Semiconductors
PrivateThe best pure-play on compound-semi deposition — but the AI-optics repricing (€6.3bn cap, ~74x trailing P/E, +21% over avg PT) now prices the opto boom as permanent while a Q1 that printed a -38% EBIT margin shows how brutal the trough is; own the tool moat, fade the multiple.
Research
The verdict
The best pure-play on compound-semi deposition — but the AI-optics repricing (€6.3bn cap, ~74x trailing P/E, +21% over avg PT) now prices the opto boom as permanent while a Q1 that printed a -38% EBIT margin shows how brutal the trough is; own the tool moat, fade the multiple.
AIXTRON makes the deposition equipment that grows the crystalline layers of compound (non-silicon) semiconductors — principally MOCVD (metal-organic chemical vapor deposition) and CVD systems. It does not make chips; it makes the machines chipmakers use to epitaxially grow gallium-nitride (GaN), silicon-carbide (SiC), gallium-arsenide (GaAs), indium-phosphide (InP) and related III-V materials onto wafers. Those materials become power electronics, RF, LEDs/micro-LEDs, lasers and photonic components. In semicap terms AIXTRON is the "ASML of compound semiconductors" for the epitaxy step — a narrow, deep, tools-and-consumables franchise, not a broad process-equipment vendor.
Business model. Sell a high-ASP capital tool (a G10 platform system runs into the multiple €-millions), then earn a long recurring tail on spares, upgrades, and service as the installed base grows. Revenue is lumpy and shipment-timed — a single quarter can swing ±50% on when big multi-tool orders ship (Q1/2026 rev €59.4m vs Q4/2024 €226.9m). The order book, not the quarterly P&L, is the real health signal.
Three end-market segments (FY2025 equipment revenue mix):
Customers (Lens 2 detail): compound-semi foundries and IDMs — Lumentum, Nokia, SMART Photonics (named opto buyers 2026), Chinese SiC makers, LED/micro-LED and RF houses. Contract structure is book-and-ship capital orders — not take-or-pay, not subscription; concentration risk is by end-market cycle and by geography (China) rather than by a single named customer.
Map: upstream inputs → AIXTRON (integrator) → epi-wafer/chip maker → device OEM → end system.
Chokepoints / single-source dependencies: (1) AIXTRON is itself a near-duopoly chokepoint in high-end MOCVD — for AI-optics InP epitaxy at scale, the practical choice is AIXTRON (G10-AsP) or Veeco. (2) China gallium/germanium export controls and US-led equipment export controls are a two-way chokepoint: China restricts Ga/Ge feedstock (hurts customers), the West restricts advanced-tool exports to China (hurts AIXTRON's largest growth market). (3) Precursor gas handling and 300mm GaN process readiness are capability chokepoints AIXTRON is spending to own (Herzogenrath innovation center, first 300mm GaN).
No segments.csv on disk — all ``. FY2025 equipment-revenue mix and direction:
| Segment | FY2025 share | Trend | Cause |
|---|---|---|---|
| SiC + GaN power | 57% | Decelerating hard into 2026 | EV/industrial SiC glut, Western under-utilization; propped by China G10-SiC volume in H1/2025 |
| Optoelectronics | 23% (FY25) → ~70% of Q1/26 orders | Accelerating sharply | AI 800G/1.6T optics, InP lasers; G10-AsP wins at Lumentum/Nokia/SMART |
| LED incl. micro-LED | 15% | Weak/flat | micro-LED project delays; smallest leg |
The whole story in one line: the demand engine is rotating out of power (57% of the base) into opto (the growth) — but opto was only ~23% of FY2025, so the mix shift is early. Geography: China is the swing region for SiC (stabilizer in 2025, hostage to export controls in 2026); US/EU benefit from CHIPS-adjacent SiC/GaN investment. Geographic revenue split not disclosed at segment level in sourced material → n/a for exact geo-EBIT.
Q1/2026 (reported ~30 Apr 2026) — a deliberate revenue trough with a violent margin hit, paired with a booking surge:
The market reaction / what was priced in: shares fell ~9% on the weak Q1/FY guide back in early 2025; the Q1/2026 order beat + FY guidance raise is what re-rated the stock toward €60. The tape now trades the order book and the opto narrative, and looks through the printed loss — classic early-cycle semicap behavior.
