Soitec

Phase A — Understand the business

Company Overview

Soitec makes engineered semiconductor substrates — not chips. It sells the specialty wafer that a foundry (GlobalFoundries, STMicroelectronics, TSMC, Samsung) builds transistors on top of. Its core process, Smart Cut™, uses hydrogen ion implantation plus wafer bonding to transfer an atomically thin, defect-controlled layer (silicon, SiC, piezoelectric, III-V) onto an insulating base — a "layer-transfer" technique it has industrialised for 20+ years and protected with a large patent estate.

Three reporting divisions (FY2026 revenue €592m total):

Business model & payment terms. Revenue is wafer sales into a concentrated set of foundries and IDMs under multi-year, take-or-pay-style supply agreements carrying capacity commitments, minimum-purchase volumes and supply-share percentages. That contract structure is a double-edged sword: it locks in demand visibility but concentrates the customer book. Key customers/partners: GlobalFoundries (largest, RF-SOI + FD-SOI + Photonics-SOI), STMicroelectronics (SiC cooperation + FD-SOI), Skyworks (multi-year POI/Sky5 agreement), plus the CPO ecosystem (NVIDIA, Broadcom, Marvell, Intel) pulling Photonics-SOI.

Supply Chain

Upstream inputs → Soitec → foundry → chipmaker → device. Named stakeholders along the chain:

• Upstream (Soitec's own suppliers): bulk 300mm silicon "handle" and "donor" wafers bought from the very wafer majors it sits above — Shin-Etsu, SUMCO, GlobalWafers/Siltronic, SK Siltron — plus SiC polySiC substrates for SmartSiC, and specialty gases/implant consumables. Soitec buys commodity wafers, adds Smart Cut IP, and resells at a large value-add markup.
• Soitec fabs: Bernin (France, HQ), Pasir Ris (Singapore), Hsinchu (Taiwan), plus a licensed 200mm partner Simgui (Shanghai).
• Direct customers (foundries/IDMs): GlobalFoundries (the anchor), STMicroelectronics, TSMC, Samsung, plus SAW-filter makers (Skyworks, Qorvo) for POI.
• End OEMs / demand pull: smartphone RF front-ends (Apple, Samsung, Chinese handset OEMs) for RF-SOI/POI; AI datacenter / hyperscalers (via NVIDIA/Broadcom CPO) for Photonics-SOI; EV makers for SmartSiC.

Chokepoint — the load-bearing fact of this whole dive. GlobalFoundries disclosed in its own 20-F that it relied on Soitec for ~71% of its SOI-wafer spend, that replacing Soitec "would take an extended period… on commercially acceptable terms," and that near-term alternate supply is not assured. That is a genuine single-source chokepoint running in Soitec's favour — GF cannot easily route around it. But read the reverse: GF is also Soitec's single largest customer, so the same relationship is Soitec's top customer-concentration risk. The chokepoint and the concentration are the same wire, seen from two ends.

Competitive Advantages (moats)

The moat is real and unusually clean for a materials company:

1. Smart Cut IP + two-decade process know-how. SOI-by-layer-transfer is a yield-and-defectivity game refined over 20 years. The single strongest moat tell in the whole file: Shin-Etsu Handotai — the world's largest silicon-wafer maker — licenses Smart Cut from Soitec rather than engineer around it. When the biggest player in your input market would rather pay you a royalty than build a substitute, the moat is structural, not marketing.
2. Standard-setting / switching costs. RF-SOI on Soitec substrates is an entrenched industry standard; a foundry that has qualified a process on Soitec wafers faces long, expensive requalification to switch.
3. Photonics-SOI scale. For silicon-photonics/co-packaged-optics substrates, Soitec is described as "arguably the best-positioned photonics substrate name in the world," with 20 years of experience and the only substrate line at industrial scale for the AI-optics ramp.
4. Bargaining power. Strong over customers in RF-SOI/photonics (few alternatives, qualified-in). Weaker over its own commodity-wafer suppliers (Shin-Etsu et al. are larger). Market share: ~28% of the specialised SOI-wafer market (2024) and effectively the dominant Smart-Cut-substrate supplier globally.

