What we said in October — and what happened
The toll held, the chips slipped, the level was never reached, and six details were wrong
On 10 October we published Qualcomm at $175.50. Our thesis was that it is two businesses. One is a toll: a royalty on almost every phone that connects to a mobile network, which Apple had just renewed even as it removed Qualcomm's modem from its iPhones. The other is a chip business losing its best customer, squeezed by memory costs that had cut Chinese phone production, and betting on cars and data centres to replace the loss. We valued the whole at about $160 a share, told readers to wait, and to buy below $140. We said to watch three things: the jury verdict in Qualcomm's case against Arm, the November results, and whether the chip margin would climb back from 26%.
The shares rose to $185.41 on 29 October, then fell with the chip sector. On 20 November they closed at $149.94, 14.6% below our price. The basket of chip shares we compare them with fell 33.8% over the same weeks. The lowest close, $149.94, came on the last day, and $140 was never reached. Here is the account.
| What we said on 10 Oct | What happened by 20 Nov | Grade |
|---|---|---|
| 'Wait; buy below $140' | Low close $149.94 (20 Nov), lowest print $148.25 (5 Nov). A buyer at $175.50 is 14.6% down; one who waited has not bought. | ✓ Waiting was right |
| 'The toll survives'; Apple's renewal 'extends that confidence into the 2030s' | Apple, Huawei and Transsion renewed; the finance chief said Qualcomm 'held the rates'; no major phone-maker renewals until fiscal 2030. | ✓ Right, stronger |
| Watch 'the Arm verdict next week' | On 13 October the jury found Arm in breach of Qualcomm's licence and liable for interfering with a customer, with $1 of damages per claim. The judge has not yet ruled on the remedy. | ◆ Won, value open |
| 'Whether chip margins climb back from 26%' | They fell to 24% in the fourth quarter and are guided to 20–22% in the first quarter of fiscal 2027. | ✗ The other way |
| November results 'will show the first full quarter of the Apple step-down' | Wrong quarter. Apple's step-down runs from the September to the December quarter, which Qualcomm reports in February. | ✗ Wrong quarter |
| Cars 'already a real success' | Automotive revenue $1.99bn, +89%, Qualcomm's first $2bn quarter in cars. | ✓ Right, better |
| Owner earnings $6.70 a share, 'a trough'; $8.50 'a normal year' | Fiscal 2026 owner earnings were $4.85 a share. The trough was lower than we said. | ✗ Too high |
| Dividends + buybacks 126% of free cash flow; net debt $2.3bn → $7.0bn | For the full year: 135%; net debt $7.4bn; and $4.0bn of new bonds sold on 9 November. | ✓ Right, worse |
Our errors. Re-checking the October report against the 10-K, the releases and our price history, we found six details that were wrong. First, we used about 1,050m shares, the count on 27 July, the day before Qualcomm issued 18m shares to buy Modular. The count at 27 September was 1,072m, so the market value at $175.50 was about $188bn, not $184bn, and the toll was worth about $59–65 a share, not $60–67. Second, we wrote that Huawei had stopped paying when its licence expired and that its licence was "gone". On 5 October, four days before we published, Huawei and Qualcomm announced a multi-year patent licence. Qualcomm now lists Huawei among its renewals. Huawei no longer buys Qualcomm's chips, which is the part we had right. Third, we said the shares had passed through $140 "in both April and July". They traded below $140 only between February and April. The July low was $147.61 at the close and $142.89 during the day. Fourth, trailing GAAP earnings per share were $8.64, not $8.76. Our feed carries the wrong figure for the September 2025 quarter (fault five in Part IX). Fifth, we dated the holdings of Vanguard (10.5%) and BlackRock (8.7%) to December 2025. The proxy dates them to May 2024 and December 2023, and Vanguard has since split its reporting. Sixth, we said Modular cost "~$3.1bn in stock". The 10-K gives an accounting price of $2.6bn, plus $342m of the shares paid to Modular's leaders, which will be expensed as pay over four years.
