Infineon Technologies

Phase A — Understand the business

Company Overview

Infineon is the world's #1 power-semiconductor company and #1 automotive-semiconductor company — a German IDM (integrated device manufacturer: it designs and fabs its own chips) that sells the discrete power switches, microcontrollers, sensors and security chips that move and control electricity. It is not a logic/compute company; its franchise is energy conversion and control. FY2025 revenue was €14.662bn, −2% YoY, the second down-year of an automotive/industrial inventory correction.

How it makes money. Physical chips sold to OEMs and their tier-1 suppliers, mostly on design-win/backlog rather than take-or-pay — a socketed part (e.g. an EVEnterprise valueWhat it would cost to buy the whole company: its market value plus its debt, minus the cash you would get with it. Often very different from market cap. traction inverter module, a MOSFET, an eSIM) once designed into a platform generates multi-year recurring unit volume. There is no meaningful subscription/services line; the "recurrence" is the multi-year design-in lifecycle of automotive and industrial platforms (5–10 year model cycles), which is what makes the backlog visible and the cyclicality slow-moving.

The reorganization — read this first. Effective 1 July 2026 (five days before this dossier), Infineon collapsed its long-standing four divisions into three:

• Automotive (~50% of revenue) — the old ATV plus the Marvell Automotive-Ethernet business.
• Power Systems (~30%) — power for AI datacenters, industrial, renewables (the old GIP + most of PSS power).
• Edge Systems (~20%) — sensors, connectivity, security, microcontrollers (the old CSS + PSS sensing/connectivity).

Management framed it as moving "beyond a product-centric approach toward solutions based on system understanding". Analytically this matters: it is Infineon formally re-organizing around the AI-datacenter-power opportunity — carving "Power Systems" into its own P&L so the market can see the datacenter ramp cleanly. This dossier analyzes the historic four-segment structure (where the disclosed numbers live) and maps to the new three where relevant.

Key products: IGBTs and IGBT modules, SiC (CoolSiC™) and GaN (CoolGaN™) power devices, automotive microcontrollers (AURIX™), traction inverters, MOSFETs, radar sensors, power-management ICs and DC-DC converters, secure elements / eSIM (the Cypress-derived CSS franchise), and now automotive Ethernet PHYs (Marvell).

Main customers: the global auto complex (tier-1s: Bosch, Continental, ZF, Denso, Vitesco; OEMs directly for EV platforms), industrial/renewables (solar inverter makers, HVAC, industrial drives), and — the new leg — AI-datacenter power supply chains anchored by NVIDIA (Infineon is a named partner in NVIDIA's MGX 800V HVDC ecosystem).

Competitors: STMicroelectronics, onsemi, NXP, Texas Instruments, Renesas, ROHM, Vishay, Power Integrations, Navitas (in power); NXP and Renesas (in automotive MCUs); for SiC specifically STMicro, onsemi, ROHM and a rising wall of Chinese players (SICC, TanKeBlue, BYD's captive fabs).

Supply Chain

Upstream → Infineon → end customer, named at every node. web-only (no supply-chain.md on the shelf; KB wiki absent).

Upstream inputs:

• SiC substrates/wafers — historically bought from Wolfspeed, Coherent (II-VI), SK Siltron, ROHM/SiCrystal; increasingly Chinese suppliers SICC, TanKeBlue. Infineon signed multi-year SiC substrate supply deals to de-risk (it is not fully vertically integrated in SiC crystal growth, unlike onsemi/Wolfspeed). This is a genuine dependency and a cost lever — Chinese substrate entry drove a ~30% SiC wafer price drop in 2024.
• Silicon wafers — standard 200mm/300mm silicon from Siltronic, SUMCO, Shin-Etsu.
• Foundry (partial) — Infineon is an IDM (fabs most power devices in-house at Villach (Austria), Dresden (Germany), Kulim (Malaysia), Regensburg, Cegléd) but outsources some advanced-logic/MCU wafers to TSMC and UMC. It is building a major new 300mm frontend in Dresden ("Smart Power Fab," ~€5bn, EU Chips Act supported).
• Manufacturing equipment — ASML, Applied Materials, Lam, Tokyo Electron, AIXTRON (for SiC/GaN epitaxy).
• Back-end/OSAT — in-house assembly/test in Malaysia, plus OSAT partners.

