QCOM (Trans): DC ramp underway; non-handset rev. to rise 60%

The following is Dolphin Research's curated Trans of$Qualcomm(QCOM.US) FY26Q4 earnings call.

Core takeaways from the print. Quick recap.

1. Shareholder returns: Returned $2.0bn to shareholders this quarter. Buybacks and dividends were $1.0bn each.

2. FQ1 (Dec. qtr) guide

a. Total: revenue of $10.4bn–$11.2bn and Non-GAAP EPS of $2.15–$2.35. Non-GAAP opex Approx. $2.8bn.

b. QTL: revenue of $1.3bn–$1.5bn with EBT margin of 72%–76%. Reflects normal seasonality.

c. QCT: revenue of $9.0bn–$9.6bn with EBT margin of 20%–22%. Handsets down mid-single-digit QoQ as Android growth is offset by continued Apple declines; data center revenue >$650mn; Auto and IoT combined are roughly flat QoQ.

3. FY27 framework

a. Non-handset: Auto, IoT and data center combined to drive QCT non-handset revenue up >60% YoY starting in FQ1, with full-year mix surpassing 50% of QCT. This reweights the portfolio toward non-handset growth.

b. Data center: FY27 revenue of $5.0bn, ramping each quarter and exiting the year at an Approx. $7.0bn annualized run-rate. FY29 target is $15.0bn.

c. FY29 target: non-handset revenue of $40.0bn, with data center contributing >$15.0bn. The balance comes from Auto and IoT.

4. Margins and definitions

a. Pricing and GPM: Broad price increases to reflect higher semiconductor input costs, starting to show in FQ2 and fully reflected in H2 FY27 in QCT GPM. This underpins a recovery toward the low end of the 48%–50% historical band.

b. Custom chips: Shipping custom parts to two hyperscalers starting FQ1, accretive to QCT OP margin but dilutive to GPM. Mix will lift profitability despite lower gross margin.

c. Segment reporting: From FY27, data center will be consolidated into QCT; combined with AI accelerators and CPU platforms investing ahead of revenue, this drags QCT EBT margin by 300bps. Company-level guide and QCT's long-term 30% margin target are unchanged.

2. Earnings call details

2.1 Key executive remarks

1. Licensing (QTL): Long-term renewals completed with Apple, Huawei and Transsion, keeping the revenue run-rate unchanged. No major handset OEM renewals for several years; next renewal cycle is FY2030–FY2033. The company expects its inventions to remain foundational in the 6G era.

2. Data center: Transitioning from co-development to revenue in FY27 with solid order coverage. a. AWS: Agreement spans multi-gen custom AI inference accelerators and optical interconnects, with up to $60.0bn potential over 10 years; custom accelerators began shipping in Oct., and optical interconnect ramps later this month. Another optical interconnect customer has been secured. b. Products: AI 250 shows power and bandwidth better than expected, with early yields tracking plan; two significant potential customers were added this quarter. DragonFly C1000 CPUs are shipping to Meta as planned. c. Software: Modular has open-sourced all MAX code; Modular Cloud launched in the Sep. qtr, processing tens of trillions of tokens monthly.

3. Auto: FY26 exit ARR Approx. $8.0bn, $1.0bn above prior quarter's outlook. a. FQ4 Auto revenue exceeded $2.0bn for the first time, up 89% YoY, driven by volume ramp of integrated modules on the 5th-gen platform. b. The 5th-gen Snapdragon Digital Chassis entered commercial shipments in the Sep. qtr, marking the largest per-vehicle content uplift in history; in FY27, the company could rank among the largest auto chip suppliers by revenue.

4. Handsets: China Android bottomed in the Jun. qtr and rebounded QoQ this quarter, while agentic AI lifts per-device compute content. a. The Apple modem transition cadence is fully incorporated in plans. b. The first wave of agentic smartphones launched on Snapdragon, with high-end Android share expected to strengthen; leading AI firms aim to enable on-device 100B-parameter models by 2028. c. An AI Agent app topped 10mn downloads in 40 days; Meta, Alibaba and others have launched AI glasses.

