Microchip Technology

Phase A — Understand the business

Company Overview

Microchip designs, manufactures and sells embedded-control semiconductors — microcontrollers (MCUs), analog/mixed-signal, FPGAs, memory, timing, connectivity and data-center silicon — sold as a "Total System Solution" (hardware + software + dev tools) to ~101,000 unique customers across automotive, industrial, aerospace & defense, communications, consumer and data center. It is one of the canonical analog/MCU "jelly-bean" compounders: tens of thousands of proprietary, long-lived SKUs, each tiny, collectively sticky, sold heavily through distribution.

FY2026 revenue $4,713.1M (+7.1% YoY) split three ways:

• Mixed-signal Microcontrollers — $2,355.4M (50.0%), +4.7% YoY. Proprietary → stable ASPs.
• Analog — $1,329.0M (28.2%), +14.9% YoY (the fastest-growing leg).
• Other — $1,028.7M (21.8%), +3.4% (FPGA, memory, timing, foundry/licensing, aerospace).

The reportable-segment cut is two segments: semiconductor products $4,549.3M and technology licensing $163.8M (SuperFlash IP — 100% gross margin, +25% YoY, the highest-quality revenue line).

Customers / channel / concentration: ~47% through distributors, 53% direct. Only one >10% customer/distributor — Arrow Electronics at 12% of net sales. End-customer concentration is genuinely low; the risk is channel concentration and the bullwhip it transmits. Contract structure: no long-term distributor agreements (terminable at will, with non-cancellable-order carve-outs); price concessions + stock-rotation rights to distributors create variable-consideration accruals; some demand carried in LTSAs (long-term supply agreements), several of which were cancelled/modified in the downturn. Take-or-pay does not protect this revenue — it is short-cycle, book-and-ship, with limited visibility outside LTSAs.

Supply Chain

A genuine IDM-lite model — Microchip owns a large share of its own fabs and back-end, which is the structural cost-and-resilience story. Named map upstream → company → end customer:

• Wafer fabrication (front-end): Internal — Fab 4 (Gresham, OR, 8-inch, 0.11–0.5µm) and Fab 5 (Colorado Springs, CO, 6-inch + SiC/8-inch expansion, paused through FY27); Fab 2 (Tempe, AZ) closed May 2025 (downturn capacity cut, tech moved to Fab 4/5). 35% of FY26 sales came off own US fabs; 65% from outside foundries — including all 300mm and some 200/150mm specialty. External foundries (named in industry: TSMC, UMC, X-FAB, SkyWater for specialty/SiC) supply the leading-edge and overflow — but Microchip does not name a single dominant foundry, which is the point: diversified, not TSMC-captive like the fabless peers.
• Assembly & test (back-end): Internal facilities in Thailand and the Philippines carried ~67% of assembly / ~69% of test in FY26; specialized A&D A&T in Germany, France, Ireland, UK, Philippines, Thailand, US. Balance outsourced to OSAT subcontractors (ASE, Amkor in industry).
• Inputs: silicon wafers, lead-frames, substrates, gold/copper bond wire, gases/chemicals; "multiple sources… but in some cases single or limited".
• Distribution → end customer: Arrow (12%) + a broad distributor network → ~101,000 OEMs.

Chokepoints: (1) outside 300mm foundry dependence for the newest products; (2) Asia back-end concentration (Thailand/Philippines) — geopolitical + disaster single-region risk; (3) ~75% of sales to foreign customers, China ~18%, Taiwan ~15% of revenue → direct tariff / export-control / cross-strait exposure. The owned-fab base is the moat and the operating-leverage liability (unabsorbed-capacity charges hammer gross margin in a trough — see Lens 5).

Competitive Advantages (moats)

The moat is real but mid-tier within an elite peer set. Sources of durability:

• Switching costs / design-in lock-in. MCUs are designed into a customer's board with Microchip's dev tools, compilers, IP libraries; re-spinning to a competitor is costly and slow. Microchip explicitly leans on its development-tool ecosystem and "familiarity/adoption of dev tools" as a future-selection driver. ASPs on proprietary MCU/analog lines "remained relatively stable" through the worst inventory glut in a decade — direct evidence of pricing power.
• Breadth / Total System Solution. ~50% MCU + 28% analog sold into the same sockets; one-stop silicon supply. The catalog is the moat — tens of thousands of long-lived SKUs no single competitor replicates.
• Owned manufacturing → cost. "One of the lowest-cost producers in embedded control" via internal fab + A&T capturing the manufacturing margin.
• Licensing flywheel. SuperFlash embedded-flash IP licensed to foundries and IDMs — a high-margin, asset-light annuity that also seeds the ecosystem.