Chokepoints / single-source risks:

1. SiC substrate supply — the biggest input dependency; mitigated by multi-sourcing but exposed to Chinese pricing and any Wolfspeed disruption (Wolfspeed filed Chapter 11 in Sept 2025 ).
2. Kulim 200mm SiC fab ramp — the single most important internal capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. bet; execution here gates the €7bn SiC ambition.
3. TSMC/UMC advanced-node dependency for MCUs — modest but real for the AURIX franchise.

Downstream: Infineon → tier-1 auto suppliers / EV OEMs (design-win socketed), → industrial OEMs (inverters, drives, HVAC), → AI-datacenter power supply chain: Infineon power stage → power-shelf/PSU makers (Delta Electronics, Vertiv, Lite-On) → NVIDIA MGX racks → hyperscalers (Microsoft, Meta, Google, Amazon, CoreWeave). The datacenter node is new and is where the incremental growth narrative lives.

Names or it didn't happen — delivered. The one gap: exact 2026 SiC-substrate supplier mix is not publicly disclosed (n/a).

Competitive Advantages (moats)

1. Scale + IDM cost position in power. Infineon is the volume leader in power discretes/modules; power semis are a process/packaging game, not a bleeding-edge-node game, so Infineon's decades of trench-IGBT and module know-how, plus in-house fabs, give it a durable low-cost, high-reliability position. Automotive-grade qualification (AEC-Q, zero-defect, 15-year field life) is a real barrier — you don't switch a validated traction-inverter part casually.

2. Switching costs / design-in stickiness. Once a part is designed into an auto or industrial platform, it stays for the model life. This is the moat that makes the cyclicality slow and the recovery visible via backlog — but note it cuts both ways (a lost socket is lost for years).

3. Breadth across all three power materials — Si + SiC + GaN. Uniquely, Infineon plays silicon, silicon carbide and gallium nitride (bolstered by the GaN Systems acquisition), which is exactly the pitch to NVIDIA's 800V HVDC architecture ("from grid to core, all relevant semiconductor materials"). Few rivals span all three at scale.

4. Bargaining power. Over customers: strong in automotive power where it is often the reference/lead supplier, weaker in commodity MOSFETs. Over suppliers: moderate — genuinely exposed on SiC substrate (see Lens 2). Net: Infineon needs the auto complex roughly as much as it needs Infineon, but in the datacenter-power land-grab it currently has scarcity value.

Moat verdict: wide and durable in power/automotive (scale, qualification, switching costs, materials breadth); narrower and more contestable in SiC, where Chinese substrate + captive OEM fabs (BYD) are compressing the value pool faster than bulls assumed in 2023.

Segments

our figures is empty → all figures. Historic four-segment structure (pre-1-Jul-2026 reorg).

FY2024 (year ended 30 Sep 2024), ~€15.0bn total:

Q2 FY2026 (quarter ended ~Mar 2026), €3,812m total, +6% YoY / +4% QoQ, segment margin 17.1%:

• Green Industrial Power: €403m, +15% QoQ — rebound off a seasonally weak Q1; driven by power infrastructure, HVAC, home appliances.
• Power & Sensor Systems: +8% QoQ — driven by AI-server/datacenter demand + automotive radar + mobile.
• Automotive and CSS contributed to the beat but exact Q2 FY2026 per-segment revenue for ATV/CSS was not separately disclosed in the coverage I could source (n/a at the euro level; ATV remains ~half of group).

Trend read:

• Automotive — the ~56% anchor; went through a deceleration/decline through FY2024–FY2025 (EV-demand slump + customer inventory digestion), now re-accelerating in FY2026 as backlog recovers. Marvell Ethernet ($225–250m revenue, 60% GM, ~$4bn design-win pipeline to 2030) is bolted on here.
• PSS/Power — the accelerating segment and the AI-datacenter beneficiary. This is why the reorg carved "Power Systems" out.
• GIP — renewables/industrial; cyclically depressed then recovering; secular tailwind (grid, renewables) intact.
• CSS/Edge — steadiest, lowest-margin-volatility (secure elements, eSIM, connectivity); the Cypress-derived franchise.

Geography: majority of revenue is Greater China + Asia-Pacific (the auto/industrial manufacturing base) and Europe; the U.S. footprint grew with Cypress and Marvell. Exact FY2026 geo split n/a.

---

Phase B — Measure performance

Earnings Result (latest print — Q2 FY2026, reported May 2026)

All; press release INFXX202605-082 is the primary document.