We also rewrote the first lines of the October Letter, which used Warren Buffett's 1991 words about technology as if they were our own. They are his, and they are now quoted as his. Two judgements deserve grading as well. We named the wrong quarter for Apple's exit, so the hardest numbers are still to come. And we treated $6.70 of owner earnings as a trough, when the full year came in at $4.85. Everything else we re-checked was right: the segment figures, the margins, the Apple concentration, the Amazon warrant, the 2029 targets, the dividend history, the chief executive's sales, the debt and the values our model gave.
★ Down 15%, while chips fell a third
A rise into the results, three heavy days, and a level never reached
| Date | What happened | QCOM that day |
|---|---|---|
| 13 Oct | The Arm jury found for Qualcomm on two of three claims, with nominal damages (10-K; Bloomberg Government). | +1.4% |
| 29 Oct | The high since our X-Ray, 5.6% above our price; no specific news. | $185.41 |
| 30 Oct | Intel's weak forecast set off a chip sell-off; the basket fell 8.2% (as reported). | −11.1% |
| 4 Nov | Starlink chose Qualcomm for its US 5G network, announced that morning; results after the close. | +5.3% |
| 5 Nov | Profit guidance below forecasts and a 20–22% chip margin, on a day when the chip basket fell 12% (Reuters, CNBC). | −9.5% |
| 18 Nov | No cause found; the chip basket rose. | −4.1% |
| 20 Nov | A broad AI sell-off: the S&P 500 −3.7%, the chip index −10.7% (AP). | −7.5%, $149.94 |
A chip-sector fall, mostly. Most of Qualcomm's worst days were the sector's worst days. The basket of chip shares lost a third of its value between our X-Ray and 20 November, as investors began to doubt that spending on AI data centres would continue at its recent pace. Qualcomm, which earns little from that spending today, fell less than half as much. Only one of its bad days was about its own numbers: 5 November, after the results, when the guidance for the December quarter came in below what analysts had expected. Even then the sector fell further.
Why the price held up better than the business. That is the uncomfortable point of this follow-up. In the weeks since October, Qualcomm's licensing business became more secure. Its chip business reported lower margins, lower cash flow and higher stock pay than we had assumed. The market seems to have treated Qualcomm as a phone company hiding from an AI sell-off. That may be a fair view of the next quarter. It does not change what the business earns. Our estimate of its value fell by more than its price did.
Where the price sits. At $149.94 the shares are 7.4% below the $161.96 average Qualcomm paid for its own shares during fiscal 2026, 7% below the price at which Amazon's warrant can be exercised, 25% below the $200 at which the chief executive last sold, and 40% below the May high. The market value is about $160bn.
★★ The toll held; the chips slipped
Cars up 89%, phones down 26%, and a chip margin guided to 20–22%
On 4 November Qualcomm reported the quarter to 27 September. Adjusted revenue was $10.44bn, down 7%, inside its own range; adjusted earnings per share were $2.17, down 28%, also inside the range. Within the chip business, revenue from phone chips fell 26% to $5.14bn, as Chinese phone makers kept cutting production because of memory shortages and Apple bought fewer modems. Revenue from cars rose 89% to $1.99bn, and from other devices (IoT) 4%. Licensing revenue fell 7%, with a margin of 70%.
