Cirrus Logic

Phase A — Understand the business

Lens 1 · Company Overview

Cirrus Logic is a fabless analog/mixed-signal semiconductor designer headquartered in Austin, TX (incorporated 1984, public since 1989), with its second engineering hub in Edinburgh, Scotland (the former Wolfson Microelectronics audio team). It sells into "the world's top mobile and consumer applications". The business is a single reportable segment reported in two product lines:

• Audio — amplifiers, codecs, smart codecs (codec + integrated DSP), ADCs/DACs, standalone DSPs, plus the SoundClear software/algorithm portfolio. FY26 Audio revenue $1,159.9M (58% of total).
• HPMS (High-Performance Mixed-Signal) — camera controllers (autofocus / optical image stabilization), haptic & sensing drivers, and battery/power ICs. FY26 HPMS revenue $837.4M (42% of total), +10% YoY — the stated growth engine.

Total FY26 net sales $1,997.4M (+5% YoY). The whole company is built around one go-to-market motion: win a custom design slot inside a flagship device, then expand dollar-content per device over successive product cycles ("an initial design win can often lead to an opportunity to further increase our content with a customer over time").

Customer structure — the defining fact. Components are proprietary and single-sourced from CRUS's view, so the "end customer" is the device OEM specifying the part. One end customer, Apple, was ~91% of FY26 net sales (FY25 89%, FY24 87% — concentration is rising, not falling); the ten largest end customers were ~96%. In the most recent quarter (Q3 FY26) Apple was an even sharper ~94%. Contracts are short-term POs with no minimum-purchase commitment: "most of our customers can stop incorporating our products… with limited notice… and suffer little or no penalty". There is no take-or-pay protecting CRUS; the only take-or-pay in the model runs the other way (CRUS owes GlobalFoundries — see Lens 2).

Payment terms / cyclicality. Sales are ~99% international (Apple's contract manufacturers in China/India/Vietnam), denominated in USD; backlog is a near-useless forward indicator (visibility "principally within the next 60 days"). This is a consumer-electronics, smartphone-cycle-levered business with effectively one anchor tenant.

Lens 2 · Supply Chain

Map: silicon wafers (foundry) → assembly/test (OSAT) → Cirrus design/IP → Apple's contract manufacturers → iPhone/Mac → consumer.

Upstream named stakeholders:

• Foundries (front-end wafers): primarily GlobalFoundries (fabs in Singapore and Germany) and TSMC (fabs in Taiwan). The newest smart codec is a 22-nanometer part; CRUS just joined Apple's "American Manufacturing Program" and is working with GlobalFoundries to manufacture for the first time at the Malta, New York fab — a geographic/process-diversification move pulled by Apple.
• Assembly & test (OSAT, back-end): Advanced Semiconductor Engineering (ASE), Amkor Technology, STATS ChipPAC, SFA Semicon, and Siliconware Precision (SPIL).

Chokepoints / single-source dependencies:

1. The GlobalFoundries Capacity Reservation Agreement (the one hard liability). Signed July 2021, amended Feb 2025, it obligates CRUS to buy a defined wafer quantity every quarter through calendar-year 2026, with shortfall payments. CRUS paid a $60M non-refundable reservation fee (balance $10M at Dec-25) and pre-paid $195M for wafers (balance $33M at Dec-25 / $14.7M at Mar-26). This means if Apple cuts orders, CRUS is still on the hook for wafers → excess inventory / margin drag. It is the inverse of a moat: a take-or-pay on the cost side with no take-or-pay on the revenue side. It expires end-2026 — a real catalyst/overhang (see Lens 8/11).
2. Most products are single-foundry, single-OSAT. CRUS concedes it does "not typically manufacture the majority of these products at more than one foundry or more than one assembly and test subcontractor," and that transferring to a backup "would likely be substantial" in cost/effort.
3. AI/datacenter crowding-out: management explicitly flags that "foundries we use have faced increased demand associated with data center build-outs and AI-related applications," pressuring price/allocation/lead-times. CRUS is a low-priority customer for leading-edge capacity relative to the GPU/HBM crowd.

Names present → this lens passes. The single most important supply-chain fact is that CRUS's only long-term purchase commitment is to GlobalFoundries, not from Apple.

Lens 3 · Competitive Advantages (moats)

What the moat actually is — design-in lock-in + audio IP depth, narrow but deep.