Where the moat is thin: SmartSiC — SiC-by-layer-transfer has not proven a moat; Chinese SiC-substrate price crashes forced a €41m impairment in H1 FY26. The Smart Cut edge does not automatically transfer to every material.

Segments

our figures is empty (foreign filer), so all figures are unless noted:

FY25 comparison base (approx., scope/FX-adjusted): Mobile ~€524-546m, Edge & Cloud AI ~€198-216m, Auto & Industrial ~€123-129m. The mix shift is the story: two years ago Mobile was ~60%+ of a ~€1bn business; today it is 52% of a €592m business, and the market is repricing Soitec off the ~15-17% of revenue that is AI-photonics.

Phase B — Measure performance

Earnings Result (FY2026 full-year, ended 31 Mar 2026; released 2026-05-27)

All:

• Revenue €592m, −34% reported (−30% cc & scope). Q4'26 €202m, −38% YoY but +25% sequential — the sequential turn is the tell that the trough is in.
• Gross profit €96m (16.3% margin) vs €286m (32.1%) FY25 — a ~1,580bp gross-margin collapse from under-loaded fabs.
• EBITDA €151m (25.4% margin) vs €298m (33.5%) FY25.
• Operating loss −€131m vs +€119m FY25 — swung to a loss. Includes the €41m SmartSiC impairment (H1) plus other write-downs.
• Net loss (group share) −€220m (EPS −€6.22) vs +€92m / +€2.56 FY25. "Current" (underlying) net result only −€14m — i.e. ~€206m of the loss is non-cash impairment/one-offs, not operating cash burnCash burnHow much more cash goes out than comes in, per period. The clock on a company with no profits..
• Balance sheet is the bright spot: FCF +€63m (vs a consensus ~€6m — a large beat, and the proximate trigger for the +14% post-print pop). Operating cash flow €202m (flat YoY). CapexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. €135m (−41%). Net debt just €56m, net debt/EBITDA 0.4x, cash €562m. Gross debt €620m after repaying €325m OCEANE 2025 convertible (1 Oct 2025) and issuing a €222m Schuldschein; avg maturity 3.7yr, avg cost 3.07%, €120m undrawn lines.

Read: a textbook cyclical trough — earnings cratered and swung to a headline loss, but the company generated cash through the bottom (positive FCF, deleveraging, no dilutionDilutionIssuing new shares, so each existing share owns a smaller slice of the same company.) and printed a sequential Q4 upturn. The quality of the loss (impairment-driven, not cash) matters.

Earnings Calls (sentiment trend)

No transcripts on the shelf; sentiment read from press releases/coverage. The tonal arc over the last ~4 reporting events:

• FY25 (May 2025): cautious — annual guidance withdrawn citing "ongoing volatility"; Q1'26 flagged down ~20% on Imager phase-out.
• H1 FY26 (Nov 2025): worst tone — net loss €66.7m on €230.7m H1 sales; SmartSiC impairment; overhang of the CEO succession (announced Oct 2025).
• FY26 (May 2026): pivot to cautiously optimistic — management leads with Photonics-SOI as "a powerful growth driver addressing surging AI data-center demand" and "a key pillar of the Group's expansion," frames the trough as behind, and guides Q1'27 +15% YoY.
• Recurring new phrases: "photonics," "AI data center," "co-packaged optics," "asset fungibility." Phrases retired: the confident RF-SOI/5G supercycle language of 2021-22. The narrative has been deliberately rotated from smartphone-RF cyclical to AI-photonics secular growth.

Comps

Soitec has no true pure-play peer — it is a specialty layer-transfer monopoly, not a bulk-wafer maker. The nearest listed comparables are the commodity 300mm-wafer majors (different, more cyclical, lower-value business). Multiples below are marked n/a where I could not pull a clean, dated figure; I will not fabricate a multiple.