| Fiscal 2026 | Q4 (to 27 Sep) | Full year | vs a year earlier |
|---|---|---|---|
| Revenue (adjusted) | $10.44bn | $43.24bn | −7% / −2% |
| Phone chips / cars / IoT | $5.14bn / $1.99bn / $1.89bn | $24.07bn / $6.00bn / $7.13bn | Q4: −26% / +89% / +4% |
| Chip margin (QCT, before tax) | 24% | 27% | 29% / 30% |
| Licensing revenue / margin | $1.30bn / 70% | $5.56bn / 72% | 72% / 72% |
| EPS: adjusted / GAAP | $2.17 / $1.17 | $10.54 / $12.70 | GAAP year includes a $5.32 tax benefit |
| Free cash flow / stock pay | $1.82bn / $0.87bn | $8.65bn / $3.45bn | FCF −33%; stock pay +24% |
| Q1 fiscal 2027 guide: revenue / adjusted EPS | $10.4–11.2bn / $2.15–2.35 | Chip margin 20–22% | About −12% / −36% |
The chip margin. In October we said to watch whether the chip margin climbed back from 26% as Qualcomm's price rises took hold. It went from 26% to 24%, and the guide for the December quarter is 20–22%. About 3 points of that come from moving the loss-making data-centre business into the chip segment, so the like-for-like figure is about 23–25%. Management says the benefit of its price rises will come "in the second half of fiscal '27", and it has kept its long-term target of 30%. It also said that custom data-centre chips are "dilutive to gross margins". So the cleanest part of the chip business, phones, is shrinking, and the fastest-growing part earns less on each dollar of sales.
The cash. Free cash flow for the year fell 33% to $8.65bn. Part of the fall is temporary: a final $663m payment of the 2018 tax on overseas profits, and a $2.0bn increase in inventory, which management says was to secure supply in a year of shortages. Part of it is not: capital spending more than doubled to $2.5bn, and the company expects it "to increase from fiscal 2026 levels in the near term". Stock-based pay rose 24% to $3.45bn, partly because Qualcomm replaced two years of cash bonuses for its broader leadership with a share award. Free cash flow less stock pay, our owner earnings, was $5.20bn, or $4.85 a share, against $9.08 a year earlier.
The guidance. For the December quarter Qualcomm expects revenue of $10.4–11.2bn and adjusted earnings per share of $2.15–2.35, down about 36% from a year earlier. Phone-chip revenue is expected to fall again, "as growth in Android revenues is offset by the continued ramp-down in Apple product revenue". For fiscal 2027 the plan is that revenue outside phones grows by more than 60%, that the data-centre business reaches $5bn, and that this more than replaces Apple. The guide also implies that stock pay in the quarter rises to about $1.2bn before tax, about 38% more than in the September quarter (our arithmetic from the guidance).
★ Three renewals and a $2 verdict
The licensing business is safer than we said; the Arm case is won but not yet worth anything
The renewals. Qualcomm's fourth-quarter release headlined "Renewed Long-term QTL License Agreements" with Apple, Huawei and Transsion. Apple's we knew about: it begins in April 2027 and was the reason we called the toll durable. Huawei's is new to us, though not to the market: it was announced on 5 October, a cross-licence covering 5G, computing, AI and networking, with Qualcomm also buying some of Huawei's US patents. Transsion is the largest seller of phones in Africa. Asked on the call whether the new terms were worse, the finance chief, Akash Palkhiwala, said "you should take it as we were able to hold the rates and the royalties". The chief executive, Cristiano Amon, said the next renewal cycle runs from fiscal 2030 to 2033.
This matters more than any quarter. The toll is the part of Qualcomm that customers cannot avoid, and in October we valued it at about a third of the share price. A licensing business whose largest customers have renewed at the same rates, with no major renewal due for three or four years, is about as predictable as a technology business gets. We raised our confidence in it, not our estimate of what it earns: licensing earned $4.0bn before tax in fiscal 2026, the same as the year before, at 72%.
