STMicroelectronics

Phase A — Understand the business

Lens 1 · Company Overview

STMicroelectronics N.V. is a top-tier European semiconductor IDM — it designs and fabricates, owning 14 main manufacturing sites and ~140,000 wafer-starts/week (200mm-equiv) of front-end capacity. 48,000 employees; ~200,000 customers; >21,000 active/pending patents; R&D 17.3% of revenue ($2.04B). It is incorporated in the Netherlands, HQ'd operationally in Geneva, and triple-listed (NYSE: STM; Euronext Paris CAC 40; Borsa Italiana FTSE MIB).

How it makes money. Selling chips into four end-markets — Automotive, Industrial, Personal Electronics, and Communications Equipment/Computers & Peripherals — across four reportable segments (reorganized eff. Jan-2025):

• AM&S (Analog, MEMS & Sensors) — FY25 rev $5,085M (-6.3%)
• P&D (Power & Discrete, incl. SiC) — $1,685M (-31.5%)
• EMP (Embedded Processing — STM32 MCUs, automotive MCUs, ADAS, secure) — $3,580M (-7.1%)
• RFOC (RF & Optical Comms — space/LEO, optical interconnect, audio/positioning) — $1,436M (-4.9%)

Contract structure. Mostly point-of-shipment product sales (30–90 day terms); 72% OEM / 28% distribution. Notably, STM increasingly runs multi-annual capacity-reservation and volume-commitment agreements with key customers (binding take-or-pay-style commitments with commitment fees recorded in revenue) — a post-COVID structural shift, especially as carmakers now contract chips directly rather than only through Tier-1s. Backlog is short-cycle and volatile; ST entered 2026 with a higher backlog than it entered 2025 — a green shoot.

Concentration flag (carry to Lens 13): largest customer Apple = 17.7% of FY25 revenue, up from 14.5% (2024) and 12.3% (2023), spread across all four segments. Customer concentration is rising precisely because the rest of the book (autos/industrial) fell faster.

Lens 2 · Supply Chain

Upstream inputs → ST → end customer, named at every node:

• Equipment suppliers (front-end steppers/scanners/etchers/implanters; back-end bonders/testers) — unnamed but the usual ASML/Applied/Lam/TEL cohort ``. ~$1.3B of equipment on order for 2026 delivery.
• Materials — silicon, SiC, GaN and glass wafers, lead frames, mold compound, chemicals, gases, water; explicit "multiple-sourcing strategy" to blunt price/disruption risk.
• Foundry partners (outsourced ~25% of silicon value in 2025): STM is a hybrid IDM — it buys leading-edge logic/FinFET from external foundries (TSMC/GlobalFoundries-class) ``, while running its own analog/power/MEMS/SiC fabs. Foundry purchase commitments total $774M.
• Internal front-end fabs: Agrate (Italy, incl. 300mm Fab 3 shared with Tower Semiconductor), Catania (Italy, SiC), Crolles (France, 300mm), Rousset/Tours (France), Norrköping (Sweden, SiC substrate), Ang Mo Kio (Singapore).
• Internal back-end (assembly/test): Bouskoura (Morocco), Calamba (Philippines), Kirkop (Malta), Muar (Malaysia), Marcianise (Italy), Rennes (France), Shenzhen (China, JV with Shenzhen SEG).
• Strategic JVs: GlobalFoundries (Crolles 300mm, €7.5B project) and Sanan Optoelectronics (Chongqing, China — captive 200mm SiC foundry, ~$3.2B buildout); GaN development pact with Innoscience (STM holds a ~$127M stake).
• End customers (named): Apple, Bosch, Continental, Denso, HP, Mobileye, Samsung, SpaceX, Tesla, Vitesco. SpaceX ties to RFOC/space; Tesla/Bosch/Continental/Denso/Vitesco are the auto spine.

Chokepoints / single-source: (1) Apple as a single 17.7% node; (2) European energy — multi-year PPAs (TSE solar in France) and a $1.3B power/efficiency commitment underpin the carbon-neutral-by-2027 goal but also lock in cost; (3) geographic concentration of PP&E in Italy ($2.6B) + France ($1.5B) + Netherlands ($4.1B) — a high-cost-labor manufacturing base exposed to European energy and politically-sensitive restructuring.