Bargaining power: Strong over a fragmented ~101k-customer base (no customer >12%); weaker over leading-edge foundries (it needs TSMC more than TSMC needs it for 300mm). Where the moat is thinnest: Microchip itself flags Chinese/Taiwanese competitors that "copied, cloned, pirated or reverse engineered" its lines — the China domestic-substitution wave (see Lens 13) is the structural threat the moat does not fully repel. Versus TXN (scale + 300mm internal cost lead) and ADI (higher-performance analog mix), Microchip is the broad, sticky, mid-performance, more-leveraged name — a good business, not a great one.

Segments

By product line (FY26 vs FY25, $M):

By reportable segment (gross profit is the only segment P&L Microchip discloses — it does not allocate opex/assets):

By geography (FY26): Asia 49.9% ($2,352.9M), Americas 29.5% ($1,391.3M), Europe 20.6% ($968.9M). Within Asia: China ~18%, Taiwan ~15% of total. All three regions grew in FY26; Europe was the weakest into the downturn and is recovering last.

The trend that matters: analog and licensing are out-growing the MCU core, mix is tilting toward higher-margin proprietary content, and management says automotive + industrial now >50% of sales (vs ~40% in 2020) — a deliberate shift toward higher-content, longer-cycle sockets.

---

Phase B — Measure performance

Earnings Result

The story is a violent down-cycle (made worse by a ransomware hit) now in a sharp V-recovery. The 3-year income statement:

Revenue fell ~42% peak-to-trough (FY24→FY25) and operating income collapsed from $2.57B to $0.30B — textbook semiconductor operating deleverage on a high-fixed-cost owned-fab base. FY26 is the inflection year: +7.1% revenue, but the recovery is back-end-loaded and accelerating hard within the year.

The latest prints (the live signal):

• Q3 FY26 (Dec 2025): net sales $1,186.0M, +15.6% YoY; GM 59.6%, op margin 12.8%; net income $62.7M (vs −$53.6M loss a year prior).
• Q4 FY26 (Mar 2026): net sales $1,311.2M, +35.1% YoY, +10.6% QoQ [derived: 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. FY26 $4,713.1M − 9M $3,401.9M = $1,311.2M; corroborated web: IR press release "$1.311B"]; GAAP GM 61.0% / non-GAAP 61.6%; GAAP op margin 16.6% / non-GAAP 30.6%; GAAP EPS $0.21 / non-GAAP $0.57. The non-GAAP operating margin snapping back to 30.6% is the recovery thesis in one number — the model's incremental margins are ferocious off the trough.
• Beat: Q4 revenue beat consensus, led by data-center / AI connectivity wins.

Guidance (Q1 FY27, June 2026 qtr): net sales $1.442–1.469B (mid $1.456B; +35.3% YoY, +11.0% QoQ), non-GAAP GM 62.25–63.25%, non-GAAP EPS $0.67–0.71. Tone shifted decisively bullish: "largest booking month in almost four years," strong expedite activity, early distributor restocking.

Balance-sheet flags: inventory cut $1,293.5M → $1,035.4M (251 → 185 days); distributor inventory 33 → 26 days (low end of the 17–43 historical band) — the channel is clean, which is bullish for sustained sell-in. But cash fell $771.7M → $240.3M (−$531.4M) as they paid down the $1.2B 2025 Notes and funded the dividend partly with new debt. Receivables rose $204.5M on the revenue ramp (normal). Net income $230.0M but OCF $962.1M (D&A $689.3M + SBC $255.4M do the heavy lifting).

Market reaction: stock up ~56% in 3 months / ~110% in a year, all-time high $105.91 on 2026-05-08. The recovery is not a surprise to the tape — it is the consensus trade.