| Matter | What happened | What it is worth |
|---|---|---|
| Apple licence | Renewed from April 2027, while Apple 'will primarily use its own modem products' (10-K). | ✓ Royalties kept |
| Huawei licence | Multi-year cross-licence announced 5 October; financial terms not disclosed. | ✓ Back in the toll |
| Transsion licence | Renewed (Q4 release). | ✓ Renewed |
| Qualcomm v. Arm (Delaware) | 13 Oct: Arm breached Qualcomm's licence and interfered with a customer relationship; did not breach the second licence. Damages $1 per claim. Remedy, including a possible five-year suspension of royalties, not yet decided (10-K). | ◆ Liability won; value open |
| Arm v. Qualcomm (Nuvia), appeal | Argued before the Third Circuit on 21 October; no ruling (10-K). | ◆ Open |
The Arm verdict. Qualcomm designs its own processor cores under licences from Arm, which tried in 2024 to cancel one of them. In October a jury in Delaware found that Arm had broken its main licence with Qualcomm by failing to deliver what it promised, and had improperly interfered with Qualcomm's relationship with a customer, reported to be Meta (Bloomberg Government). It awarded $1 of damages on each claim. What matters is a clause in the licence that, according to analysts at Futurum, could suspend Qualcomm's royalties to Arm for five years after an uncured breach, which Futurum puts at about $300m a year (as reported). Arm argues that the clause cannot be enforced. The judge has heard the arguments and not ruled. We count the verdict as a win in principle and value it at nothing until there is a ruling. In the meantime the chief executive says it has "no implications on our CPU roadmap".
The data centre becomes a business
A $5bn target, two large customers, a warrant and a loss
In October we called Qualcomm's data-centre plans "an option, not yet a business". The fourth-quarter call made them more concrete. Qualcomm expects more than $650m of data-centre revenue in the December quarter and $5bn in fiscal 2027, a target first set at its investor day in June, which our October report did not mention. Management said "the majority of the $5 billion in fiscal '27 come from the two hyperscaler customers, so one of them being Amazon"; custom chips for Amazon began shipping in October, and a processor for Meta is "on track". On 4 November Starlink chose Qualcomm's equipment for its US 5G network, revenue the company describes as more of a fiscal-2028 story.
Three cautions temper that. First, the business lost money: the segments outside phones, licensing, cars and devices, which are mostly the data centre, lost $607m before tax in fiscal 2026 on $476m of revenue. Second, part of the price of Amazon's business is paid in Qualcomm shares. The warrant over up to 25m shares at $161.26 has a fair value of $1.8bn, which Qualcomm will deduct from revenue as Amazon's purchases are recognised; about 4m shares had vested. Third, custom chips for a handful of giant buyers earn lower gross margins than phone chips, as management said. So the data-centre business can grow quickly and still do little for owner earnings for several years.
| Data centre | Now | Read |
|---|---|---|
| Revenue guide, Q1 fiscal 2027 | > $650m | ✓ Above forecasts (Reuters) |
| Target, fiscal 2027 / 2029 | $5bn / > $15bn | ◆ Mostly two customers |
| Loss before tax, segments outside the core (fiscal 2026) | −$607m on $476m of revenue | ✗ Not yet profitable |
| Amazon warrant | Up to 25m shares at $161.26; $1.8bn fair value deducted from revenue | ◆ Paid partly in shares |
| Gross margin effect | 'Accretive to our operating margins, but dilutive to gross margins' | ◆ Lower quality revenue |
What we would need to see. Revenue of $650m in the December quarter, a loss that narrows rather than widens, and a customer other than Amazon and Meta paying in cash. If those come through, the 2029 plan becomes believable and our central value rises. If the data-centre revenue arrives with a larger loss, it will be growth that owners pay for rather than growth that pays owners.
The moat, rechecked
Deeper around the patents; still shallow around the chips
| The claim (October) | Evidence now | Verdict |
|---|---|---|
| The patent toll: every phone pays | Apple, Huawei and Transsion renewed; rates held; no major renewals until fiscal 2030. | ✓ Deeper |
| Premium Android chips: Snapdragon in most flagships | The chief executive says Snapdragon's share of Samsung's flagships is now 'north of 70%' as a baseline. | ✓ Holding |
| Customers building their own chips | Apple 'will primarily use its own modem products… going forward' (10-K). Chinese makers' own chips continue. | ✗ As feared |
| Cars: design wins locked in for years | Revenue +89% in the quarter, +52% in the year; run-rate near $8bn a year (call). | ✓ Widening |
We keep the moat at 7, with a change of weight. The toll, which we had already rated highly, is now more secure: three renewals, the rates held, and a long gap before the next ones. The car business has the kind of advantage that comes from being designed into a model years before it is sold, and those designs are now turning into revenue. Against that, the core phone-chip business has no moat against its own customers. Apple has left, and the 10-K now says so plainly. Every large customer that can afford to build its own chip is trying to. That is why the toll and the cars deserve high multiples and the phone chips do not.