Lens 3 · Competitive Advantages (moats)

What protects it:

• IDM + proprietary process estate. STM owns differentiated process tech — FD-SOI (it shipped the first 18nm FD-SOI MCU, STM32V8, with embedded phase-change memory), BCD/VIPpower smart-power, SiC, GaN, BiCMOS, SiPho. Combining design + fab is a real differentiator vs. fabless peers for analog/power/automotive parts that need long-lived, qualified, automotive-grade processes.
• STM32 ecosystem (the crown jewel). The broadest general-purpose 32-bit Arm-Cortex-M MCU franchise, with a 140+-model edge-AI "model zoo," TouchGFX, STM32Cube tooling — high developer switching costs and design-win stickiness. This is the closest thing STM has to a durable platform moat.
• Automotive qualification + design-in cycles. Auto parts are designed in for 5–10 years; ST is a leading automotive analog/MCU/SiC supplier with direct OEM relationships now (not just Tier-1).
• Government-funded capex. EU Chips Act "Integrated Production Facility" status (Oct-2025); €2B Italian grant on a €5B Catania SiC plant; ~€2.9B French support on the €7.5B Crolles JV; Chinese support on Sanan SiC; IPCEI grants. State backing lowers ST's effective cost of capacity — a structural subsidy peers without European footprints don't get.
• Scale: 21,000+ patents, $2B/yr R&D, 200k customers.

Bargaining power — honest read: weak-to-mixed. Against Apple (17.7%) and large carmakers contracting directly, ST is increasingly price-taking — FY25 revenue fell on a ~6% ASP decline plus volume, and P&D ASPs fell ~29%. In commodity power/discrete and SiC, Chinese entrants (see Lens 13) are compressing price. The moat is real in differentiated MCU/FD-SOI/automotive-analog, thin-to-absent in commodity discretes and merchant SiC.

Lens 4 · Segments

Revenue & operating income by reportable segment [all research-layer: filings/20-f-2025-q4.md]. Note: the filing's "% of net revenues" column for segment OI is garbled in places (e.g. it prints P&D at "14.7%" against a -$275M loss); the correct operating margins, computed from the sourced $-figures, are below ``:

Trend & cause:

• P&D is the wound — it swung from +$263M OI to a -$275M loss as SiC/power ASPs collapsed ~29% on EV-slowdown + Chinese price pressure. This is decelerating and margin-negative.
• RFOC is the bright spot — highest margin (18.5%), and the segment that houses the AI-data-center/optical-interconnect and space/LEO growth (Q4'25 RFOC rev +30.5% q/q).
• AM&S and EMP held double-digit margins but earnings roughly halved-to-quartered off 2023 — operating deleverage on lower volume.
• The "Others" hole grew to -$974M (from -$525M / -$414M), driven by unused-capacity charges $416M (vs $370M / $120M) + the $376M restructuring — i.e. underutilization is the swing factor crushing consolidated margin from segment-level ~10% to reported 1.5%.

Geography (by shipment): Asia-Pacific $7,455M (63%), EMEA $2,449M (-26.4%, the hardest hit), Americas $1,896M. By billing entity, Singapore ($6.56B) and Switzerland ($2.95B) dominate — a tax/treasury structure, not where value is made.

---

Phase B — Measure performance

Lens 5 · Earnings Result

FY2025 (year ended 2025-12-31) — a trough year [all research-layer: filings/20-f-2025-q4.md]:

• Net revenue $11,800M, -11.1% (ASP -6%, volume -5%). On top of 2024's -23.2%, revenue is down ~32% from the 2023 peak.
• Gross margin 33.9%, -540bps (lower manufacturing efficiency, mix, lower capacity-reservation fees, FX, higher unused-capacity charges). Down from 47.9% at the 2023 peak — a ~1,400bp gross-margin destruction over two years.
• Operating margin 1.5% GAAP / 4.7% non-GAAP (after stripping $376M impairment+restructuring). Operating income $175M GAAP vs $1,676M in 2024.
• Net income $166M; diluted EPS $0.18 GAAP / $0.53 non-GAAP (vs $1.66 / $4.46 in 2024 / 2023).
• Q4'25 was a net LOSS of $(30)M, -$0.03 EPS GAAP (non-GAAP +$0.11) — the trough quarter.
• Balance-sheet flags — all benign: net cash position $2,789M; total liquidity $4,922M vs $2,133M debt; FCF still positive $265M; no covenant breaches; goodwill only $315M (no impairment). The downturn hit the P&L, not solvency.
• Effective tax rate spiked to 55% (vs 17%/11%) on a $80M valuation-allowance build + $66M uncertain-tax-position charge — flattered-then-punished by the low pre-tax base. Watch item, not a red flag.