• Revenue €3.812bn, +4% QoQ, +6% YoY.
• Segment Result €653m; Segment Result Margin 17.1% (up from 16.7% in Q2 FY2025).
• Vs. consensus: coverage split — one outlet framed it as a "Q2 miss" on the headline print while emphasizing the raised guidance, another as a clean beat-and-raise. Read: the quarter was roughly in line-to-soft, but the guidance raise was the story.
• Drivers: AI-related revenues nearly tripled YoY; GIP +15% QoQ, PSS +8% QoQ; automotive backlog recovering. Drag: "ongoing challenges in high-voltage e-mobility" (i.e. the EV-traction/SiC piece is still soft).
• Guidance / outlook — RAISED: FY2026 revenue now expected to rise significantly YoY (prior: "moderate"), to >€16bn (~>$18bn at 1.17 EUR/USD); adjusted gross margin low-to-mid-40s%; segment result margin ~20%. Q3 FY2026 guided to ~€4.1bn revenue, high-teens segment margin, at assumed 1.17 EUR/USD. AI-datacenter power revenue reaffirmed at €1.5bn FY2026 → €2.5bn FY2027.
• Tone shift: decisively more constructive than the FY2024/early-FY2025 "recovery not yet in sight" posture (see Lens 6). The raise + reorg-into-Power-Systems is management leaning into the AI-power cycle.

Balance-sheet flags (FY2025 close, 30 Sep 2025): adjusted FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. €1.8bn (12.3% of revenue); reported FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. ~€900m; gross debt €6.8bn, net debt €4.7bn, gross cash ~€2.1bn (~14% of sales); net leverage ~1.4x, gross ~2.0x (at self-defined ceiling); FY2025 investments ~€2.5bn. Leverage at the gross ceiling is the one number to watch — it constrains large M&A without equity.

Earnings Calls (sentiment trend)

transcripts/ empty → from call coverage.

Sentiment has swung from defensive to constructive over ~6 quarters:

• Q3/Q4 FY2024 (mid–late 2024): CEO Hanebeck — recovery from the EV slump "not yet in sight"; cut FY2024 guidance by €100m to €15bn; announced 1,400 job cuts; Q3 revenue −9.5%. Recurring theme: customer inventory digestion, weak EV demand, "in line with long-term targets through the cycle" (the defensive framing).
• Q1 FY2025 (Feb 2025): guided FY2025 to a ~2% revenue decline, Q1 down ~18% QoQ on inventory correction. Trough messaging.
• Q2–Q3 FY2025 (mid 2025): "revenue growth confirms expected economic recovery," but flagged tariff-dispute uncertainty and negative currency (weak USD) for H2. Cautiously turning.
• Q1–Q2 FY2026 (early–mid 2026): "successful start to fiscal 2026," "growth prospects improved," full-year guidance raised; the dominant new phrase is AI datacenter power (revenue "nearly tripling," €1.5bn→€2.5bn target).

Phrases picked up: "AI data center power," "800V HVDC," "system solutions," "Power Systems." Phrases dropped: "recovery not yet in sight," "inventory correction/digestion" (now largely past-tense). The tone inflection is real and management-led.

Comps

Provenance-critical. Multiples with source/date or n/a. Never fabricated.

Read: Infineon at ~30.6x fwd P/E / ~27.4x EV/EBITDA sits mid-pack on P/E (cheaper than STM 39.8x and onsemi 35.1x, richer than NXP 20.9x) but rich on EV/EBITDA (27.4x vs onsemi 17.8x, STM ~21x) — the EV/EBITDA premium reflects both the AI-datacenter re-rate and Infineon's higher debt load inflating EV. NXP is the cheap value comp; STM/onsemi are the direct SiC/power comps and both carry cycle-trough-depressed earnings inflating their P/E. The multiple is not screening-cheap — this is not a "buy the laggard on a low multiple" setup; you're paying up for the #1 franchise + AI optionality. 5yr-avg-ROE and several peer multiples are n/a rather than fabricated, per provenance discipline.

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

Mostly.