- That owner earnings recover to about $7 a share and then grow about 8% a year for ten years, against $4.85 in fiscal 2026.
- That the chip margin returns towards 30% as price rises take hold, even as lower-margin data-centre chips become a larger share of sales.
- That stock pay stops growing faster than revenue, so that the buyback starts to shrink the share count again.
The competition
Customers building their own chips, and giants in the data centre
| Rival | Where it meets Qualcomm | The latest | Read |
|---|---|---|---|
| Apple (own modems) | Modems in iPhones. | Primarily its own modems 'going forward' (10-K); keeps paying royalties under the renewed licence. | ✗ Chip lost, toll kept |
| Samsung (Exynos) | Processors in Galaxy phones. | The chief executive calls a Snapdragon share above 70% 'the new baseline with Samsung'. | ✓ Holding |
| MediaTek | Mid-range and increasingly premium Android. | No new figures since October; the larger seller by units. | ◆ Steady pressure |
| Chinese phone makers' own chips | Xiaomi and others. | China-based customers were 42% of fiscal-2026 revenue, from 46%. | ◆ Slowly leaving |
| Nvidia, Broadcom and others | AI chips and custom silicon in data centres. | Qualcomm is a new entrant with two large customers. | ◆ Long odds, big prize |
The pattern has not changed. Qualcomm's chip customers are mostly companies large enough to make their own chips, and over time most of them try. Apple has done it, Samsung keeps trying, and Chinese phone makers are building their own. Qualcomm keeps its place by staying a generation ahead in performance and in modems, and it is still doing that at the top of the Android market. Cars are different: the customers are less able to build chips themselves, and they choose suppliers for many years at a time. That is why the car business is the most valuable part of the chip business, and the one that has grown fastest.
A word on the phone market. Qualcomm expects the handset market to have shrunk by "a low teens to mid-teens" percentage in 2026, because memory shortages raised costs and cut production, and to be "flat to slightly down" in 2027. An analyst at Bernstein called that view possibly "aggressive" (Reuters). If phone production recovers in 2027, Qualcomm's chip numbers will look better than its guidance. If memory stays scarce, they will not.
Management, insiders & capital allocation
Buybacks that barely shrank the shares, stock pay rising fast, and new debt
Where the cash went. In fiscal 2026 Qualcomm generated $8.65bn of free cash flow. It paid $3.85bn of dividends and spent $7.82bn buying back 48m shares, at an average of $161.96. That is $11.67bn, 135% of free cash flow. It also paid $1.13bn of tax on shares vesting to employees, which is a buyback by another name. The difference came from the balance sheet. Cash and securities fell from $12.5bn to $7.7bn, and net debt rose from $2.3bn to $7.4bn. On 9 November Qualcomm sold $4.0bn of new bonds, at 5.0–5.8%, mostly to repay $2.0bn of notes due in May 2027.
What the buyback bought. Very little reduction in the share count. Qualcomm repurchased 48m shares and issued 46m: about 29m to buy Alphawave and Modular, and the rest to employees. Shares outstanding fell by 2m, or 0.2%. Over ten years the diluted count has fallen 28%, so the long record is good. But in 2026 the buyback mostly offset acquisitions paid in shares and rising stock pay, and it was partly paid for with borrowed money. In October we said the company described its buybacks as offsetting dilution rather than as a judgement about price. The 2026 figures show what that means in practice.
Stock pay. Stock-based pay rose 24% to $3.45bn in fiscal 2026, 8% of revenue, and the guidance for the December quarter implies about $1.2bn in a single quarter. The 10-K says part of the rise comes from replacing cash bonuses for fiscal 2026 and 2027 with a two-year share award for the broader leadership. Paying people in shares rather than cash flatters free cash flow. That is why we subtract stock pay from it.