Market-share/margin data. Valuation anchor (clean, dated): EV/Sales 6.37x, P/B 2.80x, ROE −15.2%, ROIC −1.43%. The point: the bulk-wafer peers trade as cyclicals; Soitec commands a large premium (~6.4x sales on trough revenue) that is entirely a bet on the Smart Cut moat + photonics S-curve, not on the current P&L. On any normalised-earnings basis it is expensive; on the current loss it has no P/E.

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

:

• 2021-22: RF-SOI/5G supercycle → rally to ~€230 (2022 peak), ~€8bn market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth..
• 2022→2025: brutal multi-year smartphone/RF inventory correction + auto/China weakness → long grind down.
• Spring 2025: profit warning / guidance withdrawal → sharp drop.
• Sep-Oct 2025: CEO Barnabé resignation announced → management-credibility overhang.
• Dec 2025: 9-year low ~€20-22 (52-wk low €22.62).
• Jan 2026: new CEO Rémont appointed; FAMES FD-SOI pilot line (€830m EU funding) inaugurated → sentiment inflects.
• H1 2026: photonics re-rating — stock ~5x to a 52-wk high of €200.50.
• 27 May 2026: FY26 print, FCF €63m vs €6m consensus + Photonics-SOI >$100m → +14% pop.
• Since: profit-taking pullback to ~€98-104 (as of 2026-07-09/10).

Pattern: historically the stock traded on the smartphone/RF cycle; it now trades as a thematic AI-photonics momentum name with extreme volatility (a €22→€200→€98 round-trip in ~7 months). The market reacts most to (1) narrative/theme shifts, (2) cash-flow surprises, and (3) management credibility — increasingly less to the RF cycle it used to live on.

Phase C — Judge people & books

Management

• New CEO: Laurent Rémont (54), effective 1 April 2026. Ex-Infineon SVP heading the RF & Sensors business; former CTO of Kontron AG; 15+ years at STMicroelectronics. Archetype: professional operator with deep RF + substrate-adjacent domain credibility — arguably a sharper technical/commercial fit than his predecessor for both the RF-SOI base and the photonics pivot. But he is ~3 months into the seat — an unproven variable at exactly the inflection the stock is pricing.
• Predecessor: Pierre Barnabé (CEO from 2022, ex-Atos) resigned 30 Sep 2025 "for personal reasons," term ended 31 Mar 2026. A CEO exit during a profit warning is a governance yellow flag; Soitec published a formal "departure conditions" document (21 Nov 2025) — worth scrutinising for severance size (potential red flag; exact figure not pulled).
• Capital allocation: disciplined through the downturn — capex cut 41% to €135m, positive FCF held through the trough, OCEANE convert repaid in cash (no dilution). That is a good trough-management scorecard. The blemish is SmartSiC: a strategic SiC push (new Bernin plant, ST cooperation) that required a €41m write-down when Chinese pricing collapsed — capital deployed into a leg where the moat didn't hold.
• Skin in the game / ownership: dominated by strategic holders, not management — see Lens 10 / the ownership note. Insider founder ownership is not the story here; state and strategic control is.

Forensic Red Flags

our figures empty; figures.

• Impairments / earnings quality: the €220m net loss is ~94% non-current (underlying net result only −€14m). Positive: the loss is largely non-cash. Watch item: SmartSiC (€41m) and other write-downs mean the balance sheet still carries capitalised SiC/other assets that could impair further if utilisation stays low.
• Cash flow vs earnings — healthy divergence: operating cash flow €202m against an operating loss of −€131m; FCF +€63m against a −€220m net loss. Cash generation far exceeds accounting earnings (the opposite of the classic red flag). Heavy D&A (~€130-150m/yr ) is the reconciling item — Soitec is capital-intensive, so EBITDA materially overstates earnings power and net income swings negative at trough loading. That is the single most important accounting nuance for valuation.
• Leverage/liquidity: clean — net debt €56m, 0.4x EBITDA, €562m cash, €120m undrawn, no near convertible cliff after the Oct-2025 OCEANE repayment.
• Segment reporting: the FY25→FY26 segment restatement (Imager reclassification, cc-vs-reported) makes clean YoY comparison slightly harder — not manipulation, but a diligence friction; use the company's stated YoY %, not derived bases.
• Revenue recognition / receivables/inventory: no disclosed anomaly found; in a destocking trough, watch inventory and take-or-pay minimum-purchase true-ups at concentrated customers.