FY2025 actuals:
Balance-sheet flags: pristine — equity ratio ~85%, D/E ~0.01, net cash, current ratio ~5.6. The FY2025 cash build (€65m→€225m) came from working-capital release as revenue fell — a good sign of discipline, but partly a cyclical inventory drawdown that reverses when shipments ramp (Q2/2026+). Watch inventory rebuild against the €359m backlog.
FY2026 guidance — RAISED (with Q1 prelims, Apr 2026):
No transcripts/ on disk — ``, from call highlights/IR speeches across FY2024→Q1/2026:
Phrases that entered the vocabulary: "AI-era photonics," "chip-to-chip / rack-to-rack / datacenter-to-datacenter," "G10-AsP benchmark," "300mm GaN," "multi-tool orders." Phrases they stopped leaning on: micro-LED (the 2021-2022 hope), and the confident SiC/EV-supercycle framing of 2023. The tone shift is from defending a cyclical trough to selling a secular opto growth story — genuine, but note management is now incentivized to keep the opto narrative loud.
Peer set: high-end deposition/epitaxy + broad semicap. Multiples are with source/date or `n/a`. AIXTRON's own multiples are off sourced price + FY2025 actuals.
| Company | Ticker | Mkt cap | EV/Sales | Fwd P/E | EV/EBITDA | Div yield | Note |
|---|---|---|---|---|---|---|---|
| AIXTRON | AIXA.DE | €6.31bn | ~11.0x `` | n/a (trailing P/E ~74x ``) | n/a | ~0.25% `` | Pure compound-semi epitaxy; deep trough earnings inflate P/E |
| Veeco | VECO | ~$4.60bn | ~5.4x P/S | 22.7x | 23.1x | 0% | Direct MOCVD rival; Axcelis merger pending |
| ASML | ASML | very large | n/a | ~42-51x | ~43x | small | EUV monopoly; the "premium semicap" anchor |
| Applied Materials | AMAT | large | n/a | ~26x | ~20x | small | Broad-line semicap benchmark |
| AMEC | 688012.SS | n/a | n/a | n/a | n/a | Shanghai STAR-listed | China rival; state-backed, share-taker |
| Lam Research | LRCX | large | n/a | n/a | n/a | small | Broad semicap; not sourced this pass |
Read: on EV/Sales ~11x AIXTRON trades at ~2x Veeco (5.4x) and near ASML-monopoly territory — extraordinary for a €556m-revenue, cyclically-troughed name. The ~74x trailing P/E is a trough-earnings artifact (net profit fell to €85m; Q1/2026 was a loss), so P/E overstates richness — but even normalizing to a mid-cycle ~€130-150m net income `` implies ~42-48x, still a full premium to AMAT (26x) and Veeco (23x). The market is paying an ASML-like multiple for the AI-optics call option. That is the entire bull-bear crux.
Pattern — what actually moves AIXA: (1) order intake / backlog far more than reported revenue (it's an early-cycle book-to-bill stock); (2) guidance direction (cuts crush, raises re-rate); (3) China export-control headlines (binary geopolitical risk); (4) the AI-optics narrative (now the dominant driver). The market reacts to forward booking signals and the opto story, and looks through trough margins.
n/a; no insider-transactions.csv). AIXTRON is widely held, no controlling shareholder — a professional-manager governance profile.No filings on disk; income-statement/balance-sheet review is `` off reported figures. AIXTRON reports under IFRS (German HGB/IFRS), audited — not SEC-registered.
n/a); AIXTRON's compensation SBC is modest by semicap standards `` and it reports clean EBIT (not a heavy adjusted-EPS storyteller).Regulatory findings (required). Per regulatory/regulatory-findings.md (fetched 2026-07-01): AIXTRON has no CIK — no SEC EDGAR search possible; total_sec_findings: 0. Non-SEC web check ("Aixtron" (FTC/DOJ/FDA/EU/consent decree/fine/penalty) enforcement): no material enforcement action found. The material regulatory exposure is export-control, not misconduct — AIXTRON operates under German/EU/US dual-use export-control regimes; recurring headline risk that China SiC/GaN/opto tool sales get restricted (US pressing allies). Historical note: AIXTRON's 2016 attempted acquisition by China's Fujian Grand Chip was blocked on US/German national-security grounds — the definitional precedent that AIXTRON tools are geopolitically sensitive. No litigation/AAER/fraud findings. Verified via absence of SEC CIK (no LR/AAER possible), web enforcement search, and public disclosure as of 2026-07-01; 10-K Item 3 n/a (foreign filer, no EDGAR).