Insiders. Since our X-Ray, executives have sold about 11,600 shares under pre-arranged plans, about $1.9m, and nobody has bought. No officer or director has bought a share on the market since October 2025.
The numbers
A tax gain in the earnings, a halving in owner earnings, and six faults in our feed
Fiscal 2026, to 27 September. GAAP revenue was $43.12bn, down 3%. GAAP earnings per share were $12.70, but $5.32 of that was a one-off tax gain, when Qualcomm released a $5.7bn reserve against tax assets after new guidance from the IRS. Without it GAAP earnings were about $7.38 a share. Adjusted earnings, which exclude stock pay and other items, were $10.54, down 12%. Operating cash flow was $11.15bn, capital spending $2.49bn and free cash flow $8.65bn. After $3.45bn of stock pay, owner earnings were $5.20bn, $4.85 a share. Excluding the final $663m payment of the 2018 tax on overseas profits, they were about $5.46.
The balance sheet. At 27 September Qualcomm had $4.64bn of cash and $3.05bn of marketable securities against $15.11bn of debt: net debt of $7.43bn, about half a year of operating profit before depreciation. The bonds are rated A2 and A, with stable outlooks. Inventories rose 32% to $8.64bn, and the company has $18.6bn of commitments to buy supply, $13.8bn of it within a year. Both are large against a business whose phone revenue is falling, and both are worth watching.
Earnings quality. Three things flatter the reported numbers: the tax gain in GAAP earnings; adjusted earnings that leave out $3.45bn of stock pay; and, from fiscal 2027, a data-centre business whose largest customer is paid partly in warrants. One thing depresses them: the inventory build and the tax payment in this year's cash flow. On balance we treat about $7 a share as a normal year of owner earnings for fiscal 2027: analysts' adjusted earnings forecast of about $9.68, less stock pay after tax, plus some inventory coming back as cash. That is below the $8.50 we used in October.
| What the screen shows | What is wrong, and the number to use |
|---|---|
| Stock pay, Q4 fiscal 2026: −$15m | The filed figure is <b>$873m</b>. The annual total ($3.45bn) and the first three quarters are right. |
| Net debt $11.36bn | Adds $880m of leases and ignores $3.05bn of marketable securities. Net debt is <b>$7.43bn</b>. Flagged in October; still unfixed. |
| Dividends: latest declaration September | The $0.92 declared on 21 October (payable 17 December) is missing, and the dividend summary on the overview is empty. |
| Revenue by segment, fiscal 2026: $89.9bn | Old and new segment rows added together. Revenue was <b>$43.12bn</b>: chips $37.21bn, licensing $5.56bn. Flagged in October; still unfixed. |
| EPS, September 2025 quarter: −$2.77 | The filed figure is <b>−$2.89</b>. This produced our own fourth error ($8.76 instead of $8.64). |
| Analysts: 30 buy, 35 hold, 5 sell | Barely changed since October and inconsistent with the feed's own monthly history (38 analysts on 1 November). Several targets were cut after the results. Treat as stale. |
★ The dividend
Safe, slow-growing, and paid by a company returning more cash than it makes — tested
On 21 October Qualcomm declared a quarterly dividend of $0.92, payable on 17 December, unchanged since the March raise of 3.4%. At four times $0.92 the yield is 2.45% at today's price, against 2.1% in October. The dividend took $3.85bn in fiscal 2026, 45% of free cash flow and about a third of adjusted earnings. Qualcomm normally raises it in March.