• Apr 2020 — Cypress acquisition closes (€8.3bn net) — the deal that built the CSS/microcontroller franchise; multi-year re-rate driver.
• 2021–2023 — shortage super-cycle — the great chip shortage drove Infineon to record FY2023 revenue/earnings; the stock ran hard on pricing + auto-content growth.
• Aug 2024 — Q3 FY2024 miss + "recovery not in sight" + FY24 guide cut + 1,400 job cuts — shares fell as much as ~5.9% intraday. The EV-slump air-pocket.
• Nov 2024 — FY2025 guided to revenue decline — reinforced the down-cycle; the 52-week low near €30.82 was made in this window / early 2025 (tariff + EV-demand trough).
• Apr–Aug 2025 — Marvell Automotive-Ethernet acquisition ($2.5bn) announced/closed — strategic re-rating of the automotive-content story.
• 2025–2026 — the AI-datacenter-power pivot (NVIDIA MGX 800V partnership, €1.5bn→€2.5bn AI target, GTC 2025) — the dominant up catalyst; drove the recovery to a 52-week high near €88.83 and the current ~€76.

Pattern: the market reacts to (a) automotive/EV demand inflections (the down-moves) and (b) the AI-datacenter-power narrative + strategic M&A (the up-moves). It is a cyclical franchise re-rating on a secular growth graft — the tape rewards evidence the AI leg is real and punishes any sign the auto/SiC cycle is rolling over again. Earnings prints matter less than guidance direction and AI datapoints.

---

Phase C — Judge people & books

Management

CEO — Jochen Hanebeck. Infineon lifer: joined 1994, ran the Automotive division 2008–2016, COO 2016–2022, CEO since April 2022, contract renewed to 31 March 2032. An operator/engineer, not a financier — the archetype that fits an IDM.

1. Track record: presided over the FY2023 record, then navigated the FY2024–25 EV downturn "in line with long-term targets through the cycle," and is now executing the AI-datacenter pivot + reorg. As ex-head of Automotive he owns the segment that is both the anchor and the cyclical liability — credible on the core, and the one who architected the Marvell/Ethernet automotive-content expansion.
2. Tenure & skin in the game: long tenure, long runwayRunwayHow long the cash lasts at the current rate of spending. It shortens the moment spending rises, which is why a figure taken from a quiet quarter flatters. (2032). Insider ownership is modest (professional-manager, not founder) — our figures absent, exact stake n/a. CEO comp assessed as "appropriate" by third-party analysis.
3. Capital allocation: dividend payout ratio ~45%, "well supported by cash flows"; buybacks are small and employee-program-only (not a return lever — Infineon explicitly runs limited buybacks purely to service employee equity plans). The real capital-allocation story is M&A + capacity: Cypress (€8.3bn, 2020), GaN Systems, Marvell Automotive-Ethernet ($2.5bn, 2025), and ~€2.5bn/yr of frontend capex (Kulim SiC, Dresden 300mm). RoCE guided to only mid-single-digit % in the current environment — a genuine flag: heavy capex + big goodwill from acquisitions is depressing returns on capital during the build phase.
4. Red flags: none of the promotional/related-party kind. The legitimate concerns are capital-intensity (RoCE dilutionDilutionIssuing new shares, so each existing share owns a smaller slice of the same company. while the fabs ramp) and acquisition digestion (Cypress goodwill, now Marvell). Governance is standard German (supervisory board, AGM-approved).
5. Archetype: professional manager / insider-operator, appropriate for a scaled IDM in a capacity-and-M&A-heavy phase. Execution-competent; the question is capital-allocation returns, not integrity.

Forensic Red Flags

Web-only. No EDGAR shelf; figures /. IFRS reporter.

Income statement: the key artifact is the gap between adjusted and reported gross margin — Q2 FY2025 was 40.9% adjusted vs 38.7% reported; FY2026 guidance is quoted on an adjusted basis (low-to-mid-40s). The adjustments are largely purchase-price-allocation amortization from Cypress (and now Marvell) and restructuring. Not aggressive, but the ~2pt adjusted-vs-reported wedge is real and recurring — value the business on reported margins, discount the "adjusted" headline.

Balance sheet:

• Goodwill & intangibles from Cypress (€8.3bn deal, 2020) + Marvell ($2.5bn, 2025) are large relative to equity. Cypress PPA generated substantial goodwill and amortizable intangibles (exact 2026 goodwill carrying value n/a). Impairment risk exists if the automotive/CSS end-markets underperform the acquisition case — this is the single biggest accounting-risk item. No impairment has been flagged to date.
• Net debt €4.7bn, net leverage ~1.4x, gross at the 2.0x ceiling — investment-grade, but the gross ceiling limits further debt-funded M&A (Infineon has said it may fund ~30% of big deals with equity — dilution risk).
• Inventory/receivables: the FY2024–25 story was elevated inventory (own + channel) during the correction; the FY2026 recovery narrative is the drawdown of that inventory. Watch for any renewed inventory build outrunning revenue as a leading sign the recovery is stalling. Exact FY2026 DIO/DSO n/a.

