Built by Claude FableThematic Exposure (6/10)
Bottom line: Qualcomm has real oligopoly positions (premium Android SoC, cellular modem, auto digital cockpit, cellular licensing, smart glasses), so the gate is passed. In FQ3'26 the AI story showed up in revenue in only one place: auto, up 61% y/y. Data center is still pre-revenue, at roughly $0.1B a quarter. The core handset franchise is shrinking: down 20% y/y, Apple is leaving, and Exynos and MediaTek are taking share. What changed in FQ3'26 is better visibility (purchase orders, wafer starts, the BMW ADAS win, a higher auto run rate), not a reacceleration in AI revenue.
About 19% of FQ3'26 revenue is AI-driven, and only about 1% is pure AI (data center). If you count premium Android phones sold on on-device AI, the figure is about 60%, but that revenue is not growing because of AI today.
| Bucket | % of revenue | AI link |
|---|---|---|
| Data center | ~1% (<a href="https://daloopa.com/src/174834428">$88M</a>) | Pure AI, but mostly the acquired Alphawave business |
| Automotive | ~16% | AI compute for cockpit and driver-assist; +61% y/y |
| AI PC (Snapdragon X) | ~2% (est.) | Sold as an AI PC; revenue not disclosed |
| AI-driven subtotal | ~19% | |
| Premium Android handsets | ~40% (est.) | Sold on on-device AI, but AI is not adding revenue yet |
| Other handsets, IoT, licensing | ~40% | Little or none |
- Auto is about 85% of the AI-driven bucket, and its growth is not yet split into AI content per car versus volume or price, so 19% is a generous count.
- Data center is what moves this number: management targets $5B in FY27 (about 11% of today's revenue base) and $15B in FY29. The first test is data center revenue in the December quarter.
- Premium-handset, AI PC and IoT splits are estimates; Qualcomm does not break them out.
| Segment | FQ3'25 | FQ4'25 | FQ1'26 | FQ2'26 | FQ3'26 | FQ3'26 y/y | FQ3'26 % of rev |
|---|---|---|---|---|---|---|---|
| QCT Handsets | 6,328 | 6,962 | 7,824 | 6,024 | 5,086 | -19.6% | 51.1% |
| QCT Automotive | 984 | 1,053 | 1,101 | 1,326 | 1,588 | +61.4% | 16.0% |
| QCT IoT (incl. PC, XR, industrial) | 1,681 | 1,806 | 1,688 | 1,726 | 1,830 | +8.9% | 18.4% |
| QTL Licensing | 1,318 | 1,410 | 1,592 | 1,382 | 1,278 | -3.0% | 12.8% |
| Total revenue | 10,365 | 11,270 | 12,252 | 10,599 | 9,947 | -4.0% | 100% |
| Data center rev (disclosed) | n/a | n/a | n/a | 97 | 88 | new | ~0.9% |
Y/y growth trajectory (computed from the Daloopa figures above and prior-year quarters FQ3'24 handset 5,899, FQ4'24 6,097, FQ1'25 7,574, FQ2'25 6,929; auto 811, 898, 961, 959; IoT 1,359, 1,683, 1,549, 1,581):
| y/y growth | FQ3'25 | FQ4'25 | FQ1'26 | FQ2'26 | FQ3'26 |
|---|---|---|---|---|---|
| Handsets | +7.3% | +14.2% | +3.3% | -13.1% | -19.6% |
| Automotive | +21.3% | +17.3% | +14.6% | +38.3% | +61.4% |
| IoT | +23.7% | +7.3% | +9.0% | +9.2% | +8.9% |
| Auto + IoT (non-handset) | +22.8% | +10.8% | +11.1% | +20.2% | +28.3% |
Annual (Daloopa): handsets $16.8B FY21, $25.0B FY22, $22.6B FY23, $24.9B FY24, $27.8B FY25. Auto $1.0B to $1.4B to $1.9B to $2.9B to $4.0B (FY21 to FY25, about a 42% CAGR). IoT $6.6B FY25, up 22%. QTL $5.6B FY25, flat for three years. FY25 total revenue was $44.3B, with handsets at 63%. In FQ3'26 handsets were 51% of total revenue, and non-handset was 40% of QCT (vs 28% in FY25).
Mix is shifting fast. Roughly half of that shift comes from handsets shrinking and half from non-handset growing.