• Switching costs at the design level. Parts are "specific to our customer's system architecture and frequently cannot be sold to other customers". Once Cirrus silicon is co-designed into an iPhone audio/power subsystem, it is locked for that product generation — ripping it out means a re-spin and re-validation. This is real but generational, not perpetual: every new device cycle reopens the socket.
• Analog/mixed-signal IP + talent. ~4,340 patents worldwide (~1,620 granted US); 71% of 1,668 employees in R&D; ~22% of revenue reinvested in R&D ($434M FY26). Low-power, high-precision analog is a craft discipline where design experience compounds and is hard to hire into quickly — the structural reason Apple keeps outsourcing it rather than fully in-housing.
• Bargaining power — weak, and structurally so. CRUS sits across the table from a customer that is ~91% of its revenue and ~3-4% of Apple's BoM cost. The 10-K is unusually candid: customers "demand favorable commercial terms," "pressure us on price reductions," and CRUS has "experienced pricing pressure from certain key customers" with ASPs that "will decline from time to time". Gross margin holding at ~52-53% despite annual ASP step-downs is the tell that the cost side (yield + mix) is doing the work, not pricing power.

Net: a genuine but asymmetric moat — strong enough to keep the socket across generations and hold ~53% gross margins, not strong enough to set price or to survive the anchor tenant deciding to dual-source or in-house. Bull framing: "sticky custom silicon partner to the best hardware company on earth." Bear framing: "a captive supplier whose moat is Apple's current preference."

Lens 4 · Segments

By product line (FY24 → FY25 → FY26, $M):

Trend & cause: HPMS is the accelerant, Audio is the mature base. Audio grew low-single-digits (smartphone unit strength + PC share gains, offset by ASP step-downs). HPMS grew 10% on higher smartphone unit volume plus custom non-smartphone content — and management states HPMS "presents the largest opportunity to drive application diversification". CRUS reports one operating segment (CODM = the CEO, evaluates on consolidated net income), so there is no segment-level operating income or EBITDA disclosure — operating income by product line is n/a — not disclosed.

By geography (ship-to of contract manufacturer; ~all is Apple supply chain) — Q3 FY26 nine-month, $M:

The geographic mix is a map of Apple's contract-manufacturing diversification away from China (India/Vietnam surging), not of CRUS's own customer diversification. End-demand is still one device family.

---

Phase B — Measure performance

Lens 5 · Earnings Result

Two data points anchor this: the audited FY26 full year (10-K) and the Q4 FY26 print (reported May 6-7, 2026).

FY2026 full year (audited):

• Net sales $1,997.4M (+5.3% YoY) — record.
• Gross profit $1,054.2M, gross margin 52.8% (FY25 52.5%, FY24 51.2%) — slow grind up on mix.
• Operating income $460.4M, op margin 23.0% (FY25 21.6%, FY24 19.2%) — real operating leverage as R&D held flat ($434.0M, −0.2%) on a growing top line.
• Net income $414.4M, net margin 20.7%; diluted EPS $7.85 (FY25 $6.00, FY24 $4.90 — a 3-yr EPS CAGR of ~27%, flattered by buybacks).
• Effective tax rate fell to 16.6% from 25.5% — the single biggest EPS tailwind in FY26, driven by the OBBBA (One Big Beautiful Bill Act, July 2025) permanently killing the TCJA R&D-capitalization/GILTI drag. ~$0.80-1.00 of the FY26 EPS jump is tax, not operations.
• Operating cash flow $650.6M (FY25 $444.4M); capex only $14.8M (fabless → ~0.7% of sales). FCF ≈ $635.8M, ~31.8% FCF margin.
• Balance sheet: fortress. Cash + marketable securities = $800.9M + $86.7M + $266.2M = $1,153.8M; zero debt (revolver undrawn); goodwill $435.9M (Wolfson legacy); inventory cut to $240.9M from $299.1M. Net cash ≈ $1.15B, ~16% of market cap.

Q4 FY26 print (the catalyst):

• Revenue $448.5M (+6% YoY), gross margin 53.0%.
• Non-GAAP EPS $1.95 vs $1.69 consensus — beat ~15%. (GAAP Q4 diluted ≈ $1.58, derived as FY26 $7.85 − 9M $6.27.)
• Q1 FY27 guide: revenue $430-490M (mid $460M, +13% YoY) vs $414.7M consensus — a clean beat-and-raise.
• Stock made a new ~52-week high; Stifel → $197, KeyBanc → $200, Barclays → $140 (Equal Weight) post-print.