Q1 2026 (reported Apr-23-2026) : revenue **$3.10B (+23% y/y, +21.4% ex-NXP-MEMS)**, GM 33.8%, net income $37M, **GAAP EPS $0.04 / non-GAAP $0.13** — the non-GAAP figure *missed* the ~$0.18 consensus by ~27%. **Q2'26 guide: revenue $3.45B, GM ~34.8%** (incl. ~100bps unused-capacity drag) .

Market reaction — the tell: the stock surged ~8% on the AWS announcement (Feb-2026) and held up despite the Q1 EPS miss ``. The market has stopped trading STM on the print and started trading it on the recovery slope + AI optionality. That is the single most important behavioural fact in this dossier.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the shelf (transcripts/ empty); synthesized from filing MD&A + web ``:

• 2024 calls: defensive — "automotive woes," inventory correction (mgmt attributed ~60% of the decline to inventory correction, ~30% to demand), repeated guidance cuts; SiC 2025 target cut $2.0B → $1.8B ``.
• H2-2025 → Q1-2026 calls: tone inflected to transformation + recovery — "manufacturing reshaping," "$800M cost savings by 2027," sequential revenue growth, backlog up entering 2026, and a loud new AI-data-center pivot (silicon photonics, doubled DC revenue target). The recurring new phrase is "AI/data-center"; the phrase they stopped leaning on is "automotive growth."
• Net: management narrative has rotated from apology to ambition. Credible on the cost program; the AI framing is real but small relative to the base (Lens 11/13).

Lens 7 · Comps

Peer set: broad-line analog/power/auto/MCU IDMs. Multiples are ``, June 2026, and should be treated as approximate — sources diverge widely on STM's own forward P/E (a hazard of forecasting through a trough):

Read: the entire power/auto-analog cohort has re-rated hard in 2026 (Infineon EV/EBITDA 11→27x; TXN +80%; STM 16.94 → ~75). On 2027 consensus, STM (~27x) is in line with TXN-ish quality multiples despite materially lower through-cycle margins and ROE — i.e. STM is priced like a premium analog franchise on recovery math. GuruFocus flags it "significantly overvalued, ~124% above fair value" ``.

Lens 8 · Stock-Price Catalysts (moves >5%, last ~5y) ``

• Jul-25-2024: -15.3% to $33.47 on the Q2-2024 revenue-guidance cut — the event that triggered the securities class action ``.
• 2024 full year: -50%; trough $16.94 on Apr-7-2025 (-57% from the mid-2024 peak) `` — the macro/tariff + auto-inventory capitulation low.
• H2-2025: steady recovery as inventory correction wound down ``.
• Feb-9-2026: +~8% on the multi-billion AWS AI-data-center deal + Amazon equity warrants ``.
• Apr-23-2026: stock rose despite a Q1 EPS miss — recovery + AI narrative dominating ``.
• To ~$74.79 by late June 2026 ``.