Earnings Calls (sentiment trend)

No transcripts on the shelf (transcripts/ empty); reconstructed from filings + IR releases, labeled accordingly. Tone arc over the last ~4 quarters:

• FY25 (trough): "weak business," "large inventory correction," restructuring, 10% headcount cuts, Fab 2 closure, LTSA cancellations, covenant relief sought. Defensive, survival-mode.
• Mid-FY26: "improvement in our business after customers reduced excess inventory," "significant revenue growth mode," inventory declining.
• Q4 FY26 (latest): "largest booking month in almost four years," strong expedite/booking/sell-through, data-center design momentum. Confident, growth-forward.

Recurring phrases now: "Total System Solution," "data center and AI," "book-to-billBook-to-billNew orders divided by orders filled. Above 1 means the backlog is growing; below 1 means the company is working through it faster than it is replacing it. improving," "distributor restocking," "operating leverage." Phrases they stopped saying: "inventory correction," "covenant relief," "resize manufacturing." The language has flipped from balance-sheet-defense to demand-offense — a clean sentiment inflection. (Flag: management is the founder who built the company and is mid-turnaround — read the optimism with that incentive in mind; see Lens 9.)

Comps

Peer table — Microchip vs the analog/MCU complex. Microchip's own multiples are from arithmetic on filing figures + the live $102.23 price; peer multiples are, June 2026.

5-yr avg ROE: n/a (would require a 5-year equity/NI series the shelf doesn't hold). Dividend yield: ~1.78% [est: $1.82 declared annual ÷ $102.23]; IR cited ~2.5% at a lower price.

Read: the entire analog group trades at mid-20s EV/EBITDA on depressed/recovering earnings — these are recovery multiples, not value multiples. Microchip at ~26.5x run-rate non-GAAP EV/EBITDA and ~37x forward run-rate P/E is roughly in line with the group, not a discount — despite carrying the most leverage and the least-covered dividend of the cohort. There is no valuation cushion here: you are paying full analog-upcycle price for an early-upcycle balance sheet.

Stock-Price Catalysts (>5% moves, last ~5 yrs)

Pattern from the public record:

• 2021–early 2022: post-COVID supply-shortage super-cycle; record revenue ($7.6B FY24 peak run-rate), stock to highs. Catalyst = demand/pricing + LTSA bookings.
• 2022–2023: Fed-driven multiple compression + early signs of inventory build; down-moves on guidance cuts.
• Aug 2024: Play ransomware attack shut fabs, impaired order fulfilment — a discrete operational shock layered on the cyclical down-move.
• Nov 2024: Moorthy retires, Sanghi returns as interim CEO — leadership/turnaround catalyst.
• Late 2024–early 2025: trough — revenue down 45–48% YoY in the worst quarters; stock −35% over 52 weeks (Jan-2025); dividend-cut fear; $1.45B preferred raise (Mar 2025) to delever. Down-moves on revised/cut guidance.
• 2025–2026: V-recovery — sequential beats, "largest booking month in 4 years," data-center wins → +56% in 3 months, ATH $105.91 May-2026.

What the market actually reacts to for MCHP: (1) the inventory/booking cycle (book-to-bill, distributor days) above all; (2) guidance revisions; (3) increasingly, data-center/AI design-win narrative; (4) balance-sheet/dividend survivability at the trough. It is a cyclical bookings stock with a turnaround-and-deleveraging overlay — not a steady compounder being re-rated on fundamentals.

---

Phase C — Judge people & books

Management

Founder-led turnaround — the single most important qualitative fact about MCHP today.

• Steve Sanghi (CEO & President, Chair). Built Microchip: CEO Oct 1991 → Mar 2021, Chair since 1993. Returned as interim CEO Nov 18, 2024 when Ganesh Moorthy retired at 65; made permanent CEO July 2, 2025. Age 69. Track record: turned a tiny memory company into a ~$7.6B-peak diversified embedded leader via a serial-acquisition playbook (Atmel, Microsemi, Microchip's SST) — quantified: decades of dividend growth, the analog/MCU consolidation machine. He is the architect of the model now under stress.
• Capital-allocation history — mixed, and central to the bear case. The acquisition binge built the franchise but loaded $5.5B of debt and $8.7B of goodwill+intangibles (Lens 10); the legacy of that strategy is negative ~$2.3B tangible common equity [est]. On the income side, capital return has been aggressive: $8.61B cumulative common dividends to date, ~$2.4B of buybacks historically. ROE/ROIC cratered in the trough (FY25 ~breakeven) and is rebuilding. The defining recent decision — maintaining and even raising the dividend ($0.455/qtr) through a zero-earnings year, funded partly by issuing the 2026 convertible — is shareholder-friendly signaling but capital-allocation that prioritizes the dividend's 20-year streak over deleveraging speed.
• The nine-point recovery plan: manufacturing optimization (largely complete), inventory normalization, opex discipline (10% headcount cut), and a product push (Switchtec PCIe, 10BASE-T1S, PIC64 RISC-V, data-center connectivity). Execution to date is visible in the numbers — margins and bookings are inflecting on schedule.
• Skin in the game / comp: Sanghi took a 20% salary cut as part of the executive reduction program; comp is salary + 200% MICP + quarterly evergreen RSUs. Insider-ownership % n/a (our figures absent); Sanghi is a large long-tenured holder by reputation but I won't fabricate a figure.
• Archetype: founder/operator brought back to fix his own creation — historically a strong setup (aligned, knows where the bodies are), with the standard risk that the returning founder re-runs the old M&A-and-dividend playbook into a structurally tougher (China-substitution) market.