Regulatory findings (required). Per regulatory/regulatory-findings.md (2026-07-10): Soitec has no SEC CIK — no EDGAR Litigation Releases or AAERs are searchable, and none exist. Non-SEC web search ("Soitec" (FTC OR DOJ OR EU OR consent decree OR settlement OR fine OR penalty) enforcement) surfaced no material enforcement action, fine, or litigation as of 2026-07-10. The company's own French/AMF disclosures (URD) would carry any material legal proceedings; none flagged in coverage. Conclusion: no material regulatory or legal findings — verified via (a) SEC EDGAR EFTS (n/a — no CIK), (b) web search, and (c) coverage of AMF/URD disclosures, as of 2026-07-10. The live geopolitical exposure is a shareholder/strategic matter (French-state + Chinese NSIG holdings — see below), not an enforcement one.

Phase D — Project & stress-test

Forward Projection (EPS, FY27-FY29)

No clean sell-side consensus EPS for FY27/FY28 was sourceable (n/a); the projection below is bottom-up from FY26 actuals + guidance (Q1'27 +15% cc&scope; capex ~€100m; ~95% of FY27 hedged at 1.19 €/$; profitability capped by low fab loading, FX, lower funding). Structural inputs: RF destock ending, Photonics-SOI compounding off >$100m, FD-SOI design-win ramp, Auto/SiC soft; heavy D&A ~€140m/yr; ~35.7m shares, ~25% tax, ~€15m net interest.

Worked example (base FY27): €592m ×1.15 ≈ €680m; EBITDA €680m ×28% ≈ €190m; less D&A ~€140m = EBIT €50m; less €15m interest = €35m; ×0.75 tax = **€26m net → EPS ~€0.73**. Implication: even the base path only returns Soitec to ~€2.80 EPS by FY29 — a ~35x P/E three years out at €98.62. The multiple only looks defensible on the bull case (~16-22x FY29). Base call to track (candidate forecast, not logged per --watchlist rules): SOI.PA returns to positive current-net-result / ~breakeven-to-slightly-positive statutory EPS in FY27, base EPS ≈ +€0.70, resolves 2027-03-31.

Bull vs Bear

Bull case. A Smart Cut monopoly with a customer (GF) that has disclosed it can't replace 71% of its SOI supply — a rare, durable chokepoint moat. Layered on top is the single best secular driver in tech: AI-datacenter interconnect. Co-packaged optics (NVIDIA Spectrum-X/Quantum-X, Broadcom, Marvell) is shifting from pluggables to CPO for ~30% energy savings, with AI datacenter capex compounding >20% through 2030 — and Soitec is the only industrial-scale Photonics-SOI substrate, already >$100m and inflecting early. Optionality stack: FD-SOI (FAMES 7nm-and-below pilot, €830m EU-funded), 28Si FD-SOI for quantum. Balance sheet is fortress (0.4x net leverage, €562m cash, positive trough FCF). If RF-SOI merely normalises while photonics scales, revenue and margins re-rate together.

