Frequently asked questions
Can Qualcomm have a move like AMD as AI moves to the edge?
Partially, and on a delay. AMD's re-rate needed a depressed multiple, named hyperscaler contracts, consensus EPS forced sharply up, and second-source scarcity. Qualcomm has the multiple (about 13.6x fiscal 2028 consensus EPS), the contracts (HUMAIN 200MW of AI200 racks; Amazon custom silicon with warrants tied to up to $60 billion of purchases), and the scarcity. What it lacks near term is the earnings slope: consensus fiscal 2027 EPS of about $10.20 dips below fiscal 2026 as Apple modem revenue exits. The DalalBytes bull case of $430 by fiscal 2029 is the AMD script on a two-year delay, roughly 2.4x, not 180% in a year.
What happens to Qualcomm when Apple finishes switching to its own modems?
The chip revenue ends and the royalties continue. Teardowns show the iPhone 18 Pro and Duo already use Apple's C2 modem worldwide, with Qualcomm's X80 left only in the US iPhone 18 Pro Max, and every 2027 iPhone is expected to use an Apple modem. Management guided Apple product revenue down about 50% sequentially into the December 2026 quarter. But on September 24, 2026, Apple renewed its global patent license with Qualcomm effective April 2027, so Qualcomm keeps collecting royalties on the standards inside Apple's own modem.
What is Qualcomm's data center business?
An inference-first entry built on the AI200 accelerator (up to 768GB of LPDDR per card, liquid-cooled racks), the AI250 with near-memory computing due 2027, an Oryon-based server CPU, and custom silicon for hyperscalers. HUMAIN committed to 200MW of racks from 2026. Amazon signed a multi-year custom silicon deal in September 2026 with revenue starting in the December quarter and warrants tied to up to $60 billion of purchases over ten years. Management targets $5 billion of data center revenue in fiscal 2027 and over $15 billion by fiscal 2029.
Should investors believe Qualcomm's roadmap and fiscal 2029 plan?
Partially, and the report audits it. Management's June 2026 plan targets $40 billion of non-handset revenue by fiscal 2029 (data center over $15B, automotive $10B, IoT over $14B) and non-GAAP EPS above $18. The track record is two-sided: the 2021 automotive target was beaten a year early (fiscal 2025 auto revenue of $3.96B against a ~$3.5B five-year goal), while the 2021 IoT target of up to $9B by fiscal 2024 missed by about 40% (actual $5.42B). DalalBytes' base case assumes the plan lands at roughly 80 percent, with data center near $10 billion rather than $15 billion.
What an AMD move actually requires?
Four ingredients: a depressed starting multiple, named contracts large enough to rewrite the earnings path, consensus forced sharply higher, and second-source scarcity against NVIDIA. Qualcomm scores three and a half. Multiple: ~17x fiscal 2027E, ~13.6x fiscal 2028E, the cheapest in large-cap AI semis (Broadcom ~46x trailing, Marvell ~89x). Contracts: HUMAIN 200MW is 0.2GW against AMD's 12GW of headlines, but the Amazon instrument is structurally identical to AMD's deals, custom silicon plus warrants. Estimates: the slope is the problem, fiscal 2027 EPS dips before fiscal 2028 recovers to ~$12.95. Scarcity: genuine. The mechanism has fired once already in miniature: +20% in a day on the October 2025 AI200 announcement.
How do the businesses break down?
QCT chips were $38.4B of fiscal 2025 revenue at a 30.4% EBT margin (handsets $27.79B, automotive $3.96B, IoT $6.62B, from $22.57B / $1.87B / $5.94B in fiscal 2023); QTL licensing was $5.58B at a 72.4% EBT margin. Handsets were $5.09B in Q3 FY2026 (-20% on the memory crisis and OEM inventory drawdown). Automotive hit a record $1.59B (+61%, 23rd straight double-digit quarter) on the Snapdragon Digital Chassis, against a design-win pipeline raised to $65B in June 2026, with BMW naming Qualcomm its lead compute silicon provider for cockpits and automated driving and Ride Pilot shipping in the Neue Klasse iX3. One honest debit: GM's 2028 flagship AD platform moves to NVIDIA Thor. IoT was $1.83B (+9%): Snapdragon X2 PCs at 80 TOPS with new Surface models and Google's Gemini-native Googlebook program, XR2 in Meta Quest, AR1 in Ray-Ban Meta glasses, and Dragonwing industrial with a $7B+ design pipeline.