Red flags (governance): returning-founder/Chair-CEO duality (concentrated power, weak board independence at the top); the dividend-over-balance-sheet bias; and the M&A-built intangible stack that flatters nothing and must amortize down ($431M/yr drag). None are disqualifying; all are watch-items.

Forensic Red Flags

Forensic pass across IS / BS / CF. The accounting is clean and conservative — Microchip is not an aggressive reporter — but the capital structure and a few line items carry real risk.

• Earnings vs. cash flow — favorable, not a red flag. FY26 OCF $962.1M dwarfs net income $230.0M, bridged by D&A $689.3M + SBC $255.4M + a $264.3M inventory release. Cash conversion is strong; no earnings-quality concern on the operating line.
• Stock-based comp flatters non-GAAP — material. SBC $255.4M (FY26), up from $180.4M — ~5.4% of revenue and >FY26 GAAP net income. The Q4 non-GAAP op margin (30.6%) vs GAAP (16.6%) is a 14-point gap, much of it SBC + intangible amortization. Pay for the recovery in GAAP, not non-GAAP.
• Goodwill & intangibles — the balance-sheet's defining feature. Goodwill $6,695.5M + intangibles $2,033.4M = $8,728.9M, against total equity $6,432.4M → tangible common equity ≈ −$2,297M [est: 6,432.4 − 6,695.5 − 2,033.4]. No impairment taken (only $4.1M), and the recovery supports the carrying value — but any re-trough re-opens impairment risk, and book value is accounting fiction here.
• Leverage / liquidity — the real risk. Gross debt $5,536.9M, cash $240.3M → net debt ~$5.3B [est]. Gross debt/EBITDA 4.7x on FY26 GAAP EBITDA (~$1,179M), falling to ~2.4x on Q4 run-rate [est]. They needed covenant relief (max total-leverage stepped to 6.25x mid-2025, ratcheting back to 3.50x) and as of 2026-03-31 were in compliance. Maturity wall: FY27 $386.9M, FY28 $1,000M, FY29 $1,000M, FY30 $1,900M, FY31 $1,250M — chunky but laddered; the 2024 convert ($1.25B, $121.80 strike) is puttable June 2027 if stock < strike (a real refinancing pinch only if the stock round-trips).
• Revenue recognition — standard, low risk. Ship-and-bill with distributor variable consideration; a 100bp price-concession change moves the refund liability only ~$4.0M. No channel-stuffing tell (distributor days are at the low end). Receivables up on the ramp, not outrunning revenue abnormally.
• The hidden liability — Malaysian tax. A Malaysian IRB assessment for FY2020 could mean income tax + penalties up to MYR 1.9B (~$480.2M) if upheld; payment is stayed pending appeal, possible resolution within ~18 months. Microchip "firmly believes it is without merit" and has reserved per ASC 740, but this is a ~$480M contingent cash bomb sitting against a $240M cash balance. The US IRS transfer-pricing dispute (FY2007–2015) was settled Sept 2025 — that overhang is gone; Malaysia is not.