Cash flow vs earnings: reported FCF (€900m) is roughly half adjusted FCF (€1.8bn) because of frontend-building capex. That is disclosed and legitimate (Infineon defines "adjusted FCF" ex-frontend-buildings) but means the headline "€1.8bn FCF / 12.3% of sales" flatters the true cash conversion during the capex-heavy phase. Use reported FCF for valuation.

SBC: standard for a large-cap semi; not a dominant non-GAAP flatterer the way it is for US growth-semis (Infineon's adjustments are PPA-amortization-led, not SBC-led). Lower concern here.

Regulatory findings (required sub-section).

• SEC (EDGAR LR/AAER): regulatory/regulatory-findings.md (generated 2026-07-06) states: "Infineon Technologies has no CIK — it is public and not required to file with the SEC. No EDGAR enforcement search is possible.". Zero SEC findings — because it is out of SEC jurisdiction, not because it was cleared.
• Non-SEC enforcement (web search): No material recent FTC/DOJ/EU-competition/consent-decree action against Infineon surfaced in this pass for the current period. (Historical note for completeness: Infineon was among DRAM/SRAM manufacturers fined in the 2000s antitrust price-fixing cases — a legacy matter, not current-management, not material to today's thesis.) Current-period material enforcement: none found.
• Item 3 (Legal Proceedings) equivalent: Infineon files an IFRS Annual Report (2025 AR), not a 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes.; its "Legal risks / litigation" disclosures are in the risk report / notes. I did not fetch the full 2025 AR litigation note in this pass (n/a at the line-item level) — flag for a future refresh with the AR PDF.
• Verdict: No material regulatory or legal findings for current management — verified via the (inapplicable) SEC EDGAR EFTS, web enforcement search, and press coverage as of 2026-07-06. The gap: German-filing litigation notes were not read line-by-line (fetch the 2025 AR to close).

---

Phase D — Project & stress-test

Forward Projection

Bottom-up from FY2025 actuals + FY2026 guidance. Every input labelled; output. Reported in EUR (fiscal years ending 30 Sep). No our model create per --watchlist rules.

Anchors: FY2025 revenue €14.662bn. FY2026 guided >€16bn, ~20% segment margin, low-to-mid-40s adj. gross margin. AI-datacenter €1.5bn FY26 → €2.5bn FY27. Share count ~1.31bn. FY2025 basic EPS from continuing ops was €0.18 in Q2 and roughly ~€0.90–1.00 for the FY.

Because a clean FY2025 full-year EPS is not sourced, EPS paths below are explicitly and should be treated as directional, not precise.

Base-case logic: FY2026 = the €16.2bn guide at ~20% segment margin. FY2027–28 = auto recovery (mid-single-digit) + AI-datacenter ramp (€1.5→€2.5→~€3.5bn, the standout growth engine) + industrial normalization, partially offset by SiC price erosion and heavy D&A/capex. Operating leverage lifts EPS faster than revenue as the FY2024–25 restructuring (1,400 jobs) + fab utilization recovers. Bull adds share-gain in datacenter power + faster auto/SiC + margin to low-20s. Bear = auto double-dips or Chinese SiC/EV competition caps datacenter margins and auto content — revenue flattens near €16bn and EPS stalls ~€1.10–1.30.

At €76.26, base FY2026e ~€1.25 EPS ⇒ ~61x current-year; on FY2027e ~€1.65 ⇒ ~46x; on FY2028e ~€2.00 ⇒ ~38x. Note this reconciles only loosely with the sourced 30.6x "forward" P/E — the gap means either (a) my EPS estimates are conservative vs. the sell-side, (b) the market "forward" P/E is on a further-out or adjusted-EPS basis, or (c) consensus adjusted EPS is materially above my reported-basis estimate. I flag this conflict explicitly rather than reconcile it silently — the sourced 30.6x is the tradeable number; my ~38–46x on reported EPS says the reported-earnings valuation is richer than the adjusted-EPS screen suggests. Do not trust a single EPS figure here without the FY2025 AR to anchor the base.

