 2026, in a 13F filed on 14 August. None of its four earlier filings, back to June 2025, lists Qualcomm. A 13F shows a holding, not a reason, and I did not find Mr Gerstner commenting on Qualcomm.³⁴ On 4 November 2026 Mr Amon said fiscal 2027 starts a non-handset inflection, with non-handset revenue projected to grow by more than 60% and a $5B Data Center target reiterated.² On my count fiscal 2026 non-handset revenue was about $13.6B (automotive, IoT and the small segments), so 60% means at least $21.8B; the release does not define the term C (calculated). The 10-K says Data Center shipments begin to ramp in the first quarter under custom-silicon agreements with large cloud buyers, and the Amazon warrant vests as Amazon buys, up to $60B of payments.³⁴

In plain words: phone chips shrink, the royalty stream holds, and the newer businesses grow fast enough to replace the phone profit and then add to it.

Against reality

If they do what they say and hit their targets, the price is possible. That is my bull case, $168, and the price is 2% above it. It needs $17B of data-center revenue by fiscal 2030 at a 32% margin before stock pay, the full Amazon warrant vesting, automotive at $13.8B, the chip margin back at 29%, and handsets holding near $20.5B. Operating income after stock pay has to go from $8.7B to $16.7B in four years, above the $15.9B record.

Today’s numbers are a long way from that. Fourth-quarter operating income was $1.4B, an annual rate of $5.4B. Handsets fell 26%. The data center and its small neighbours lost $607M on $476M of revenue. Stock pay is rising. Each could change, but the price already assumes they will.

It helps to multiply it out. The bull case needs five things: the data center near $17B by fiscal 2030, perhaps 35%; a margin of 30% or more, 60%; automotive near $14B, 55%; handsets near $20B, 45%; royalties and the Arm cases staying benign, 85%. Multiplied, that is about 4%, or one chance in 23. They are my guesses, they are not independent, and I do not use them in the values E (estimate).

What a 10x would take. Ten times $171.47 is $1,715, or $1.8 trillion of market value on today’s shares. At 25 times owner cash, a rich multiple for a chip company, that needs about $74B of owner cash, which is 7.3 times the best year so far. If owner cash were 25% of revenue, better than the best year’s 23%, that means about $294B of revenue, 6.8 times fiscal 2026, or 21% growth a year for ten years. Qualcomm’s revenue grew 7.6% a year from fiscal 2017 to 2026. The share price has risen 2.5 times in ten years, 9.7% a year before dividends. A tenfold price is not plausible for a company that starts at this size, and I would say so about any chip company that does.

Where I come down

Three things moved me: the royalty business holding its revenue for four years and its margin for three, and being renewed, the car business reaching $2B a quarter, and low cash taxes. They are why my bear is $44 and not lower. Two things did not. The data center is still $476M of revenue across small segments, and handset revenue fell 26% in the quarter. What would raise my values is a quarter with Data Center revenue above $1.5B at a margin of 25% or more, stock pay held under $4.5B a year, or evidence that Apple’s loss is smaller than I fear. Until then the price sits above my bull case.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• China. Customers headquartered in China, including Hong Kong, were 42% of fiscal 2026 revenue. The 10-K says trade tension, national-security policy and rules on moving money out of China can affect sales and the timing of payments.³
• Customers and vertical integration. Apple and Samsung each pay 10% or more of revenue and both build their own chips. A third may follow.³
• Supply and cost. Qualcomm relies on outside factories and on memory and packaging suppliers, whose costs are rising. It has $18.6B of purchase commitments, and if demand drops it could be stuck with inventory or fees.³
• Law and regulators. The company has been fined in Korea, Taiwan and the European Union, though the Taiwan and EU fines were later largely reversed, and the 10-K says licensing rules and standards-body policies could change.^1718103
• Arm. If a court limited what Qualcomm may do with its own cores, or if Arm made new licences hard to get, the products that carry the royalty and chip revenue could be hurt.³
• Taxes. By releasing a $5.7B valuation allowance in fiscal 2026 the company now counts on future taxable income to use its deferred tax assets, and on how the new US rules are applied.³
• Deals. Goodwill is now $16.4B, 54% of equity, and the company says it expects more strategic deals.³