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. No LR or AAER names Microchip Technology in the 2021–2026 window, verified via SEC EDGAR EFTS.
• Item 3 Legal Proceedings (10-K): The material disclosed matters are tax (Malaysian IRB ~$480M exposure; IRS FY07–15 settled) and ordinary-course IP/commercial litigation — no material product-liability or securities action disclosed.
• Non-SEC enforcement (web): No material FTC/DOJ/FDA/CFPB action found. The one notable event is the Aug 2024 Play-ransomware cybersecurity incident (fab disruption + theft of employee contact data and hashed passwords) — an operational/data-security event, SEC-8-K-disclosed, not a regulatory enforcement action; management states it expects no material adverse effect.
• Net: No material regulatory or accounting-enforcement findings — verified via SEC EDGAR EFTS (LR, AAER), 10-K Item 3, and web search as of 2026-06-29. The live risks are financial (leverage, Malaysian tax), not forensic-accounting.

---

Phase D — Project & stress-test

Forward Projection

Built bottom-up from the Q4-FY26 run-rate and explicit guidance. Fiscal years FY27 (ends Mar-2027) → FY29. All inputs labeled; outputs. No our model create in the watchlist loop — logged here as the base call for a future human-gated commit.

Anchor: Q1-FY27 guide non-GAAP EPS $0.67–0.71 (mid $0.69), revenue mid $1.456B. Annualized run-rate ≈ $2.76 non-GAAP EPS / ~$5.8B revenue — and the cycle is still climbing, so a full-year FY27 sits above the Q1 annualization.

Arithmetic basis: Base FY27 = Q4 run-rate ($2.28 non-GAAP annualized) stepped up by the guided Q1 jump ($2.76 annualized) and held roughly flat through the year ≈ $2.85 [est]. GAAP EPS runs ~$1.20–1.60 below non-GAAP on intangible amortization + SBC. At $102.23, the base case = ~36x FY27 non-GAAP / ~27x FY28 [est] — i.e. the recovery to mid-cycle earnings is already in the price; you are underwriting the bull path to make money from here.

Brier base call (for a later our position log commit, not logged now): "MCHP FY27 (Mar-2027) non-GAAP EPS ≥ $2.75," p ≈ 0.60, resolves 2027-03-31.

Bull vs Bear

Bull case. A high-quality, sticky analog/MCU franchise is coming out of the worst inventory correction in a decade with a clean channel (distributor days at the low end), the largest bookings month in four years, non-GAAP operating margins already back to 30.6%, and a credible founder running a visibly-working nine-point turnaround. The mix is upgrading (analog + data-center connectivity + automotive/industrial >50%), the highest-margin licensing line is compounding, and the whole analog complex is entering a pricing upcycle (TXN/NXP raising prices). Incremental margins off a high-fixed-cost owned-fab base are explosive in recovery — a few more quarters of restock could drive earnings through the prior peak. Deleveraging is happening fast (4.7x → ~2.4x), the IRS overhang is resolved, and the dividend's 20-year streak is intact.

Bear case (permanent-impairment risks). (1) You're paying peak multiples for trough-cycle earnings — ~37x forward non-GAAP, ~26.5x run-rate EV/EBITDA, in line with peers despite the worst balance sheet in the cohort; the easy +56% has already happened. (2) China domestic substitution structurally erodes the low/mid-end MCU + analog franchise that is Microchip's core — the one threat the moat doesn't repel, and it is accelerating regardless of the cycle. (3) The dividend ate 113% of FY26 FCF and 429% of GAAP earnings, funded partly by issuing converts — a re-trough forces the choice between the dividend streak and the balance sheet, and a cut would be a violent de-rating. The Malaysian ~$480M tax bomb sits against $240M cash.

Pre-mortem (18 months out, thesis broke): the restock was a one-time channel-refill, not end-demand; bookings rolled over by late FY27; China substitution took share in the high-volume MCU tail; the stock round-tripped below the 2024 convert strike, triggering the June-2027 put and a refinancing scramble into a $5.5B debt stack; the dividend was cut to protect covenants; the multiple compressed from 37x to ~18x on lower-and-cyclical earnings → a 50%+ drawdown.

Are multiples too high? Yes, on any normalized basis — the market has fully priced the recovery to mid-cycle and is paying analog-leader multiples for a more-levered, more-China-exposed, mid-tier-moat name.

Contrarian view (what the market refuses to see): the bookings inflection is real but largely a channel-restock pull-forward after distributors ran days to the low end — the market is extrapolating a refill as a new demand super-cycle. The durable question isn't the next two quarters (those are bagged); it's whether China substitution caps the through-cycle earnings power below where a 37x multiple makes sense. The bull case needs end-demand to take the baton from restock and China to not bite — two things, both required.