Unusual vs own history: the +13% guided Q1 growth is a step-up from the ~5% FY26 full-year pace, signaling either new content ramping (smart power IC / FaceID work) or easy comps — bulls read content, bears read pull-forward. Margin is range-bound at 52-53% as always; the EPS story is volume + tax + buyback, not margin expansion.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the research-layer shelf (transcripts=0), so this is ``. Across the FY26 call sequence the management narrative is consistently and increasingly "diversification" + "content gains":

• Recurring phrases: "three-pronged strategy," "content per device," "HPMS expansion," "PC share gains across all segments," "AI-enabled PCs," and newly in FY26 — "$9B SAM by 2030," "smart power IC," and a "FaceID" collaboration with the largest customer.
• Tone shift: measurably more confident through FY26 — each quarter added a new non-smartphone vector (PCs → automotive haptics → smart-grid/energy-measurement ICs → the FaceID smart-power-IC win), and Q4 closed with a beat-and-raise and the formal $9B-SAM target. The "diversification" drumbeat is louder precisely because the Apple % keeps rising — management is selling the future mix because the present mix is more concentrated than ever.
• What they stopped saying: the post-COVID "inventory correction / soft consumer" hedging that dominated FY24 calls has faded; FY26 is a clean-growth narrative.

Caveat: the bullish call tone is the company's own framing of a single-customer business; treat the $9B-2030 SAM as TAM-marketing, not backlog.

Lens 7 · Comps

Peer set from the 10-K's named competitors + the index.

Read: at ~15x forward / ~12x EV/EBITDA / ~18.6x trailing, CRUS is the cheapest name on the board by a wide margin — TXN trades >50x forward, SYNA's EV/EBITDA is distorted high by a depressed cycle. The market is not mispricing quality; it is applying a single-customer discount. A ~20% ROE, ~32% FCF-margin, zero-debt compounder at 15x is "cheap" only if you underwrite the Apple relationship as durable. The China names (Goodix, Awinic) matter as the low-end displacement vector, not as valuation comps — I will not invent their multiples.

Lens 8 · Stock-Price Catalysts

5-year tape and what moves it (`` + research-layer TSR table). Over 5 years CRUS returned +72.7% (TSR index 100→172.65, 3/2021→3/2026) — but badly lagged the Philadelphia Semiconductor Index +153.8% (100→253.82). CRUS roughly tracked the broad Russell 3000 (+63.4%), i.e. it has not been treated as an AI/semi growth name — the market caps its multiple precisely because of the Apple tether.

What actually moves the stock >5%:

1. Apple iPhone unit/build-cycle news and "content win" headlines — the dominant driver. The May-2026 Q4 beat-and-raise + FaceID/smart-power-IC content story drove a new 52-week high; analyst PTs jumped to $197-200.
2. Quarterly beat/raise vs the ~$414M-ish guide — CRUS reliably sandbags Q1 (smartphone seasonality) and beats.
3. Any Apple-sourcing/competitive-displacement scare — the single largest downside catalyst type (see the Skyworks→Broadcom precedent, Lens 13).
4. Tax-law / OBBBA — the FY26 re-rate had a macro-policy tailwind embedded.
5. NOT broad AI capex — CRUS does not rally with the GPU complex; if anything AI competes for its foundry capacity.

Pattern: this is an "Apple-content + beat-the-sandbag" stock, not a secular-AI stock. The market reacts to the iPhone BoM and to socket-risk headlines, and discounts everything else.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO: John Forsyth — CEO since January 2020 (~6.5 yrs), longtime Cirrus insider (strategy/sales background) who took over from Jason Rhode. Track record: under Forsyth, revenue grew from ~$1.28B (FY20) to $2.0B (FY26) and diluted EPS from the ~$3-4 range to $7.85, while diversifying HPMS from a sliver to 42% of sales — a credible execution record on the stated three-pronged plan.
• Skin in the game — thin. Forsyth directly owns ~76,187 shares (~0.14% / ~$8.9M), ~220k beneficial incl. options. FY26 comp ~$10.8M, ~93% equity/bonus. He has been a net seller (~26.8k shares sold TTM, ~$2.95M, no open-market buys) — routine 10b5-1 selling, not alarming, but not the insider-accumulation pattern bulls like. (10-Q discloses GC Scott Thomas adopting a 10b5-1 to sell up to 19,123 sh.) insider-transactions.csv not on shelf → ownership detail is ``.
• Capital allocation — disciplined and shareholder-friendly, buyback-only. ~$1.2B returned since 2020, 100% via buybacks (never paid a dividend). Share count cut from ~62M (2020) to 50.6M (−18% over the period; −4.4% in the last year). FY26 alone: $280M repurchased at avg ~$107-142/sh; $274M still authorized under the March-2025 $500M program. ROE ~20.3% (Jun-26), up from ~15.7% (2021). No value-destroying M&A; the last big deal (Wolfson, 2014; goodwill $435.9M) is long-digested.
• Red flags: none material on governance. Comp is equity-heavy and aligned; no related-party dealings disclosed; auditor is Ernst & Young since 1984 with an unqualified opinion + effective ICFR. The only "promotional" tell is the $9B-2030-SAM framing — aspirational TAM marketing, common in semis.
• Archetype: professional-manager operator (not founder), executing a narrow, well-defined content-expansion strategy competently. The right archetype for a mature, single-anchor supplier; the wrong one if the situation ever demands a bet-the-company pivot away from Apple.