5. IoT and infrastructure: Robotics is on the cusp of scale, and SpaceX opens up infrastructure demand. a. IoT: Industrial, networking and robotics grew 20% YoY, offsetting consumer drag from memory constraints; FQ4 IoT revenue rose 4% YoY. b. Robotics: Dragonwing IQ10 is set for commercial scale-up, with partners running models up to 80B parameters. c. SpaceX: Supplying the Dragonwing infrastructure platform for Starlink Mobile's nationwide 5G ground network in the U.S.

2.2 Q&A

Q: To exit FY27 at a $7.0bn annualized run-rate and nearly triple full-year data center revenue, will GPM keep falling as custom chips scale?

A: Once pricing fully lands, baseline GPM returns to the low end of the 48%–50% historical band. The increases reflect higher industry semiconductor costs and start showing in FQ2, with full impact in H2 FY27. Data center will further dilute GPM on top of that, but it is accretive to OP margin, consistent with last quarter's commentary.

Q: For which workloads are high-bandwidth accelerators most favored, and how is commercialization progressing?

A: Lab energy efficiency is Approx. 0.3 pJ/bit, better than the 0.5 design target shared at Investor Day. Yields are ahead of plan, and the most recognized workload is decode. Moving from slides to silicon changed customer dialogues, strengthening confidence in both custom and in-house commercial solutions and the 2029 goal; the technology is also entering custom chip discussions and is being paired with some customers' own accelerators.

Q: How does data center revenue ramp through the year, and when does the third customer on the slide contribute?

A: FY27 revenue is $5.0bn, with FQ1 >$650mn, and exiting the year at an Approx. $7.0bn run-rate. Supply supports the $5.0bn target, and upside beyond that is supply-constrained rather than demand-limited. FY28 starts from $7.0bn, with existing customers continuing, enabling strong YoY growth; the FY29 $15.0bn target is also highly covered.

The majority of the $5.0bn comes from two hyperscalers, including Amazon. The third customer involves accelerators, with additional custom chips under advanced discussion, feeding the FY28–29 ramp; CPUs also have new potential customers beyond Meta.

Q: How big is the pricing impact on handset revenue, and what is the handset market outlook next year?

A: We expect the handset market to decline low-to-mid teens in FY26 and be flat to slightly down in FY27. China Android QCT handset revenue bottomed in the Jun. qtr, with growth in the Sep. and Dec. qtrs, tracking our prior conservative plan. The pricing contribution to handset revenue was not quantified.

Q: Any change in per-device royalty rates in the newly signed licenses?

A: The revenue run-rate is unchanged, indicating rates and royalty levels were maintained upon renewal. Contract details are confidential and will not be disclosed.

Q: Will Amazon's warrant structure be replicated with other customers?

A: Amazon's at-the-money warrants are the only exception; all other data center engagements follow a traditional semiconductor model. Future structures will be evaluated case by case, but the overall framework remains aligned with the traditional model.

Q: Any outlook for the Mar. qtr?

A: No specific Mar. qtr guide, only the full-year cadence framework. After Apple-related product revenue declines, the Dec. qtr is no longer the seasonally highest quarter; data center ramps each qtr starting from >$650mn in FQ1 and sets the quarterly shape. Auto, IoT and data center combined non-handset revenue is up >60% YoY in FY27, more than offsetting Apple declines; handset will account for less than 50% of revenue in FY27.

Q: You previously indicated growth in the Mar. qtr; do you still stand by that?

A: No firm commitment was provided; we noted that the current market modeling framework is reasonable.

Q: Is the 'flat' handset market assumption optimistic, and is China Android's recovery just pre-price-hike pull-forward?