Pattern: STM trades on (1) forward revenue/guidance (cuts punish brutally, an inflection re-rates violently) and (2) single big-customer/secular headlines (the AWS deal is the cleanest example). It is a high-beta cyclical where the direction of estimate revisions matters far more than the absolute print.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO Jean-Marc Chery (President & CEO; reappointed to 2028) — long-tenured ST insider, an engineer/operator who has run ST through the up-cycle and now the restructuring. Track record: delivered peak 2023 (47.9% GM, $4.46 EPS), then presided over the -32% revenue collapse and the first loss-making quarter in years. 2025 comp $5.64M, down from $9.47M (2024) — pay fell with performance.
• CFO Lorenzo Grandi (President & CFO) — new to the board, appointed May-22-2024, i.e. installed into the downturn; 2025 comp $2.81M, mostly fixed (no LTI vests until 2027).
• Skin in the game / alignment — mixed. Modest ownership guideline (1.5× base salary); 2024 STI achievement 60% (CEO) / 42% (CFO) vs 205% in 2022–2023 — so incentive pay genuinely tracked the collapse (good). Clawback provisions exist; no clawbacks triggered.
• Capital allocation: disciplined-but-pressured — maintained the $0.36 dividend and ran a $367M buyback in 2025 while still generating positive FCF and raising 2026 capex to $2.0–2.2B counter-cyclically into SiC/300mm. ROE collapsed with earnings (net income $166M on ~$17.8B equity ≈ <1% ROE in 2025 ``, vs mid-20s% at peak). The counter-cyclical capex is a bet on the up-cycle — right if the recovery is real, value-destructive if SiC stays oversupplied.
• Red flags (carry to Lens 10/13): (1) the securities class action naming Chery + Grandi survived a motion to dismiss; (2) April-2025 Italian-press allegations that the two board members sold stock "on the eve of earnings" — the company says these were automatic sell-to-cover transactions for Swiss tax (52,810 shares sold via the plan administrator in 2025, all tax-related) and the Supervisory Board reaffirmed support. Plausible explanation, but the optics are poor and it is now litigation exhibit material.
• Archetype: professional managers of a state-influenced national champion, not founder-owners. Governance answers to the French/Italian state bloc (Lens 13) — employment and industrial-policy objectives sit alongside shareholder returns.

Lens 10 · Forensic Red Flags

Grounded in the 20-F + regulatory/regulatory-findings.md and web ``.

Accounting-quality scan (FY2025):

• Revenue recognition: point-of-shipment; price-protection/return reserves for distribution (28% of sales) are estimated — standard but a judgment area. The multi-annual capacity-reservation commitment fees booked into revenue are worth monitoring (timing/quality of revenue) but disclosed.
• Cash vs earnings: clean — operating cash flow $2,152M dwarfs $166M net income (D&A $1,854M + the unused-capacity/restructuring non-cash charges). No divergence concern; if anything earnings understate cash generation in the trough.
• Inventory/receivables: no customer >10% of receivables; inventory written down via reserves; underutilization costs expensed straight to COGS rather than capitalized into inventory (conservative).
• SBC: $193M in 2025 (down from $222M/$236M) — modest (~1.6% of revenue), not flattering non-GAAP aggressively. The $0.53 vs $0.18 non-GAAP gap is the $376M restructuring, not SBC.
• Goodwill/intangibles: tiny ($315M goodwill, $324M amortizable intangibles); annual test passed; impairments in 2025 were on PP&E and a license tied to the footprint reshaping ($189M), appropriately taken.
• Tax: ETR 55% on valuation-allowance + uncertain-tax-position builds; under transfer-pricing audit in multiple jurisdictions; uncertain tax positions grew to $142M; deferred-tax-asset recoverability is a judgment lever.

Regulatory findings (required):

• SEC LR/AAER: none — verified via SEC EDGAR EFTS (LR + AAER) search 2021–2026.
• Securities litigation (material): two consolidated class actions (SDNY, filed Aug-23-2024) against ST + CEO + CFO allege false/misleading statements about 2024 revenue (class period Mar-14-2023 → Jan-29-2025). The court DENIED the motion to dismiss on Sep-15-2025 — the case proceeds; ST says it has strong defenses. This is the single most material legal overhang.
• Patent (10-K Item 3 / Note 26): Purdue University won a $32M jury verdict (W.D. Tex., Jun-2024); ST is appealing; loss assessed "possible," reserved accordingly.
• Non-SEC enforcement (FTC/DOJ/antitrust): web search returned no material antitrust/regulatory enforcement actions — only plaintiff-firm securities solicitations ``.
• Net: accounting quality is clean/conservative; the real Lens-10 risk is securities-litigation + governance-optics, not the books.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Built up from sourced actuals + guidance; outputs `` with arithmetic shown. ST's fiscal year = calendar year. No forecast.ts create (watchlist rule).