How does the roadmap grade out?
The expanded report audits three Investor Day plans. The 2021 plan: automotive beaten a year early, QTL exact, IoT missed by ~40%, and the mid-teens QCT growth promise missed. The 2024 plan: the $22B auto-plus-IoT fiscal 2029 target, with auto tracking ahead and PC behind (reported cumulative Snapdragon X laptop sales near 720,000 units). The June 2026 plan: non-handset revenue of $40B by fiscal 2029 (data center $15B+, auto $10B, IoT $14B+ re-cut as $8B industrial/robotics plus $6B personal AI and compute), EPS above $18, non-handset growth guided above 60% in fiscal 2027 with data center at $5B. The silicon roadmap underneath it is coherent: one Oryon CPU line from phones to a reported 250-core server part (Dragonfly C1000, Meta deploying from H2 2028), Hexagon redesigned around memory behavior, modems to the Release 19 X105, and a 50-company 6G coalition targeting 2028 pre-commercial devices. The plan is aggressive where this team has earned it (cars, contracts) and soft where it has not (consumer PC). Base case: roughly 80% delivery.
Why is the data center leg the whole debate?
Because it is the only leg that re-rates the stock. The AI200 inverts NVIDIA's architecture: LPDDR instead of HBM, up to 768GB per card, priced on cost per token rather than peak FLOPS. The AI250 adds near-memory computing (10x+ effective bandwidth) in 2027. Alphawave Semi brought SerDes and connectivity IP; Modular, acquired in 2026, attacks the CUDA software moat. Management targets $5B of data-center revenue in fiscal 2027 and $15B+ by fiscal 2029, inside a $40B non-handset goal. The honest discount: 200MW is small, the entry is three years behind AMD, and Centriq's 2018 failure is the scar tissue. Named customers and warrant alignment are what is different this time.
What do the financials show?
Revenue compounded from $35.8B (FY2023) to $39.0B (FY2024) to $44.3B (FY2025), with non-GAAP EPS of $12.03 and free cash flow rising $9.85B to $11.16B to a record $12.8B, returned at ~98% in fiscal 2025 ($8.8B buybacks plus $3.8B dividends). The balance sheet carries $10.2B of cash and securities against $14.8B of debt (net debt ~$4.7B), with a further $20B buyback authorized in April 2026 and the dividend at $0.92 quarterly (~2.1% yield). Fiscal 2026 lands near ~$42.9B and ~$10.48 normalized EPS as the memory crisis and the Apple step-down bite. Consensus: fiscal 2027 EPS ~$10.20-10.45 (the air pocket), fiscal 2028 ~$12.95-13.62. China (including Hong Kong) is 46% of fiscal 2025 revenue by customer headquarters, the largest single structural exposure. QTL's roughly $4B of annual pre-tax profit, re-anchored by the Apple renewal (September 2026) and the Huawei multi-year agreement (October 2026), is the ballast under all of it.
What are the key risks?
Data-center execution (HUMAIN pace, the first Amazon revenue in the December 2026 quarter, a second hyperscaler), the fiscal 2027 air pocket deepening on the November 4, 2026 print, handset memory pricing and China premium share versus MediaTek, licensing regulation in China and the EU, TSMC concentration at 2nm, and China geopolitics. One legal overhang cleared this week: on October 13, 2026, a Delaware jury found Arm breached its architecture license with Qualcomm, securing the Oryon roadmap that powers the PC, auto and data-center strategy through 2033.
© 2026 DalalBytes Research · For educational purposes only. Not investment advice.