| Segment | % of Revenue | Market Share | TAM | Theme Growth |
|---|---|---|---|---|
| Premium Android handset SoC | ~40% (est., most of the handset line) | Leader. Est. majority of premium Android SoC value. ~70% of Samsung Galaxy flagships (FQ3'26 call); 28% of all premium chipsets incl. Apple (Counterpoint, 2026) | Premium tier is ~half of Android SoC revenue (Counterpoint 2024) | Negative near term: QCT Android revenue down ~20% FY26 (CFO); OEMs moving to prior-gen chips. Agentic phones are still a forecast |
| Total smartphone AP | 51% (handsets) | #2: 23% in CQ2'26 vs MediaTek 31%, Apple 19%, Exynos 9% (Counterpoint) | Handset units down low teens in FY26 (CFO) | Declining; share down from 27% |
| Cellular modem (incl. Apple) | Apple product rev ~$7.5B FY26 (Rasgon calc, mgmt "fair range") | ~50%+ of baseband revenue historically. Apple share "materially lower" than 20% in fall 2026 iPhones, then ~0 | Baseband market rev +22% y/y in CQ1'25 on 5G mix (TechInsights) | Negative: Apple C1/C1X replaces QCOM. Apple revenue down ~50% q/q into the Dec quarter |
| Automotive (Digital Chassis, ADAS) | 16% | Leader in digital cockpit (est. majority of premium cockpit SoCs); "on track to be #1 auto semi by revenue." ADAS challenger | Design-win pipeline $65B (Investor Day); $10B FY29 target | +61% y/y FQ3'26; ~+60% guided for FQ4; exit run rate raised from $6B to $7B |
| IoT / edge AI / industrial | ~13% (ex-PC/XR, est.) | Fragmented; QCOM is one of several (NXP, MediaTek, Nvidia Jetson, Intel) | Industrial networking/robotics $8B FY29 target; pipeline >$7B, $3.5B won in FY26 | Industrial up double digits; total IoT ~+9%, flat guided for FQ4 |
| AI PC (Snapdragon X) | Low single digits (est., inside IoT) | ~9% of US retail Windows laptops above $600 (CQ1/CQ2'25, mgmt). Est. low-single-digit share of all PCs. Leads Windows-on-Arm (MediaTek/Nvidia entering) | ~$4B FY29 PC revenue target (FQ2'25 call) | Design wins only: 150 PCs commercialized in 2026; Chromebook/Gemini wins. Revenue not disclosed |
| XR / smart glasses | Small (inside IoT) | Near-monopoly platform (Meta Ray-Ban/Oakley, Samsung/Google glasses) | ~600M eyewear units/yr conversion opportunity (mgmt) | Strong but small in dollars |
| Data center (custom ASIC, AI200/AI250/HBC, CPU, SerDes) | ~0.9% ($88M FQ3'26) | ~0% vs Nvidia (dominant), Broadcom/Marvell (custom ASIC), AMD | Inference TAM is very large; QCOM targets $5B FY27 and $15B FY29 | From ~$0.3B FY26 to $5B FY27 (target). POs placed, wafers started; revenue begins in the Dec quarter |
| QTL licensing | 12.8% | Dominant cellular SEP licensor (Nokia, Ericsson, InterDigital far smaller) | Tied to handset ASP, capped royalty | Flat to slightly down (FQ3'26 -3% y/y); 69% EBT margin (src) |
Main competitors by segment: MediaTek (Dimensity, all Android tiers, Chinese OEMs), Samsung Exynos 2600 (S26 base models in some regions), Apple in-house (A-series and C1 modem), Huawei HiSilicon in China, Mobileye/Nvidia/Horizon Robotics in ADAS, Nvidia/Broadcom/Marvell/AMD in data center, and Intel/AMD/Apple M-series in PCs.
(a) On-device gen AI in handsets (ASP and mix). The content uplift was real in FY25 but reversed in FY26. Per Daloopa's 10-K handset bridge, FY25 handset growth came mostly from +$2.5B in higher revenue per chipset against only +$0.4B from units. By contrast, FY24 growth came from units (+$2.8B) while price per chipset subtracted -$0.5B. In FQ3'26 that tailwind reversed. CFO: "a little bit of a weaker mix within premium tier... OEMs making a choice on which chip to use and also using prior generation." Android QCT revenue is about -20% for FY26, worth more than $1.50 of EPS. Management's argument is that the "agentic smartphone cycle" will restore it ("early signs... in China, major OEMs are preparing to bring new on-device agents"). That has not shown up in revenue yet. The double-digit price increases are a cost pass-through, not AI content.
(b) AI PC. Design-win counts grew from 80+ (FQ1'25) to ~150 commercialized in 2026, and the X2 Elite received good reviews. Share is the weak point: it was stuck at ~9% of US retail Windows laptops above $600 across CQ1 and CQ2'25, and the company has not updated it since. PC revenue is not disclosed. This is still narrative.
(c) Data center inference. Disclosed data center revenue is $97M in FQ2'26 and $88M in FQ3'26 (mostly Alphawave connectivity). The target is $5B in FY27 and $15B in FY29, which is about 17x in a year. What is real: two hyperscaler custom-silicon engagements (Meta and Microsoft were named at Investor Day per press). CFO: "we do have POs from these engagements, and so we've already started wafers," with revenue starting in the Dec quarter. HBC Gen 1 has taped out, with launch in mid-2027. Caveats: the first revenue is low-margin custom ASIC work (a 1.5 to 2 point drag on QCT gross margin), the accelerator and CPU have no silicon proof yet ("customers want to see silicon"), and Humain (200MW AI200/AI250 announced in Oct 2025) was not mentioned on the FQ3 call. Broadcom/OpenAI and other custom-silicon programs compete for the same sockets.
