Qualcomm

Phase A — Understand the business

Lens 1 · Company Overview

Qualcomm is two businesses bolted together by one patent portfolio, and you cannot value it without separating them.

QCT (Qualcomm CDMA Technologies) — the chip company. Designs and sells integrated SoCs (Snapdragon application processors), modems, RF front-end, and connectivity silicon. It is fabless — designs the chips, outsources manufacturing. FY2025 revenue $38,367M, EBT $11,670M (30% EBT margin). Three end-markets: Handsets $27,793M, IoT $6,617M, Automotive $3,957M.

QTL (Qualcomm Technology Licensing) — the patent company. Grants licenses to Qualcomm's cellular-essential IP (3G/4G/5G SEPs); collects a royalty on essentially every modern smartphone sold globally, whether or not it contains a Qualcomm chip. FY2025 licensing revenue $5,582M, EBT $4,043M — a 72% EBT margin. This is the highest-quality earnings stream in the company and the reason QCOM is not just a cyclical chip name.

QSI / "Other" (Qualcomm Strategic Initiatives, plus the new QGOV government and Data Center businesses) is immaterial today — QSI equipment-and-services revenue was ~$0 in FY2025 — but the Data Center sub-segment is where management is staking the next growth narrative (Lens 12).

Contract structure / payment terms. QCT is transactional hardware (sell-through to OEMs, no take-or-pay). QTL is recurring royalty under multi-year license agreements; the load-bearing risk is that those agreements expire between fiscal 2027 and 2031 and must be renewed, sometimes via arbitration that can reset terms unfavorably. Revenue is highly concentrated: Apple, Samsung and Xiaomi each comprised ≥10% of consolidated revenue in FY2025.

Plain terms: Qualcomm makes the brains for premium Android phones, taxes the entire smartphone industry through patents, and is now trying to convert that engineering base into automotive, PCs, and AI data-center silicon before its single largest chip customer (Apple) finishes walking out the door.

Lens 2 · Supply Chain

Upstream inputs → Qualcomm → end customer, named at every node ]:

Upstream (foundry / assembly — Qualcomm is fabless, this is the chokepoint):

• Wafer fabrication: TSMC, Samsung Electronics, GlobalFoundries. Leading-edge Snapdragon is overwhelmingly TSMC — a single-source dependency at the bleeding-edge node, shared with every other premium SoC designer (Apple, NVIDIA, MediaTek, AMD).
• Assembly & test (OSAT): Advanced Semiconductor Engineering (ASE), Amkor Technology, Siliconware Precision (SPIL), STATSChipPAC.
• IP / instruction set: Arm Holdings — Qualcomm licenses the Arm architecture; the Nuvia/Oryon dispute (Lens 10/12) is the live chokepoint on the custom-CPU roadmap.
• High-speed connectivity IP: Alphawave (acquired Dec 18, 2025, $2.3B) — now in-house SerDes for the data-center push.

Qualcomm (design + integration): SoC/modem/RF/connectivity design, system software, the AI Engine.

Downstream (customers / OEMs):

• Handset OEMs: Samsung, Xiaomi, and the broader Android premium tier (Oppo, Vivo, Honor, Motorola); Apple as a modem-only buyer (thin MDM, lower margin) — and a departing one.
• Automakers (Snapdragon Digital Chassis): dozens under the ~$45B design-win pipeline (; Lens 4).
• PC OEMs (Snapdragon X): Microsoft (Copilot+ PC reference), Dell, HP, Lenovo, Samsung.
• Data center (emerging): "a leading hyperscaler" custom-silicon ramp + sovereign-AI / CSP opportunities, initial shipments guided for the Dec 2026 quarter.
• QTL licensees: effectively every smartphone maker on earth.

Chokepoints: (1) TSMC leading-edge capacity — non-substitutable in the near term; (2) Arm — both an IP supplier and a litigation adversary; (3) China — 46% of revenue ships into China/HK (Lens 4), a geopolitical single-region dependency; Huawei is permanently lost as a product customer (export controls).