Bull vs Bear

Bull case. Infineon is the only scaled Western power-semi that spans Si + SiC + GaN, arriving exactly as AI-datacenter power becomes the fastest-growing demand pool in semis — racks going 120kW → 500kW → 1MW and the industry shifting to 800V HVDC, where Infineon is a named NVIDIA MGX partner with a €1.5bn→€2.5bn revenue ramp already in the guide. Simultaneously the automotive down-cycle is inflecting up (backlog recovering, inventory cleared), and Infineon has added auto content via Marvell Ethernet (software-defined-vehicle exposure). It is the #1 franchise in both power and auto semis — structurally advantaged, with a secular-growth graft (AI + electrification + renewables) on a recovering cyclical base. Capital allocation is disciplined (45% payout, real capex into share-gaining capacity, no financial-engineering buybacks). The bull owns a re-rating from "German EV-cycle laggard" to "AI-power infrastructure compounder."

Bear case (2–3 things that could permanently impair or de-rate):

1. SiC value-pool collapse. The €7bn/30%-share SiC ambition was underwritten in 2023's EV euphoria. Since then Chinese substrate players took ~40% share and drove wafer prices −30%, Wolfspeed went bankrupt (Sept 2025), and BYD makes its own SiC. If SiC becomes a low-margin commodity faster than Infineon can ride the cost curve, the single biggest capex bet (Kulim) earns a mediocre return — impairing RoCE, not just growth..
2. AI-datacenter power is now consensus and contestable. Every power-semi (STM, onsemi, Monolithic Power, Navitas, Vicor, plus in-house hyperscaler efforts) is chasing the same 800V rack. €2.5bn FY27 is real but it's ~13% of revenue — not enough to offset a soft auto cycle, and the margin/share may compress as competition floods in. The market is already paying for this leg (~27x EV/EBITDA).
3. Capital intensity + goodwill. ~€2.5bn/yr capex, net leverage at the ceiling, RoCE guided to mid-single-digits, and €10bn+ of goodwill/intangibles from Cypress + Marvell that could impair in a prolonged downturn. This is a capital-heavy business earning sub-par returns during the build — the return on the current investment cycle is unproven.

Pre-mortem (18 months out, thesis broke): Automotive recovery stalled (EV demand soft, China local-content substitution accelerating), SiC turned into a commodity price-war that hit Infineon's and everyone's margins, the AI-datacenter ramp came in on revenue but below expected margin as competition intensified — and the ~30x/27x multiple compressed toward NXP's ~21x on a "great franchise, mediocre returns, capital-heavy" re-rating. Stock back toward the mid-€50s.

Are multiples too high? On EV/EBITDA (27.4x), yes-ish for a business with mid-single-digit RoCE and cyclical auto exposure — you're paying a growth multiple for a capital-heavy cyclical with a real (but contested) AI graft. On fwd P/E it's mid-pack. Net: priced for the pivot to work.

Contrarian view (what the market is refusing to see): the crowd is fixated on the AI-datacenter €2.5bn headline — but that's now in the price and increasingly contested. The underappreciated, higher-conviction leg is the automotive/industrial margin-normalization: a #1 franchise coming off a forced restructuring (1,400 jobs) into a demand recovery has powerful operating leverage that lifts reported margins back toward the FY2023 peak — a boring, high-probability earnings-recovery story that doesn't need SiC or AI to be a home-run. The risk the market is under-weighting is the mirror image: Chinese vertical integration (BYD captive SiC, local auto-chip substitution) permanently shrinking Infineon's China auto/SiC TAM — a slow, structural erosion that no quarter will make obvious until the design-wins are already lost.

Devil's Advocate (short-seller)

Dismantling the bull case.