| Test | Qualcomm | Result |
|---|---|---|
| 1. Covered by earnings | 34% of fiscal-2026 adjusted EPS; 38% of the fiscal-2027 forecast; 49% of GAAP EPS without the tax gain | ✓ Pass |
| 2. Covered by free cash flow | 45% of fiscal-2026 free cash flow (2.2× cover); 74% after stock pay | ◆ Pass, thinner |
| 3. Growing | $2.02 (FY2016) → $3.62 (FY2026), 6.0% a year; the last three raises +6.25%, +4.7%, +3.4% | ◆ Slowing |
| 4. Never cut | The annual total has risen every fiscal year since at least 2007, though the quarterly rate was frozen from March 2018 to March 2020 | ✓ Pass |
| 5. Balance sheet behind it | Net debt $7.4bn, about half a year of operating profit before depreciation; rated A2/A | ✓ Pass, weaker |
| 6. Cash returned vs cash earned | Dividends + buybacks: 65% (FY23), 70% (FY24), 98% (FY25), 135% (FY26) of free cash flow | ✗ Fail |
The dividend is safe; the total payout is not sustainable. To threaten the dividend, free cash flow would have to fall below about $3.9bn, less than half of a weak year. That is very unlikely. What cannot go on is the combination: a dividend growing slowly, plus buybacks larger than the remaining cash, paid for with savings and new bonds. Something will give. We expect it to be the buyback rather than the dividend, and we would welcome that, because buying back shares with borrowed money at 31 times owner earnings is not good value for owners. We expect the March raise to be small, around 3–4%, in line with the last two.
★ Valuation
Cheaper on adjusted earnings, dearer on owner earnings
Our method. As in October, we value Qualcomm on owner earnings, free cash flow less stock pay, over ten years at 10% a year, with a multiple on the tenth year, and we leave out the net debt as we did then. Re-running October's assumptions reproduces our October values of $78, $160 and $249 exactly. What changed is the starting point. In October we assumed a normal year of $8.50 a share. Fiscal 2026 came in at $4.85, stock pay is rising, capital spending is rising, and the data-centre growth will earn thinner margins. We now start the central case at $7.
| Path | Owner earnings start | Growth, years 1–5 / 6–10 | Exit multiple | Value today |
|---|---|---|---|---|
| Phones shrink, the data centre slips | $5.50 | 1% / 1% | 11× | ~$61 |
| Central case ◆ | $7.00 | 7% / 4% | 15× | ~$127 |
| October's growth and multiple, lower start | $7.50 | 7% / 4% | 16× | ~$141 |
| The 2029 plan is met | $8.00 | 13% / 6% | 18× | ~$221 |
| What $149.94 implies | $7.00 | ~8% a year, constant | 15× | $149.94 |
What is priced in? At $149.94, starting from a normal $7 a share and a multiple of 15 on the tenth year, the price assumes owner earnings grow by about 8% a year for ten years. Starting from what fiscal 2026 actually produced, $4.85, it needs about 14%. That is roughly Qualcomm's own 2029 plan, which calls for $40bn of revenue outside phones and adjusted earnings above $18 a share. The plan may be met. Cars are ahead of it. But a price that requires the plan leaves no room for the parts that go wrong.
The toll and the rest. Another way to see it: licensing earns about $3.5bn a year after tax. At 15–20 times that is worth $49–65 a share. The rest of the price, about $85–100 a share, pays for a chip business whose earnings after stock pay were about $4.5 a share in fiscal 2026, roughly 20 times. That is a full price for a business whose margin is falling and whose largest customer has left.
Our level. Our central value is now about $127–141, down from $160, so the $140 we set in October would now be paying full value. We lower the level to $110: about 18% below the middle of our central range, a smaller discount than we would ask of a weaker business, because the toll gives a floor that few chip companies have. At $110 Qualcomm would trade at about 11 times next year's expected adjusted earnings and 16 times a normal year's owner earnings, with a dividend yield of 3.3%, and our central case would earn about 12% a year. The shares last traded near there at the end of fiscal 2023. Wait. Buy below $110.
Risks, lawsuits & controversies
Verified 22 November 2026 against the fiscal 2026 10-K, 8-Ks, Forms 4 and attributed reports
Legal. In Qualcomm v. Arm the jury found for Qualcomm on 13 October; the judge has not ruled on the remedy. Arm's appeal in the Nuvia case was argued before the Third Circuit on 21 October. The Federal Circuit dismissed ParkerVision's appeal on 30 September and sent the case back to the lower court. The fourth quarter included a $120m payment to a customer under a legal settlement, which Qualcomm excluded from its adjusted revenue.