Lens 3 · Competitive Advantages (moats)

The QTL patent moat is genuine and durable — and it is the whole investment case for the bull. Qualcomm holds the foundational cellular-SEP portfolio; the royalty is levied per-device on the industry's installed base, insulated from Qualcomm's own chip share. 72% EBT margins that barely moved YoY ($4,043M on $5,582M FY25; $4,027M on $5,572M FY24 ) prove the pricing power. The moat's expiry is legal/contractual, not competitive — it erodes only if renewals reset rates or SEP enforcement weakens, not if a rival builds a better product.

The QCT moat is real but narrowing. Qualcomm's edge is system integration — modem-to-antenna RF, the on-device AI Engine, and now custom Oryon CPUs — at premium-tier power/performance. Switching costs exist (OEM software/integration lock-in, modem certification across global carriers). But:

• MediaTek has closed much of the gap at the high end (Dimensity flagships) and dominates mid-range.
• Apple designs its own A-series SoC and is now building its own C1 modem in-house — the single most important moat breach in the story.
• Premium Android is a structurally flat-to-shrinking TAM.

Bargaining power: Strong over most Android OEMs (they need Qualcomm's premium SoC + the SEPs). Weak over Apple (Apple is exiting) and weak over TSMC (Qualcomm needs leading-edge capacity more than TSMC needs Qualcomm). The QTL leverage over the entire industry is the asymmetry that matters — that is where Qualcomm "needs them less than they need Qualcomm."

Net moat verdict: a wide, decaying-clock licensing moat wrapped around a narrowing, cyclical product moat. The bull thesis is that custom data-center/PC silicon widens the product moat back out before the Apple exit and any QTL rate reset bite.

Lens 4 · Segments

By product (QCT end-markets), FY2025 vs FY2024:

The trend has now reversed in the latest tape. Q2 FY2026 (3 mo. ended 2026-03-29) QCT end-markets:

Read: the FY2025 handset growth was a cyclical recovery + the last full year of high Apple share; in Q2 FY2026 handsets are down 13% YoY as the Apple ramp-down begins and the premium-Android cycle softens. Automotive (+38%) and IoT (+9%) are the designated offsets, and Automotive is delivering — but off a base ($1.3B/qtr) far too small to fill a $7B Apple hole on its own near-term.

By geography, FY2025:

Geographic concentration is extreme and unchanged: ~67% China + South Korea (Korea ≈ Samsung). This is a US/China-trade and single-customer-region risk stacked on top of the customer concentration.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print — Q2 FY2026, ended 2026-03-29)

The reported numbers are a trap if you read the GAAP headline; the operating story is a soft quarter dressed up by a tax swing.

Headline (GAAP):

• Revenue $10,599M, −3.5% YoY (vs $10,979M Q2 FY25). Below the prior-year quarter — the first clear YoY decline in the recovery.
• Operating income $2,309M vs $3,120M — operating margin compressed to 21.8% from 28.4%.
• R&D $2,463M (+11.2%) and SG&A $898M (+27.2%) — opex inflation, partly Alphawave integration.
• Pre-tax income $2,232M vs $3,105M (−28%).
• Net income $7,370M — up 162% — but only because of a $5,138M income tax BENEFIT (effective tax rate −230%), a partial release tied to CAMT/OBBB guidance.
• GAAP diluted EPS $6.88 vs $2.52 — entirely a tax artifact.
• Non-GAAP diluted EPS $2.65, beat consensus ~$2.57.

What drove it: Handsets −13% YoY (Apple drag + cycle); Automotive +38% (record); QTL +5% at 72% EBT.

Guidance — the actual signal: Q3 FY2026 revenue ~$9.6B midpoint, −7% YoY, below consensus; non-GAAP EPS $2.10–$2.30. Management is explicitly guiding the top line down as Apple fades — the tone has shifted from "diversification offsets" to "navigating the handset trough."

Balance-sheet flags: Cash & equivalents $5,435M + marketable securities $4,364M ≈ $9.8B liquidity. Inventories rose to $7,368M (from $6,526M at FYE25) — building into a softening handset market is a yellow flag worth tracking. OCF H1 FY26 $7,414M (vs $7,141M) — cash generation intact. Capex stepped up to $1,082M in H1 (vs $491M) — data-center build.

Market reaction: The stock had run to a record ~$247 on data-center hope, then de-rated; QCOM is down ~20% YTD 2026. The market is paying for the AI/data-center optionality and selling the handset reality.