(d) Auto AI. This is the one AI vector that shows up clearly in revenue. Growth accelerated from +14.6% (FQ1'26) to +38.3% to +61.4% (FQ3'26). Management credits the step-up in compute content from 4th- and 5th-gen Digital Chassis ("an order of magnitude increase... step function in silicon content"). In FQ3 Qualcomm won the lead ADAS plus cockpit compute role at BMW in a competitive process, and the Stellantis win extends the pipeline into the 2030s. The FY26 exit run rate was raised to ~$7B from $6B.
Did anything measurably change in FQ3'26?
| Signal | Before FQ3'26 | FQ3'26 | Direction |
|---|---|---|---|
| Auto y/y growth | +14.6% / +38.3% | +61.4%, ~+60% guided | Accelerating |
| Auto exit run rate | ">$6B" (FQ2'26) | ~$7B | Raised |
| Non-handset (auto+IoT) y/y | +11% / +20% | +28%, FY27 guided >60% (incl. DC) | Accelerating |
| Data center | "initial shipments later this calendar year" (FQ2'26) | POs from 2 hyperscalers, wafers started, Dec qtr revenue; HBC taped out | Firmer visibility, no revenue yet |
| FY29 non-handset target | $22B | $40B (Investor Day, 2 months before) | Raised |
| Handsets y/y | +3% / -13% | -19.6%; Apple share "materially lower" than 20% | Worse |
| Premium mix / AI content | ASP-led growth in FY25 | Weaker premium mix, prior-gen substitution | Worse |
| QCT EBT margin | 31% / 27% | 26%, guided 23 to 25% | Compressing |
Answer: yes for auto and data center visibility, no for handset AI. Measured in revenue, FQ3'26 was the first quarter where non-handset growth (+28%) clearly accelerated, and that came from auto. Data center moved from slides to purchase orders but still contributes under 1% of revenue. Handset "AI content" went backwards. Most of the AI story is still narrative. The proof point is FQ1'27 (Dec quarter) data center revenue and whether the $5B FY27 target holds.
- Competitors above 15% share per segment: total smartphone AP: MediaTek 31%, Apple 19%. Premium Android: Samsung Exynos (in base S26 models in some regions), MediaTek Dimensity. Modem: Apple (in-house), MediaTek. Auto ADAS: Mobileye, Nvidia. AI PC: Intel and AMD (more than 85% of Windows). Data center: Nvidia (dominant), Broadcom (custom). Licensing: no licensor comes close.
- Could a customer replace QCOM within 12 months? Handsets: partly yes. Apple is doing it now, Samsung shifts flagships to Exynos by region, and Chinese OEMs can move a generation to MediaTek within one launch cycle. Auto: no, because design wins lock in 3 to 5 year platforms and BMW/Stellantis are multi-generation. QTL: no, because the patents are needed. Data center: customers are not yet dependent, so the question does not apply.
- Price setter or taker: mixed. Qualcomm is pushing through double-digit price increases ("consistent with... other players"), which shows pricing power in premium. But OEMs are responding by down-binning to prior-gen chips and dual-sourcing, so it is a cost pass-through rather than true pricing power. In auto it is closer to a price setter. In data center custom silicon it is a price taker (gross-margin dilutive).
Oligopoly gate: PASS. Qualcomm holds more than 30% share in premium Android SoC, the cellular baseband (ex-Apple), auto digital cockpit, and cellular licensing. In premium Android, Qualcomm, MediaTek and Exynos are three players controlling well over 70%.
Qualcomm passes the oligopoly gate on the strength of premium Android SoC, the modem, the auto cockpit, and QTL. But by the rubric's own test it does not reach 7 to 8: its largest segment, handsets (51% of revenue), is declining about 20% y/y. In that segment Qualcomm is #2 in total AP share (23% vs MediaTek 31%) and losing share (from 27%), and Apple's modem business, about $7.5B of FY26 revenue, is going away. The AI thesis is measurable today only in auto, where growth accelerated from +15% to +38% to +61% in three quarters on rising compute content per vehicle, backed by multi-year BMW and Stellantis lock-ins. Auto is a clear leadership position in a growing theme, but it is only 16% of revenue. Data center is the biggest swing factor: $88M in the latest quarter against a $5B FY27 target. FQ3'26 upgraded it from aspiration to purchase orders and wafer starts, but the early revenue is low-margin custom ASIC work and the accelerator and CPU have no silicon yet. On-device gen AI in phones produced ASP-driven growth in FY25 (+$2.5B from revenue per chipset) but reversed in FY26 as OEMs traded down. So FQ3'26 did change the trajectory of non-handset revenue (+28% y/y, >60% guided for FY27), but the "AI reacceleration" of the whole company is still mostly narrative until data center revenue prints in the Dec quarter. Score 6/10: real oligopoly assets and one proven AI growth engine, offset by a shrinking, share-losing core and an unproven data center ramp.
Sources: Daloopa (company_id 157) as linked; QCOM earnings call transcripts FQ4'24 to FQ3'26 (FMP); Counterpoint AP-SoC share; Korea Times, Exynos share CQ2'26; Counterpoint premium SoC 2024; Investor Day coverage; ServeTheHome Investor Day; TechInsights baseband; Tom's Hardware Snapdragon PC share; Humain 200MW. Data sourced from Daloopa.