Anchors: FY25 revenue $11.8B, GM 33.9%, non-GAAP OpM 4.7%, ~915M diluted shares. Q1'26 $3.10B; Q2'26 guide $3.45B (GM 34.8%) → H1'26 ~$6.55B run-rate . Cost program targets "high triple-digit million" / ~$800M annual savings exiting 2027. AWS warrants add ~2.7% dilution as they vest at $28.38 .

Inputs: base assumes the cyclical recovery restores ~38% GM (still below the 47.9% peak), the ~$800M cost program lands, unused-capacity charges fade, and data-center revenue reaches ~$1B ` for the build]. Base FY2027 non-GAAP EPS ~$2.50–2.80 — i.e. roughly back to 2024 levels, well short of 2023's $4.46.

Reconciliation to price (the crux): at ~$74.79, the stock trades at ~27x base-case FY2027 EPS, ~45x on 2024's $1.66, and ~17x only if you put it on the 2023 peak $4.46. Consensus PT ~$64 sits below spot. EV/FY25-sales ≈ 5.5x — a rich multiple for a low-double-digit-through-cycle-margin European IDM. The market is paying a premium multiple on recovered-but-not-peak earnings, with the AI story as the justification.

Lens 12 · Bull vs Bear

Bull case. (1) Genuine cyclical bottom — Q4'25 was the trough; revenue inflected +23% y/y in Q1'26; backlog up entering 2026; auto/industrial inventory correction is over. (2) Self-help margin lever — the manufacturing reshaping (300mm Agrate/Crolles, 200mm SiC Catania) + ~$800M cost-out should restore GM toward the high-30s/40 as utilization normalizes, with operating leverage amplifying EPS off the trough. (3) AI/optical optionality — the AWS multi-billion deal validates ST's silicon-photonics/optical-interconnect and power positioning for data centers; DC revenue target doubled to ~$1B. (4) Government-subsidized capacity (EU Chips Act IPF, €2B Catania grant, €2.9B Crolles support) lowers the cost of the build. (5) Fortress balance sheet — net cash $2.8B, FCF-positive even at the bottom — pays the dividend and buys back through the cycle. (6) STM32 franchise — a durable MCU/edge-AI ecosystem moat.

Bear case (permanent-impairment risks). (1) SiC/power structurally oversupplied — P&D already loss-making (-16.3% margin), ASPs -29%, with Chinese entrants (BYD Semiconductor, StarPower) + Wolfspeed/onsemi flooding capacity into a slower-than-promised EV ramp; ST's own SiC targets have already been cut. The power-cycle may not return to prior margins. (2) Rising single-customer dependence — Apple 17.7% and climbing; an Apple socket loss or pricing reset would gut AM&S. (3) European cost base + politics — restructuring European fabs/labor is slow and politically constrained (the state owns 27.5%); the cost-out may under-deliver. (4) The valuation itself — at ~27x recovered EPS / 5.5x trough sales / above the consensus target, expectations are demanding; any 2026–27 estimate cut de-rates the multiple and the earnings simultaneously.

Pre-mortem (18 months out, thesis broke): auto/industrial recovery stalls in H2-2026, SiC pricing stays in a China-driven price war so P&D never turns, the AWS/optical ramp proves smaller and later than the headline, GM plateaus in the mid-30s instead of marching to 40, and the AI re-rating unwinds — STM round-trips from ~$75 back toward the $40s as the multiple normalizes to a cyclical-IDM 14–18x on ~$2 EPS.

Are multiples too high? On any normalized (non-peak) earnings power, yes — the AI narrative is doing a lot of multiple work on a franchise whose through-cycle margins are structurally below TXN/Infineon. Contrarian view the market is refusing to see: STM is being priced as an AI-infrastructure play when 85%+ of revenue is still cyclical auto/industrial/consumer analog-and-power facing Chinese competition — the AI tail ($1B) is wagging a $14–15B dog.