Bear case (permanent-impairment risks). (1) Valuation — ~6.4x trough sales / mid-20s-to-high-30s EV/EBITDA / no P/E, after a 5x melt-up, for a business 15-17% exposed to the photonics theme it's priced on. (2) Mobile is half the company and structurally soft — smartphone units are ex-growth; RF-SOI content gains may not restore the old €500m+ base. (3) SmartSiC proved the moat doesn't transfer — Chinese commoditisation impaired the SiC bet; a warning that "Smart Cut everywhere" is not guaranteed. Pre-mortem (18 months out, thesis broke): photonics revenue is real but smaller and slower to book than the multiple implied ($150-200m, not the €-hundreds the narrative extrapolated); mobile double-dips; the new CEO's first full year disappoints on margins under low loading + FX; the €22→€200 momentum crowd that "does not understand what the company does" reverses, and the stock round-trips toward €50. Contrarian view the market is refusing to see: the market is treating Soitec as an AI-photonics pure play when it is still an RF-SOI smartphone cyclical with a fast-growing photonics option — the re-rating has run ahead of the mix shift.

Devil's Advocate (short-seller)

Where it structurally breaks: customer concentration is the mirror of the moat — GF at ~71%-of-its-SOI-spend means a GF share-loss, in-sourcing move, or a foundry-share shift away from FD-SOI/RF-SOI hits Soitec disproportionately; the take-or-pay contracts cut both ways in a prolonged glut. Most dangerous competitor bulls underestimate: not another substrate maker but the customers' own roadmap — if hyperscaler CPO designs standardise on a different substrate approach, or if GF/foundries qualify a second SOI source (they are explicitly "developing relationships with alternate suppliers"), the chokepoint erodes. Worst capital-allocation move: the SiC build-out that impaired. Assumptions that must hold for €98: photonics scales to many hundreds of millions AND mobile stops shrinking AND margins recover to ~32%+ despite low loading — a lot to underwrite at ~6.4x trough sales. If growth disappoints 20-30%: on a normalised-earnings framework the stock has 40-60% downside toward the €50s (still a premium to bulk-wafer peers). Single permanent-impairment scenario: a credible second-source qualifies at GlobalFoundries and CPO adoption slips a generation — plausible enough (low-teens %) to demand a margin of safety the current price does not offer.

Management Questions (ordered by information value)

1. Photonics-SOI is >$100m — what is the booked backlog and design-win pipeline for FY27-FY29, and at what revenue does it become >25% of Group sales?
2. On co-packaged optics, which CPO programs (NVIDIA, Broadcom, Marvell) are you qualified into, and is your substrate spec'd single-source or dual-source in each?
3. GlobalFoundries is your anchor and depends on you for ~71% of its SOI wafers — what % of your revenue is GF, and how do the current take-or-pay minimums/penalties read through the downturn?
4. Is GF (or any customer) actively qualifying a second SOI source, and on what timeline?
5. Mobile is 52% of revenue and down 41% — what is the normalised RF-SOI/POI revenue base once destocking ends, and can it re-approach the prior €500m+?
6. After the €41m SmartSiC impairment, what is the go-forward SiC strategy — invest, hold, or exit — and what carrying value remains at risk?
7. FY27 margins are guided down on "low fab loading" — at what utilisation/revenue level does gross margin return to 30%+?
8. FAMES (7nm FD-SOI, €830m EU-funded) — what is the customer commitment and revenue timeline, and how much is grant-funded vs Soitec capex?
9. Capital allocation at 0.4x net leverage with €562m cash — buyback, M&A, or reinvest, and what hurdle rate?
10. FX: ~95% hedged FY27 at 1.19 — what is the structural €/$ sensitivity beyond the hedge horizon?
11. As new CEO, what will you do differently from the prior regime on portfolio focus and photonics scaling?
12. With Bpifrance/CEA (French state) and NSIG (Chinese) as anchor holders, how do you manage strategic/geopolitical constraints on customers and export control?
13. What is the replacement cost and lead time for a competitor to industrialise a Smart Cut-equivalent at your yields?
14. Imager-SOI is phasing out — any other legacy lines you plan to rationalise, and the revenue headwind?
15. What single external variable most threatens the FY27 recovery, and what is your contingency?