A: For FY27 we assume flat to down, with lean channel inventory, and the FY26 destock headwind will not repeat. In FY26, QCT handset revenue was impacted by market decline and OEM channel destocking; we monitor channel inventory closely and cross-validate with QTL data, and expect QCT revenue to track the market.

Price increases are focused on new platforms and models, so pull-forward effects are limited, which is why pricing impact becomes more visible in H2. Data indicates Jun. remained the bottom for China Android.

Q: Is the next-gen Samsung platform still planned at 75% share?

A: The new baseline for Samsung share is 70%+, and future Galaxy generations are planned accordingly. The prior baseline was 50%. We are also expanding supply into Samsung's wearables and other devices, where Qualcomm was not a major supplier previously, and are now seeing attach with premium Galaxy designs.

Q: Did Auto outperformance include pre-price-hike pull-forward, and is it sustainable?

A: No pull-forward; growth is driven by share gains and higher content on the 5th-gen platform. From the 3rd to the 5th gen, per-vehicle chip content increased 8x; most growth is design-win conversion to revenue, and the step-up in SoC revenue will continue.

Q: How does QCT EBT margin evolve from 20%–22% over the year?

A: The long-term 30% QCT margin target stands, and we are confident of achieving it as revenue scales over the next 2–3 years. No quarterly or annual margin guide. The 300bps drag is purely a segment-definition change (data center consolidated into QCT from FY27, with multiple product lines investing ahead of revenue) and does not affect company-level numbers.

Q: What is the size of the interconnect business today and how does it grow in FY27–28? How does industry consolidation in optical interconnect affect Qualcomm?

A: Interconnect revenue is embedded in the $5.0bn and $15.0bn data center targets and is not broken out separately. The asset came via the Alphawave acquisition; this quarter we demonstrated optics and co-packaged optics with ECOC 2026 and Lumentum.

We expect structural change in the industry, analogous to RF front-end, and are discussing various forms of strategic cooperation with multiple laser companies; the roadmap includes more optical products across 400G, 800G and 1.6T.

Q: What is the upside from new CPU customers beyond Meta, and does the Arm lawsuit affect the CPU roadmap or costs?

A: The Arm lawsuit does not affect the CPU roadmap, and customers beyond Meta are incremental. A jury found Arm in breach of the Architecture License Agreement (ALA), intentionally interfering with Qualcomm's business and engaging in unfair competition; we are confident in our ability to continue developing and shipping Oryon products.

Current $5.0bn and $15.0bn targets include CPUs only for Meta; any new CPU customer would be upside to those targets.

Q: Will all three hyperscalers contribute at scale by the end of FY27?

A: The $5.0bn is primarily contributed by two of them and continues into FY28–29. Both are multi-gen, multi-product, long-term collaborations.

Q: Will FY27 opex step up meaningfully?

A: Opex philosophy is unchanged from the past year, with FQ1 guided at Approx. $2.8bn. Investments focus on adding new capabilities, especially for data center; each Jan. we reset calendar-year employee-related items as usual.

Q: Timing, scale and scope of the SpaceX/Starlink engagement?

A: Revenue starts to show by late FY27, with more in FY28. Scale was not quantified; we noted it is nationwide in the U.S. and could be one of the largest ground networks, with deployment beginning in 2027.

Qualcomm supplies RF (including massive MIMO), digital unit accelerators and Layer 1 software; the multi-year agreement covers next-gen wireless, and we see Open RAN and AI-native trends enabling innovation in this area.

Q: What composes the FY27 $5.0bn data center revenue? Does it exclude accelerators?

A: Most comes from two hyperscalers' mainline custom AI accelerator chips, followed by interconnect. The FY29 $15.0bn includes expansion in custom and interconnect and adds CPUs and in-house AI accelerators; we are confident in meeting and exceeding $15.0bn.

<End of text>

Risk disclosure and statements:Dolphin Research Disclaimer and General Disclosure