Unusual vs. own history: The −230% effective tax rate (Q2) and the +56% rate in FY2025 (Lens 10) are both OBBB/CAMT artifacts — two consecutive quarters where GAAP EPS is nearly meaningless. Use non-GAAP and pre-tax operating income to judge the business.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts/ empty); this lens is ``.

Tracking management's framing across recent calls:

• FY2024: tone = "diversification is working," 5G recovery, automotive pipeline to ~$45B. Confident, growth-forward.
• Early FY2026 (Q1, record revenue): peak optimism — record quarter, data-center "multiple billions ahead," edge-AI-wins-the-AI-race messaging.
• Q2 FY2026: the pivot. Amon frames a "handset bottom" and pushes the story to data center (custom silicon ramp with a hyperscaler, Dec-quarter shipments) and automotive records.

Recurring phrases: "diversification," "edge AI," "Snapdragon Digital Chassis," "design-win pipeline," and now "custom silicon / hyperscaler / sovereign AI." What they've stopped emphasizing: handset share gains and Apple. The rhetorical center of gravity has moved decisively from "we own premium mobile" to "we are an edge-AI and data-center company." That is a tell that management knows the mobile leg is in secular decline and is selling the next act — credible only to the extent the data-center revenue actually shows up (it is ~$0 today).

Lens 7 · Comps

Peer table — QCOM vs key global semis. Multiples are ``, dated; where a figure was not sourced it is marked n/a (no fabrication).

Read: QCOM at ~20–23x forward is the cheapest mega-cap on this AI-adjacent list except where MediaTek's depressed near-year EPS distorts its ratio — and it is cheap for a reason: the comp set (NVDA, AVGO) is buying secular AI-compute growth, while QCOM's forward EPS still leans on a declining handset/Apple base. The bull's entire valuation argument is "re-rate QCOM toward the AVGO custom-silicon multiple as data center scales." Until data-center revenue is non-trivial, 20x is arguably generous for a name guiding revenue down 7% — not cheap. Note the consensus 12-mo price target sits below the current price ($180 avg vs ~$220 spot ), i.e. the sell-side sees the stock as already ahead of fundamentals.

Lens 8 · Stock-Price Catalysts (what actually moves QCOM)

Pattern of >5% moves over recent years:

• Apple modem news is the single biggest swing factor — every disclosure on Apple's in-house C1 timeline and Qualcomm's share guidance (70%→20%→0%) has moved the stock.
• Arm litigation milestones — the Dec 2024 jury verdict (favorable) and the Sept/Oct 2025 final judgment were positive catalysts; an Arm appeal win would be a sharp negative.
• Data-center / AI announcements — the run to a record ~$247 was AI-data-center hope; the +6% single-day "AI data center hopes build" move is illustrative.
• China demand & US/China trade headlines — given 46% China exposure, export-control and tariff news move it.
• Earnings guidance (not the print) — the Q3 FY26 below-consensus guide is the latest example.

What it reveals: QCOM trades on narrative transitions (Apple-out vs data-center-in) and binary legal/regulatory events far more than on quarterly EPS beats. The non-GAAP beats have not stopped a ~20% YTD decline — the market is trading the direction of the business mix, not the current quarter.

---

Phase C — Judge people & books

Lens 9 · Management

CEO: Cristiano Amon (President & CEO since June 2021; ~25-year Qualcomm veteran, prior president/COO who led the 5G rollout).

1. Track record. Engineer-operator who built Qualcomm's 5G leadership and secured premium SoC supply through the COVID shortage when rivals could not — a genuine operational win. As CEO he made diversification "priority number one," scaling automotive from <$2B to ~$4B and a ~$45B design-win pipeline, and IoT into a real segment.
2. Tenure & skin in the game. ~5 years as CEO, decades at the company. Insider ownership is modest (professional-manager scale, not founder) — insider-transactions.csv not present; specific % n/a.
3. Capital allocation. Disciplined and shareholder-friendly: FY2025 returned $12,596M ($8,791M buybacks at avg $155.43 + $3,805M dividends, $3.48/sh); refreshed with a $15.0B program (Nov 2024) and a further +$20B authorization in FY26 plus a dividend raise to $0.92/qtr. M&A has been bolt-on and strategic — Nuvia (custom CPUs, 2021) and Alphawave (data-center SerDes, $2.3B, Dec 2025) — augmenting capability rather than empire-building. ROIC is solid but the Alphawave/data-center bet is unproven.
4. Red flags. None of the egregious kind (no related-party self-dealing surfaced). The honest critique is strategic, not governance: Amon is betting the multiple on data center and PCs to replace Apple, and the Nuvia acquisition that underpins the custom-CPU roadmap is what triggered the Arm litigation — i.e. the growth strategy carries embedded legal risk he created.
5. Archetype. Professional manager / operator (not founder), well-suited to executing a diversification pivot but not a visionary capital allocator. The team is deep-bench semiconductor engineering.