Base off FY2025 actuals (rev €556.6m, net €85m, ~112.6m shares, EPS ≈€0.75) and management's raised FY2026 guide (rev ~€560m, EBIT margin 17-20%). Net margin ≈ EBIT margin × ~0.80 (tax) + small net-cash interest income ``. Share count roughly flat (buyback authorized but undeployed).
FY2026 (base): rev ~€560m; EBIT margin ~18.5% (mid) → EBIT ~€104m; net ~€88m → **EPS ≈ €0.78**. H2-weighted: Q1 was a loss, Q2 ~€110m, back half carries the year via the €359m backlog.
FY2027 (base): opto ramp continues + GaN utilization recovery (late-2026 into 2027) + early SiC recovery. Rev ~€680-720m ; margin recovers to ~21-22% on volume leverage → net ~€120-130m → EPS ≈ €1.05-1.15 ``.
FY2028 (base): rev ~€800m , mid-cycle margin ~23% → **EPS ≈ €1.35-1.45** . Kerrisdale's bull frames €1.06bn 2028 revenue vs consensus ~€693m — my base sits between the two, closer to consensus.
Valuation check: at €60.16, the base FY2028 EPS ~€1.40 implies a forward P/E ~43x on a two-years-out number — the price already discounts a successful multi-year opto ramp. forecast.ts create skipped (watchlist loop rule — no committed base case logged).
Bull case. AIXTRON is the arms dealer for two secular build-outs at once — AI-optics (InP lasers/detectors for 800G→1.6T, where G10-AsP is the volume benchmark and orders are landing at Lumentum/Nokia/SMART) and, eventually, the SiC/GaN power recovery (EV + 800V data-center power). It sits in a duopoly (~55-65% MOCVD share) with deep switching costs, a fortress balance sheet (85% equity, net cash), first-mover 300mm GaN, and a management team that called the opto pivot early and raised guidance into a trough. Order intake +30%, backlog €359m and building, book-to-bill well above 1. If opto is a durable multi-year S-curve (Kerrisdale: €1.06bn 2028 rev, 271-322% upside), today's €6.3bn cap is early.
Bear case (permanent-impairment risks). (1) The multiple, not the business, is the risk — ~11x EV/Sales and a ~74x trough P/E price the opto boom as permanent and linear; any order air-pocket de-rates the stock 30-50% regardless of the moat. (2) China is 2-sided and structurally deteriorating — SiC/LED share bleeding to state-backed AMEC under domestic-procurement rules, and Western export controls can wall AIXTRON out of its biggest growth region overnight (the Fujian-Grand-Chip block set the precedent). (3) Opto concentration + lumpiness — the growth is a handful of optical-module customers (Lumentum et al.); AI-optics capex is itself a derivative of hyperscaler capex, so AIXTRON is two cycles removed from end-demand and whipsaws hard. Pre-mortem (18 months out, thesis broke): opto orders were a 2025-2026 capacity pull-forward that digests through 2027; SiC recovery slipped again; China controls tightened; revenue lands flat ~€560m, EPS ~€0.80 — and a stock priced for €1.40+ EPS halves to the low-€30s (near analysts' €23.5 low PT). Contrarian view the market is refusing to see: the opto order surge is real but the Street is extrapolating a lumpy, customer-concentrated, hyperscaler-derivative order book as if it were ASML-grade recurring demand — it is not; AIXTRON is a great business trading like a monopoly it isn't.
Dismantling the bull:
Best-in-class EDA franchise temporarily wearing an Ansys-debt-and-amortization disguise — the GAAP "collapse" is accounting, not the business; the real risk is paying ~35x forward for a name whose Design-IP leg is structurally cracked and whose synergy math doesn't pay until FY2028.
A genuine deleveraging turnaround (9.0x→~1.6x net leverage) that has tripled on AI-datacenter optionality — but the stock now prices that optionality at ~62x forward earnings while Credo owns ~88% of the very AEC market Semtech is fighting to enter; the moat is real in TVS/LoRa, not yet proven in datacenter interconnect. WATCHING, not chasing.
A toll-booth on the AI-silicon boom with 86% gross margins and an $8B backlog — but priced for perfection at ~42x forward earnings while the agentic-AI upsell that justifies the multiple is not yet in the model and a fresh DOJ guilty plea caps the China optionality. Quality is not the question; the entry price is.