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

Dismantling the bull:

• Revenue concentration & the wrong kind of growth: the book is concentrating into Apple (17.7%, rising) while the diversified auto/industrial base shrank — "diversified IDM" is becoming "Apple + a cyclical tail." A consumer-electronics socket is lower-margin and re-competed annually.
• The moat is thinnest exactly where the growth is supposed to come: merchant SiC is a commoditizing, government-subsidized global capacity race; ST is not the cost leader (its European fabs are) and China (BYD/StarPower) + Wolfspeed/onsemi are structurally undercutting it. P&D's -$275M loss is the canary.
• Most dangerous competitor bulls underestimate: Chinese domestic power/SiC + MCU vendors, backed by state capex and captive EV demand, eroding ST in its largest region (APAC 63% of sales) — a share-and-price double-hit ST can't tariff away.
• Capital-allocation/governance: counter-cyclical capex into an oversupplied SiC market; a 27.5% French/Italian state bloc + a Stichting poison pill (up to 540M pref shares) mean (a) zero M&A/takeover optionality to crystallize value, and (b) governance that weighs European jobs against shareholder returns. Add the live securities-fraud class action (survived dismissal) and the insider-sale optics — the incentive/disclosure picture is not pristine.
• What must hold for ~$75: a full cyclical recovery to ~$15.5B+ revenue and GM marching back toward 40 and the AI/optical ramp delivering and the multiple staying at ~27x. If 2026–27 revenue disappoints by 20–30%, EPS misses badly (operating deleverage on a high-fixed-cost IDM), and on a cyclical 15x the stock has ~40–50% downside.
• Single scenario that permanently impairs: SiC becomes a structurally low-return commodity (Chinese oversupply + EV demand undershoot), P&D stays loss-making, and the European cost base can't be flexed fast enough — turning ST into a sub-30%-GM, low-ROE utility-like national champion that the market re-rates to tangible book. Plausibility: moderate and rising.

Lens 14 · Management Questions (ordered by information value)

1. P&D lost $275M in 2025 at -16.3% margin — at what SiC/power pricing and utilization does it return to profit, and what is your honest read on Chinese SiC oversupply over the next 24 months?
2. The AWS deal: what is the revenue ramp profile (2026 → end-of-decade), the margin vs corporate average, and how do the warrants' vesting/dilution interact with the economics?
3. Apple is now 17.7% of revenue and rising — what is the socket/program visibility, and how do you defend against an Apple pricing reset or design loss across all four segments?
4. The ~$800M cost program "exiting 2027": how much is structural vs cyclical, how much European headcount/footprint reduction does it require, and what's the political/works-council risk to the timeline?
5. What through-cycle gross margin do you now underwrite post-reshaping — is 40%+ realistic, or is the new normal mid-to-high-30s given mix and Chinese competition?
6. Counter-cyclical capex of $2.0–2.2B into SiC/300mm — what utilization/demand assumptions justify it, and at what point do you cut capacity if EV/SiC demand undershoots again?
7. How dependent is the margin recovery on EU/Chinese state funding (Catania €2B, Crolles €2.9B, Sanan), and what happens to unit economics if subsidies are clawed back or delayed?
8. On the SDNY securities class action (motion to dismiss denied) — what is the realistic exposure and timeline, and what disclosure-process changes have you made?
9. The April-2025 insider-sale allegations: walk through the sell-to-cover controls and why automatic blackout-period sales are appropriate for board members.
10. Data-center/optical (RFOC) is your highest-margin, fastest-growing segment — would you carve it out or report it more granularly so the market can value it separately from the cyclical core?
11. With a 27.5% state bloc and a poison pill, how do you think about shareholder-value crystallization absent any M&A path — bigger buybacks, a higher payout, a structural separation?
12. Backlog entered 2026 higher than 2025 — how much is genuine end-demand vs distributor restock, and what's the cancellation risk if macro/tariffs deteriorate?
13. STM32: how do you monetize the edge-AI ecosystem (the 140-model zoo, NPU MCUs) beyond unit sales — is there a software/licensing layer?
14. FX: at ~$1.16/€ the cost base hurts — how much of the margin recovery is operational vs a euro tailwind, and how are you hedged through 2027?
15. Capital return: is the $0.36 dividend a floor, and how do you prioritize dividend vs buyback vs capex if FCF stays compressed through 2026?

---