Verdict: a credible, candid operator with a clean capital-allocation record. The question is not competence — it's whether the data-center bet pays before the Apple hole opens fully.

Lens 10 · Forensic Red Flags

financials.csv was header-only; figures cited to the filings.

• Tax-driven GAAP distortion (the big one). FY2025 effective tax rate 56% (income tax expense $7,122M vs $226M FY24) because Qualcomm established a valuation allowance writing off federal deferred tax assets after OBBB (enacted 2025-07-04) reinstated immediate domestic R&D expensing — a non-cash charge that crushed FY25 net income to $5,541M despite pre-tax income rising to $12,663M (+22.5%). Then Q2 FY2026 booked a $5,138M tax benefit (−230% rate). Two quarters of near-meaningless GAAP EPS — anyone anchoring on reported net income/EPS is misreading the business in both directions. Go-forward the rate normalizes to a permanent ~14% by fiscal 2027 with CAMT (15% min) from FY2026. Not a fraud flag — a "read non-GAAP and pre-tax" flag.
• Inventory build into softness. Inventories $7,368M at 2026-03-29 vs $6,526M at FYE25 (+13%) while handset revenue falls 13% YoY — watch for write-downs if the handset trough deepens.
• Receivables: AR $4,347M roughly flat vs revenue — no obvious channel-stuffing signal.
• SBC / non-GAAP gap: The ~$2.65 non-GAAP vs $6.88 GAAP Q2 gap is dominated by the tax benefit, not SBC add-backs; SBC is present but not the distortion here. Still, the habitual non-GAAP framing deserves the usual scrutiny.
• Cash flow vs earnings: OCF ($7.4B H1) comfortably exceeds non-GAAP earnings — cash conversion is clean and better than GAAP net income suggests. No divergence red flag.
• Goodwill/intangibles: Alphawave adds goodwill/intangibles to watch for future impairment if the data-center thesis stalls.

Regulatory findings (required sub-section):

• SEC Litigation Releases: None naming Qualcomm in 2021-06-21 → 2026-06-21 (EDGAR EFTS LR search) — total_sec_findings: 0.
• SEC AAERs: None in the period.
• Item 3 / Legal Proceedings (10-K Note 7): The material item is Arm Ltd. v. Qualcomm (Nuvia ALA dispute, filed 2022-08-31) — Arm sought to force Qualcomm to cease using and destroy Oryon/Nuvia-derived CPU core technology. Resolved largely in Qualcomm's favor: Dec 2024 jury found no breach of Qualcomm's ALA; Sept/Oct 2025 the Delaware court (Judge Noreika) ruled neither Qualcomm nor Nuvia breached and denied Arm a retrial. Arm has stated it will appeal, and Qualcomm counter-sued Arm (Jan 2025) over withheld ALA deliverables. Residual appeal risk on the entire custom-CPU roadmap, but the base case is cleared.
• Non-SEC enforcement (historical context, ): Qualcomm has a long antitrust history (FTC v. Qualcomm — Qualcomm ultimately prevailed on appeal, 2020; EU and Korea FTC fines in the prior decade, largely litigated/reduced). No new material government enforcement action surfaced for the current period.

Net: clean on fraud/accounting (zero SEC findings); the only forensic story is the OBBB tax noise (cosmetic, non-cash) and an inventory build to monitor. Legal overhang (Arm appeal) is a tail risk, not a base case.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (non-GAAP EPS, FY2026 → FY2028)

Built bottom-up from the latest actuals + guidance. All inputs labeled; output `` with arithmetic. No forecast.ts create in --watchlist mode (per skill rules).