Lens 10 · Forensic Red Flags

Forensic pass over the income statement, balance sheet, cash-flow — every figure labeled.

• Earnings quality — clean. FY26 OCF $650.6M vs net income $414.4M (1.57x cash conversion); FCF $635.8M > net income. Cash flow exceeds earnings — the opposite of an aggressive-accrual red flag. Net income is not being flattered by capitalized costs (capex is trivially small, R&D fully expensed).
• Receivables / inventory vs revenue: AR $220.1M at FY26 (down from a Q3 spike of $279.0M; +2% YoY full-year vs +5% sales — clean). Inventory cut to $240.9M from $299.1M — a use of working-capital discipline, not a channel-stuffing build. No allowance for doubtful accounts (zero) given the Apple-grade counterparty.
• SBC: $62.0M for 9M FY26 (~4% of sales) — moderate for a chip designer, and notably the buyback ($280M) is 4.5x SBC, so dilution is more than mopped up (share count genuinely falling). Non-GAAP EPS ($1.95 Q4) vs GAAP ($1.58) gap is ~20%, mostly SBC + amortization — material but disclosed and ordinary; track the GAAP line.
• Goodwill $435.9M (18% of assets) — Wolfson-era, never impaired; in a single-customer business a future Apple loss is the scenario that would trigger an impairment test. No impairment indicated FY26.
• Prepaid wafers / capacity reservation — the one balance-sheet item that could bite: the GF take-or-pay means a demand air-pocket converts to excess inventory / shortfall payments. Inventory write-down judgment is the Critical Audit Matter flagged by E&Y — appropriate, given single-customer demand forecasting risk. Watch inventory + prepaid-wafer balances each quarter through the CY26 agreement expiry.
• Segment opacity: reporting one segment with only two revenue lines and no product-line profitability limits forensic granularity — not a red flag per se, but it means margin-by-product (e.g. is HPMS dilutive?) is unknowable from filings.

Regulatory findings (required sub-section).

• SEC Litigation Releases: none. "No LR found for this company in the search period" (2021-06-30 → 2026-06-30).
• SEC AAERs: none. "No AAER found".
• Non-SEC enforcement (FTC/DOJ/FDA/CFPB): web search returns no material enforcement actions, consent decrees, or fines against Cirrus Logic.
• 10-K Item 3 (Legal Proceedings): company discloses only ordinary-course litigation — "management does not believe that there are any pending matters that could potentially have a material adverse effect".
• The one material contingency (not "enforcement" but quantified tax exposure): the IRS is examining FY2017-2019 federal returns on transfer-pricing between US and UK affiliates; the final Revenue Agent's Report asserts ~$168.3M additional tax + ~$63.7M penalties (excl. interest) — CRUS has accrued no additional liability and is in IRS Appeals, prepared to litigate. Combined ~$232M+ exposure is real and unreserved beyond ASC 740 amounts; an adverse outcome "could be material." This is the single biggest off-model financial risk and it is disclosure-quality flagged, not an accounting irregularity.

Verdict on the books: No material regulatory or accounting findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-30. Earnings quality is high (cash > earnings, falling share count, conservative inventory). The watch-item is the unreserved ~$232M IRS transfer-pricing exposure, not revenue recognition.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Bottom-up from FY26 actuals + the Q1 FY27 guide. Base year: FY26 (ended Mar-28-2026) net sales $1,997.4M, op margin 23.0%, tax 16.6%, diluted shares 52.8M, diluted EPS $7.85. Buyback assumption: ~3% annual share-count reduction (consistent with the FY26 ~$280M pace + remaining $274M authorization). Tax normalizes toward ~17-18% as OBBBA international provisions phase in FY27.