Devil's Advocate (short-seller)

Dismantling the bull case. What structurally breaks the model:

• Revenue concentration → channel, not customer. No customer >12%, but ~47% flows through distributors with no long-term agreements, terminable at will. The "largest booking month in 4 years" is distributor restocking off 26-day lows — by definition a refill, not durable demand. When days normalize back to the mid-30s, the sell-in tailwind reverses into a headwind. The bullwhip that took revenue down 42% is the same mechanism now flattering bookings.
• The most dangerous competitor bulls underestimate: China. Microchip names it in its own 10-K — competitors that "copied, cloned, pirated or reverse engineered" its lines in China and Taiwan. With ~18% of revenue into China and Beijing pushing domestic-silicon substitution, the low/mid-end MCU + commodity analog tail — the volume core — is exactly what local players target first. This is structural share loss that no inventory recovery fixes.
• Worst capital-allocation move: borrowing to pay the dividend through a zero-earnings year. FY26 cash dividends ~$1.10B (common $984M + pref $108.5M) against $230M net income, partly financed by the $900M 2026 convertible — defending a dividend streak at the cost of the balance sheet during the worst of the trough. The $1.45B 7.50% preferred raised in 2025 is a 7.5% cost-of-capital admission they were near the edge.
• Assumptions that must hold for $102: (1) restock → genuine end-demand super-cycle; (2) China substitution doesn't cap through-cycle earnings; (3) GM climbs back toward 65%+ and holds; (4) the stock stays above the 2024 convert strike to avoid the June-2027 put; (5) the Malaysian ~$480M tax goes away. Break any two and the thesis cracks.
• If growth disappoints 20–30%: FY27 non-GAAP EPS falls from ~$2.85 base toward the ~$2.10 bear; at a de-rated 18–20x that's a ~$40 stock — a 50%+ drawdown from here.
• Single permanent-impairment scenario (and plausibility): a structural China-substitution shift that permanently lowers Microchip's through-cycle MCU/analog earnings power while a re-trough forces a dividend cut and a refinancing of the converts into a tougher market. Plausibility: medium — not a base case, but far from tail, and it is the asymmetric risk the +56% run is ignoring.

Management Questions (ordered by information value)

1. Of the Q4/Q1 bookings surge, how much is distributor restock vs. true end-customer sell-through, and what distributor-days level do you consider steady-state?
2. What is your through-cycle, normalized operating margin and EPS power for this portfolio at mid-cycle utilization — not the recovery peak?
3. Where are you losing or defending share against Chinese domestic-substitution competitors in low/mid-end MCU and commodity analog, and how big is that revenue tail?
4. At what stock price / date does the 2024 convertible put (June 2027) become a live refinancing problem, and what's the plan if the stock round-trips?
5. Walk through dividend policy priority vs. deleveraging: at what leverage or FCF level would you cut, and why was issuing the 2026 convert to fund the dividend the right call?
6. What is the realistic range and timing of the Malaysian IRB matter, and how would a ~$480M cash outflow be funded against a $240M cash balance?
7. What is the data-center / AI revenue run-rate today (PCIe switches, CXL, storage controllers), its growth, and its margin vs. the corporate average?
8. How much of the gross-margin recovery is utilization/unabsorbed-capacity reversal vs. durable mix, and where does GM plateau?
9. When do you resume Fab 4 / Fab 5 (SiC) expansion, and what does that do to capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. and free cash flowFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices.?
10. Buybacks vs. dividend vs. debt paydown — with ~$1.56B authorization remaining and no FY26 repurchases, what's the order of operations from here?
11. What is the right long-run debt level for this business, and is the M&A-and-leverage playbook that built the company still appropriate in a China-substitution world?
12. How exposed is the order book to tariffs / export controls given ~18% China and ~15% Taiwan revenue, and what's the mitigation?
13. What's the succession plan behind a 69-year-old returning founder-CEO, and what is the board's independence posture given the combined Chair/CEO role?
14. How durable is ASP stability on proprietary lines if a prolonged downturn returns, and how much pricing did you concede on LTSA cancellations?
15. What are the 2–3 metrics you would point to in 18 months to prove the turnaround is structural, not just a cyclical bounce?

---