Anchors:

• FY2026 non-GAAP EPS consensus ≈ $12.6; H1 FY26 non-GAAP run-rate (~$3.41 Q1 + $2.65 Q2 = ~$6.06) + a guided ~$2.20 Q3 supports a ~$11.5–12.6 full year depending on Q4.
• Apple drag: ~$7B annual revenue ($5.7–5.9B modem + $1.6–1.9B RF) phasing out — ~70% iPhone share fall 2025 → 20% fall 2026 → ~0% fall 2027. The full hit lands in FY2027–FY2028.
• Offsets: Automotive on track to ~$8B by FY2029 (from ~$4B FY25, ~+25%/yr); IoT growing high-single/low-double; data center "multiple billions" but ~$0 today, first shipments Dec-2026 quarter. Non-mobile target ~$22B by FY2029.
• QTL: ~flat $5.5–6.0B at 72% EBT, with renewal risk FY2027–2031.
• Buybacks: ~3–4% annual share-count reduction at current pace (56M of ~1.1B shares in FY25).

The call: Base ≈ flat EPS (~$12) through FY2027, re-accelerating only in FY2028 if data center delivers. This is a show-me stock — the next two fiscal years are a transition trough where the Apple loss roughly cancels the diversification gains, and the entire upside case rests on FY2028+ data-center scaling that is currently $0.

Brier seed (not logged — watchlist): "QCOM FY2027 non-GAAP EPS ≥ $12.0", p ≈ 0.45 (base case says ~flat-to-down; the Apple drag + QTL renewal risk make a clean ≥$12 a coin-flip-leaning-under).

Lens 12 · Bull vs Bear

Bull case. Qualcomm is a 72%-margin patent annuity (QTL) you get for free underneath a chip business pivoting into the two biggest secular pools in tech — automotive compute (~$45B design-win pipeline, +38% YoY, $8B by FY29) and AI data-center silicon (custom CPUs/XPUs/ASICs, Alphawave SerDes in-house, a hyperscaler ramp shipping Dec-2026). The Arm litigation that threatened the custom-CPU roadmap is won. Capital allocation is impeccable — +$20B buyback, growing dividend, ~3-4%/yr share shrink. At ~20x forward it is the cheapest AI-adjacent mega-cap; if data center re-rates it toward Broadcom's 25x on a higher EPS base, the stock works from both numerator and denominator. The market is over-discounting a known, telegraphed Apple loss and under-pricing the optionality.

Bear case (2-3 permanent-impairment risks).

1. The growth story is a timing mismatch. Apple's $7B exits in FY27-28; automotive (+$1B/yr) and data center ($0 today) cannot fill that on the same timeline. Revenue is already −3% YoY with Q3 guided −7%. The "diversification offsets Apple" math doesn't close for 2+ years — a multi-year EPS trough the ~20x multiple doesn't reflect.
2. QTL rate reset. The 72%-margin annuity that is the moat has license agreements expiring FY2027–2031; any renewal via arbitration at lower rates permanently impairs the highest-quality earnings stream. This is the under-appreciated tail.
3. Data-center credibility gap. Qualcomm has tried and retreated from server CPUs once before (Centriq, 2018). "Multiple billions" against entrenched NVIDIA/Broadcom/AMD/Arm-Neoverse incumbents is a from-zero bet; if it stalls, the entire re-rate thesis evaporates and you're left holding a declining handset taxer at 20x.

Pre-mortem (18 months out, thesis broke): It's late 2027. Apple is fully out, handsets never recovered, a flagship Android customer (e.g. a Samsung renewal) reset QTL rates lower, the hyperscaler data-center ramp slipped or stayed sub-$1B, Arm won its appeal and clouded the Oryon roadmap, and China softened on trade tension. EPS is ~$10, the multiple compressed to 14x on "structural decliner," and the stock is in the $140s — roughly the consensus PT, which already implied downside.

Are multiples too high? For a name guiding revenue down with the growth engine still at $0, ~20x forward is full, not cheap — it already prices a successful pivot. The bull needs execution; the bear just needs the status quo to persist.

Contrarian view (what the market refuses to see): The bull consensus is that "QCOM is a cheap AI re-rate." The contrarian read is the opposite — the market is too optimistic: it is paying 20x and pricing data-center success that has not shipped a dollar, while underweighting that the QTL annuity (the only thing that makes QCOM more than a cyclical) faces a renewal cliff in the exact window the Apple loss lands. The asymmetry may be to the downside, not the upside.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull.