Base arithmetic, shown: FY27 sales $1,997.4M × 1.07 ≈ $2,137M; op income @ 24% ≈ $513M; less ~$0.9M net interest expense + $38M interest income ≈ $550M pre-tax; × (1−0.175) ≈ $454M net; ÷ 51.2M diluted shares (−3%) ≈ $8.40 EPS. This sits above the lower published consensus ($7.62, statutory, "−3.8%" — a tax-normalization view) and below the higher one ($8.90); consensus FY28 ~$9.39. The spread is the thesis: the published $7.62 figure prices tax normalization eating the FY26 tax windfall; the $8.90 figure prices content-driven growth through it.

Brier forecast — NOT logged (per --watchlist rule: only forecast.ts create when genuinely committing a base case; this is an unattended breadth dossier). If promoted to a thesis, the loggable line would be: "CRUS FY27 ( end Mar-2027) non-GAAP diluted EPS ≥ $9.00, p≈0.45, resolves 2027-05-31."

Lens 12 · Bull vs Bear

Bull case. The cleanest fabless-analog compounder in the market at a value multiple. ~53% gross / ~23% operating / ~32% FCF margins, 20% ROE, zero debt, $1.15B net cash, ~18% share-count reduction since 2020, and a CEO executing a content-expansion playbook that has lifted HPMS to 42% of sales. The $9B-SAM-by-2030 map (consumer $0.9B→$1.4B, PC $1.0B→$1.3B, +75% non-smartphone SAM) gives a credible second leg in PCs / automotive / battery-power, and the FaceID smart-power-IC win shows Apple is still adding, not subtracting, sockets. At ~15x forward / 8.6% FCF yield, you are paid to wait while buybacks compound EPS even at flat revenue. If diversification de-risks the Apple % even toward 80%, the multiple re-rates toward the broad-analog 20-25x and the stock doubles.

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

1. Single-customer cliff. ~91% Apple, rising. Apple in-houses or dual-sources the audio/power subsystem (it has the cash, the talent, and a documented pattern — see Lens 13) → 30-60% of revenue evaporates in 1-2 product cycles with the GF take-or-pay still running. This is not tail risk; it is the risk, and it is structural.
2. Margin/price compression with no offset. Annual ASP step-downs are contractually normal; CRUS has zero pricing power against Apple. The day mix stops improving, gross margin rolls over and the EPS algorithm (which leans on tax + buyback, not margin) stalls.
3. Foundry crowd-out + take-or-pay asymmetry. AI/datacenter demand crowds CRUS out of capacity and raises wafer cost, while the GF agreement forces purchases regardless of Apple demand — a margin vise in any down-cycle.

Pre-mortem (18 months out, thesis broke): the iPhone 18 cycle disappointed on units and a competitor (or Apple silicon) took the smart-power-IC / a codec socket; revenue guided down high-single-digits, the GF take-or-pay produced an inventory write-down, the IRS appeal went against them ($232M cash hit), and the "diversification" PC/auto revenue was too small ($<200M) to matter. The 15x "value" multiple compressed to 11x on a lower EPS base — a 40%+ drawdown.

Are multiples too high? No — they are arguably correct. 15x forward for a 90%-single-customer supplier is neither a bargain nor a bubble; it is an efficient single-customer discount. The mistake would be calling it "cheap vs TXN" without pricing the concentration.

Contrarian view (what the market refuses to see): the consensus debate is "Apple risk vs cheap multiple," but the underappreciated swing factor is the FY26 tax windfall reversing. ~$0.84+ of FY26 EPS was the OBBBA tax cut; international provisions phase in from FY27. If the Street is anchoring its $8.90 on a 16.6% tax rate that normalizes to 18-19%, the "growth" is partly a one-time tax step that won't repeat — the bear's quietest, most quantifiable point.

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

Dismantling the bull case.