• What structurally breaks the model: Infineon's economics rest on selling differentiated power/auto silicon at a premium. Two forces attack that simultaneously — (a) Chinese commoditization (substrate, discretes, and increasingly full auto-chips, backed by local-content policy and captive OEM fabs like BYD), and (b) hyperscaler/OEM vertical integration in the very datacenter-power and EV-traction sockets the bull is excited about. Premium power silicon becoming a cost-down commodity is the structural break.
• Revenue concentration: ~half the company is Automotive, and a large share of that is Greater China — the exact market most exposed to local-content substitution and the EV price war. If China auto sockets shift domestic, a big chunk of the anchor segment erodes over model cycles (invisible until it isn't).
• Why the moat is weaker than bulls think: the SiC moat is already proving contestable — a 30% wafer-price crash and a bankrupt US pure-play (Wolfspeed) in <2 years is not what a durable moat looks like. Infineon isn't even fully substrate-integrated, so it rides someone else's cost curve on the key material.
• Most dangerous competitor bulls underestimate: not STM or onsemi — it's China Inc. as a bloc (SICC/TanKeBlue on substrate, BYD/local IDMs on devices), compressing price and taking share under policy cover, plus the hyperscalers designing power in-house.
• Worst capital-allocation risk: mid-single-digit RoCE while spending €2.5bn/yr capex + €10.8bn on Cypress+Marvell goodwill. If the auto/SiC case underperforms, that's impairment + a permanently lower return profile — value destruction dressed as "investing through the cycle."
• Assumptions that must hold for €76: (1) auto recovers and holds through FY2027–28; (2) AI-datacenter hits €2.5bn at a good margin; (3) SiC price erosion slows enough for Kulim to earn its keep; (4) no China-share cliff. Miss any two and the ~30x/27x multiple compresses.
• If growth disappoints 20–30%: FY2027 revenue ~€14–15bn instead of ~€18bn, EPS back to ~€1.10, RoCE stuck mid-single-digits — a de-rate toward NXP's ~21x fwd P/E ⇒ stock mid-€50s or lower (~25–30% downside).
• Single scenario that permanently impairs: structural loss of the China automotive/SiC value pool to domestic vertical integration — plausible on a 3–5yr horizon (medium probability), and the one that turns Infineon from a re-rating compounder back into a low-return cyclical.

Management Questions (ordered by information value)

1. RoCE is guided to only mid-single-digits during this capex cycle — at what revenue level and by which fiscal year does the current investment program (Kulim SiC, Dresden 300mm) cross your ~15% RoCE hurdle, and what happens to the plan if SiC prices keep falling?
2. Chinese substrate players took ~40% share and drove SiC wafer prices down ~30% since 2023 — what is the updated realistic 2030 SiC revenue and market-share target (vs. the original €7bn/30%), and what SiC gross margin do you now underwrite?
3. Roughly what share of Automotive revenue is Greater China, and what is your concrete plan for the risk that Chinese OEMs (and BYD's captive fabs) substitute domestic auto/SiC content over the next 3–5 years?
4. The AI-datacenter power target is €1.5bn→€2.5bn (FY26→FY27) — what gross/segment margin does that business carry, and how do you defend it as STM, onsemi, MPS, Navitas and hyperscaler in-house power all target the same 800V rack?
5. You reorganized into Automotive / Power Systems / Edge Systems on 1 July 2026 — will you report full P&L (revenue and segment result) for all three, and specifically break out AI-datacenter power within Power Systems?
6. Net leverage is at your ~2.0x gross ceiling — what is the appetite and financing mix (debt vs. equity) for the next large acquisition, and where are the gaps in the portfolio you'd fill?
7. Cypress (€8.3bn) and Marvell ($2.5bn) added ~€10bn+ of goodwill/intangibles — under what end-market scenario would you have to test these for impairment, and how close are current results to the acquisition cases?
8. Adjusted gross margin runs ~2pts above reported (PPA amortization) — when does PPA amortization roll off, and what is the sustainable reported through-cycle gross-margin target?
9. Reported FCF (~€900m) is roughly half adjusted FCF because of frontend buildings — when does frontend capex peak, and what's the normalized reported-FCF conversion you're steering to?
10. What is the design-win-to-revenue lag in Automotive today, and what specific leading indicators in the backlog tell you the recovery is durable rather than a restock?
11. On SiC substrate: how much do you now source from Chinese suppliers, will you vertically integrate crystal growth, and how do you hedge a Wolfspeed-style disruption at a key supplier?
12. GaN (CoolGaN / GaN Systems) — where does GaN out-compete SiC in datacenter and auto, and what's your GaN revenue trajectory?
13. Beyond the completed 1,400-role restructuring, what further structural cost actions are on the table if the auto recovery underwhelms, and what's the incremental/decremental margin you'd defend?
14. How exposed is FY2026–27 guidance to EUR/USD (guided at 1.17) and to tariff/trade disruption, and how much of that can you offset operationally?
15. What is the single greatest risk to the business you see on the 5-year horizon — and what are you doing about it that the market can't yet see in the numbers?

---