Supply. Inventories of $8.6bn (+32%) and $18.6bn of purchase commitments are large for a business whose phone revenue is falling. If memory prices fall and phone production recovers, they will be used; if not, they could lead to write-downs.
| Test | Today | What would change our mind |
|---|---|---|
| Q1 fiscal 2027 results, early February | Guide: adjusted EPS $2.15–2.35 | The first full quarter without most of Apple's modems. Results at the top of the range with the chip margin at 22% would steady the case; below the range would lower it again. |
| Chip margin (QCT) | 24%; guided 20–22% | A return above 26% by the second half of fiscal 2027, as management expects, would lift our central value. |
| Data-centre revenue and loss | > $650m guided; −$607m loss in FY26 | Revenue on plan with a narrowing loss would make the 2029 plan believable. |
| Stock pay | $3.45bn; ~$1.2bn implied in Q1 | Stock pay falling back below 8% of revenue would raise owner earnings; above 10% would lower them. |
| The Arm remedy | Liability found; $2 of damages | A five-year suspension of royalties would be worth roughly $1.5bn over the period, as reported; we would add it to our value. |
| The price | $149.94 | Below $110 with the toll intact and the chip margin no worse than guided: buy. |
In October I told you that Qualcomm was a toll road with a chip factory attached, that I admired the toll and was wary of the factory, and that I would wait to buy until the shares fell below a hundred and forty dollars. Six weeks later they are at a hundred and fifty. They have fallen, but by much less than the rest of the chip industry, and they have not reached my price. A reader who waited has lost nothing. I have to tell you that the waiting has not made the shares more attractive.
The toll did better than I said. Apple renewed its licence, as we knew. Huawei, which I wrote had stopped paying, had signed a new licence four days before I published, and I missed it. Transsion renewed too. The finance chief says the rates were held, and the next big renewals are three or four years away. A business that the customer cannot route around, renewed by the customers with the most reason to try, is about as good as technology gets. I value it more securely than I did.
The factory did worse. I said to watch whether the chip margin would recover from twenty-six per cent; it fell to twenty-four and is guided lower. The cash the business keeps after paying its people in shares halved, to under five dollars a share. Stock pay rose by a quarter and is rising again. And the company paid out a third more than it earned, in dividends and buybacks, and borrowed the difference. The buybacks barely reduced the number of shares, because Qualcomm also issued shares to buy companies and to pay its staff.
There is real good news in the factory, and I do not want to bury it. Cars are now a two-billion-dollar-a-quarter business, growing almost ninety per cent. The data-centre business has two large customers and a target of five billion dollars next year. If both work, the company's plan for 2029 is within reach. But the price already assumes that plan. At a hundred and fifty dollars you pay about thirty times what the business kept for its owners last year, and about twenty times what I think it will keep in a normal year.
So my estimate of the business went down by more than the price did, and I will not pretend otherwise by keeping my old number. A hundred and forty dollars was a good price for a business worth a hundred and sixty. It is not a good price for one worth about a hundred and thirty. I would wait, and buy below a hundred and ten. I also got six details wrong in October, the most embarrassing being Huawei, and I have corrected them here.
— The Buffett Lens. When the best part of a company gets better and the largest part gets worse, value the parts separately, and do not let the good news pay for the bad.
Buy Below $110The toll is safer than we said; the chip business earns less than we assumed. ★ At $149.94 — 15.5× next year's adjusted earnings but ~31× fiscal-2026 owner earnings — the price is above our new central value of ~$127–141 (from ~$160), within a range of ~$61 (phones shrink, the data centre slips) to ~$221 (the 2029 plan is met). The October level of $140 now offers no margin of safety; wait, and buy below $110. Watch February's results, the chip margin, stock pay and the Arm remedy.