• Where revenue is concentrated, and the live precedent. ~91% one customer is the whole story — and the market just watched the analog version of this risk detonate: in October 2025, Skyworks disclosed it was losing iPhone orders to a competitor (Broadcom), and Skyworks + Qorvo were forced to merge into a ~$22B combined company to survive falling Apple orders. That is the exact scenario for CRUS, in an adjacent socket, already proven to happen. The bull "sticky design-in" thesis is precisely what Skyworks bulls believed.
• Why the moat is weaker than bulls think. The switching cost is generational, not structural — every iPhone redesign reopens every socket, and Apple's own silicon team (which already does the SoC, modem-in-progress, and increasingly power management) is the most dangerous competitor, with infinite resources and an explicit history of insourcing supplier functions. Apple's in-house modem rollout (putting $5.7-7.8B of Qualcomm revenue at risk by 2027) proves the willingness to climb the BoM. Audio/power is a plausible next target.
• The China low-end vector bulls ignore. Awinic and Goodix are named competitors commoditizing amplifiers/codecs from below; the 10-K itself warns "competitors that can tolerate lower margins… pose a risk." On non-Apple Android, CRUS is already largely absent — its TAM is effectively "Apple + a little PC."
• Worst capital-allocation critique: none egregious — but 100% buyback / 0% dividend at a peaky multiple ($142/sh repurchases in Q4 FY26) returns cash at high prices, and the cash hoard ($1.15B) earning T-bill yield is arguably under-deployed for a company with one customer and no transformative M&A appetite.
• Assumptions that must hold for $144: (1) Apple keeps/grows the socket through iPhone 18/19; (2) ASP step-downs stay offset by mix; (3) the GF take-or-pay doesn't bite; (4) the IRS appeal doesn't cost $232M cash; (5) tax stays ~17%. If growth disappoints 20-30% (a partial socket loss), FY27 EPS goes to ~$5.50-6.00 and a de-rated 11-12x multiple = $65-72/sh, ~50% downside.
• Single scenario that permanently impairs the business: Apple insources or dual-sources the boosted amplifier + smart codec. Plausibility: moderate and rising — not this year (the FaceID win shows current additive momentum), but the multi-year base rate for Apple-concentrated component suppliers (Imagination, Dialog partial, Skyworks 2025) is genuinely bad.

Lens 14 · Management Questions (ordered by information value)

1. Walk us through the exact mechanism and timeline by which Apple could second-source or insource the boosted amplifier and 22nm smart codec — what makes those two sockets harder to displace than the Skyworks RF socket that went to Broadcom?
2. What is the revenue and gross-margin bridge if Apple revenue fell 20% in a single product cycle — specifically, what shortfall payments would be owed under the GlobalFoundries agreement, and what inventory write-down would result?
3. The GF Capacity Reservation Agreement expires end-CY2026. What replaces it — terms, volume commitment, pricing — and does the next agreement carry the same take-or-pay asymmetry?
4. Quantify the FaceID smart-power-IC opportunity: expected dollar-content per device, ramp timing, and whether it is sole-sourced or shared.
5. Of the $9B 2030 SAM, how much is contracted/design-won today versus aspirational TAM, and what non-Apple revenue do you underwrite for FY28-29?
6. The FY26 effective tax rate fell to 16.6% on OBBBA. What is the normalized FY27-29 tax rate as the international provisions phase in, and how much of the FY26 EPS step was tax versus operations?
7. What is your plan and price discipline for the $1.15B net cash — given you have one customer, no dividend, and no recent M&A, why is buyback-only the right answer at $140+ per share?
8. On the IRS transfer-pricing dispute ($168.3M tax + $63.7M penalties) — best/worst-case cash timing and amount, and why is nothing beyond ASC 740 reserved?
9. Where are you structurally absent in Android/China, and is competing there (vs Awinic/Goodix) ever economic, or is the strategy permanently "Apple + PC"?
10. PC revenue grew share "across all segments" — give us the absolute PC revenue number and the FY27 trajectory; when does non-smartphone exceed 20% of total?
11. How do you defend gross margin at ~53% as ASPs step down annually — how much runway is left on the cost/yield side before mix can no longer offset price?
12. What second customer could realistically reach 10%+ of revenue, and on what timeline — or is structural single-customer concentration permanent?
13. How is Apple's American Manufacturing Program / Malta NY changing your cost structure and capex, and does US fabrication compress or expand your margins?
14. What is your exposure if iPhone units structurally decline on lengthening replacement cycles (a risk you cite) — what is the content-growth rate needed just to offset flat units?
15. How are you using/defending against AI both in your design flow and in foundry capacity competition — are you being crowded out of leading-edge capacity by the GPU/HBM demand you flag in your own 10-K?

---