• What structurally breaks the money machine: QTL. Strip the 72%-margin licensing annuity (rate reset on FY27-31 renewals, or SEP-enforcement erosion) and QCOM is a mid-cycle fabless SoC vendor that deserves a low-teens multiple, not 20x. The bull case quietly assumes QTL is permanent — it is contractual and expiring.
• Revenue concentration: Apple, Samsung, Xiaomi each ≥10%. Apple is leaving (−$7B). If Samsung (Korea = 21% of revenue) vertically integrates further (Exynos) or resets its license, a second leg wobbles. Two of the three largest customers are credible self-supply threats.
• The moat is weaker than bulls think: MediaTek has caught up at the high end; Apple proved a hyperscaler-scale buyer can design out Qualcomm's modem; the on-device-AI edge is replicable. The product moat is switching costs + integration, not a true monopoly — and switching costs didn't keep Apple.
• Most dangerous underestimated competitor: Broadcom in data center (custom ASIC for hyperscalers is their franchise — Qualcomm is the late entrant), and MediaTek in handsets/auto. Bulls frame data center as greenfield; it is Broadcom's home turf.
• Worst management moves: The Nuvia acquisition that underpins the growth roadmap created the Arm litigation; the prior Centriq server-CPU retreat shows execution risk on exactly the data-center bet now being sold. Accounting is clean, but the strategy embeds self-inflicted legal/execution risk.
• Assumptions that must hold for today's price: (1) QTL renews near current rates; (2) data center reaches several $B by FY28-29 at good margins; (3) auto keeps compounding ~25%; (4) Arm's appeal fails; (5) China stays open. All five must hold. Miss two and the multiple compresses.
• If growth disappoints 20-30%: A revenue path that's −7% (Q3 guide) deepening toward −10-15% as Apple fully exits, with data center under-delivering, takes EPS to ~$10 and the multiple to ~13-14x → a $130–150 stock (−30%+ from ~$220).
• Single scenario that permanently impairs: A bad QTL renewal cycle (Samsung/Apple-licensing reset lower) coinciding with a data-center stall — the annuity shrinks just as the replacement growth fails. Plausibility: moderate, but it is the specific combination the price does not protect against.

Lens 14 · Management Questions (ordered by information value)

1. QTL license agreements expire between FY2027 and FY2031 — which specific large licensees renew in FY27-28, and what is your realistic range for blended royalty-rate change at renewal?
2. Quantify the data-center revenue ramp: what is the FY2027 and FY2028 revenue you underwrite from the hyperscaler custom-silicon programs, and what gross margin?
3. After Apple reaches ~0% in fall 2027, what is the net QCT handset revenue trajectory FY2028+ assuming no further share gains?
4. Centriq was exited in 2018. What is structurally different this time that makes the custom CPU/XPU data-center push durable against Broadcom and NVIDIA?
5. If Arm wins its appeal of the Nuvia ruling, what is the contingency for the Oryon-based PC and data-center roadmap?
6. The ~$45B automotive design-win pipeline — what conversion rate to revenue, over what years, and what is the gross-margin profile vs. corporate average?
7. With Apple, Samsung and Xiaomi each ≥10% of revenue, what is your plan if Samsung accelerates Exynos in-housing?
8. Snapdragon X PCs: what unit share of the Windows installed base do you actually expect by FY2028, and is Microsoft's Arm commitment exclusive enough to rely on?
9. Inventory rose 13% into a 13%-down handset quarter — what is the write-down risk and the normalized inventory target?
10. Post-OBBB, confirm the steady-state non-GAAP tax rate (the ~14% by FY2027) and CAMT cash impact — what is the true go-forward FCF conversion?
11. Capital allocation: with the $20B authorization, do you prioritize buyback at ~20x or preserve dry powder for further data-center M&A like Alphawave?
12. China is 46% of revenue — model the downside if US export controls or tariffs tighten on advanced-node mobile/auto SoCs.
13. What is the realistic timeline and TAM for the QGOV (government) and edge-AI businesses you've started breaking out?
14. Gross/operating margin trajectory: with R&D +11% and SG&A +27%, when does opex growth normalize below revenue growth again?
15. What single internal metric do you watch that would tell you the diversification is failing to offset mobile — and where is it today?

---