AXT, Inc.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
A narrow materials moat: VGF crystal-growth process IP, a vertically integrated China raw-material chain, and multi-year optical-customer qualifications — real switching costs inside a three-player InP oligopoly, not a physical monopoly.
AXT's durability is narrow and physical — it is one of three merchants that can ship volume indium phosphide wafers into 800G/1.6T lasers, not a software or network franchise:
- InP Oligopoly & Qualification Cycles: High-quality InP substrates are made at scale by AXT, Sumitomo Electric and JX Advanced Metals, with Freiberger a smaller Western alternative. Epitaxy houses and laser OEMs (Coherent, Lumentum, and the China EML chain) typically qualify at least two substrate vendors over multi-quarter cycles; once a 3- or 4-inch flow is in a transceiver BOM, swapping is slow. That is stickiness. It is not exclusivity — the 10-K is explicit that customers dual-source — and Sumitomo remains the quality/share reference.
- VGF Process & Vertical Raw Materials: AXT grows crystals with its own VGF furnaces in China (Beijing Tongmei) and holds stakes in more than ten raw-material JVs, including Jingmei's high-purity indium refining. Q2 raw-material JV revenue hit a record $10M. The integration is a real cost-and-supply edge versus a pure substrate polisher. It is not unreplicable process IP: Freiberger also uses VGF, and at least two GaAs competitors already ship VGF-like material.
- China Manufacturing & the Permit Throttle: All wafer production sits in China. That is why AXT can add InP capacity faster than the Japanese incumbents — and why non-China shipments need Ministry of Commerce export permits. InP was $3.6M in Q2 2025 when permits were scarce, then $13.6M in Q1 and $30.7M in Q2 once they started clearing. China laser demand (no permit required) was more than half of Q2 revenue. The same geography that funds the ramp can shut the export door.
Ten Moats Verdict
AXT is a net beneficiary of AI — 800G/1.6T and co-packaged optics need InP lasers, and the merchant substrate book is only three names deep. The AI-resilient piece is transaction embedding inside those qualifications and LTSAs, plus the China operating-license bar. Nothing here is a data, network or system-of-record franchise, and the China export-permit regime is the same fact as the capacity story: it is how the 2025 InP collapse happened and how a 2027 destock would happen. Durability is a qualified-vendor seat in a cyclical materials oligopoly, not a compounder moat.
65.0 resilient · 57.1 vulnerable · 80/20 = 63.4 · = 63
AXT sells polished compound-semiconductor wafers, not a user-facing interface or workflow.
VGF furnace designs, doping recipes and 4-inch/6-inch InP process know-how are real and accumulated over decades, but Freiberger also grows VGF InP and Sumitomo's LEC/VGF book is the quality reference — not vendor-owned logic competitors cannot copy.
The business does not derive a moat from access to or aggregation of a public data source.
Crystal-growth engineers are specialized, but the pool is shared with Sumitomo, JX, Freiberger and Chinese substrate shops, so talent is not a standalone franchise.
AXT sells substrates and some JV raw materials, not a multi-product suite that makes any one wafer stickier.
Yield and process data are operational. There is no compounding, non-replicable dataset that a competitor cannot recreate by growing crystals.
China operating licenses, InP export-permit administration, and multi-quarter customer qualifications raise the bar for a new merchant. The same MOFCOM permit regime is also the throttle on non-China shipments — a barrier, not a Visa-style lock-in.
Adding another laser OEM does not make AXT's wafers more valuable to the others; there is no Metcalfe dynamic.
Once qualified, AXT sits in a 3–5 year optical platform cycle. Q2 LTSAs with Coherent, Lumentum and a second customer, plus $47.7M of wafer prepayments, are that embedment in cash. Customers still dual-source, and the 10-K says a vendor that misses lead times loses share.
AXT is a materials supplier. Downstream epi and laser houses keep the device record.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
A narrow materials moat: VGF crystal-growth process IP, a vertically integrated China raw-material chain, and multi-year optical-customer qualifications — real switching costs inside a three-player InP oligopoly, not a physical monopoly.
Growth Score
Q2 2026 printed $47.6M (+164% YoY, +77% QoQ) on record $30.7M of indium phosphide for AI data-center optics, with non-GAAP GM 45% and a return to profit ($0.19). Q3 is guided to $66M of already-permitted or unrestricted revenue and $0.30–$0.32 non-GAAP EPS. Management now targets ~$60M of quarterly InP capacity by year-end 2026 and ~$130M by the end of 2027, with backlog above $100M and LTSAs at Coherent, Lumentum and a second optical customer. FY2025 was $88.3M and a $21.3M loss — this is a step-function off a trough, not a decade of 30% compounding, and China export permits remain the throttle on every non-China InP dollar.
Valuation Score
At the Sep 22 tape of ~$75.70 (~$5.0B on ~66.5M diluted shares) AXTI is ~23× an implied FY2026 ~$220M revenue book and ~95× the ~$0.80 FY2026 EPS consensus — a materials-company multiple that already prices the InP ramp. The stock is 52% of the way from the $55 base to the $95 bull after a 52-week range of $4.03–$140.83. Street mean target ~$77 sits on the tape, not below it.
The InP Qualification Moat
AXT's durability is narrow and physical — it is one of three merchants that can ship volume indium phosphide wafers into 800G/1.6T lasers, not a software or network franchise:
- InP Oligopoly & Qualification Cycles: High-quality InP substrates are made at scale by AXT, Sumitomo Electric and JX Advanced Metals, with Freiberger a smaller Western alternative. Epitaxy houses and laser OEMs (Coherent, Lumentum, and the China EML chain) typically qualify at least two substrate vendors over multi-quarter cycles; once a 3- or 4-inch flow is in a transceiver BOM, swapping is slow. That is stickiness. It is not exclusivity — the 10-K is explicit that customers dual-source — and Sumitomo remains the quality/share reference.
- VGF Process & Vertical Raw Materials: AXT grows crystals with its own VGF furnaces in China (Beijing Tongmei) and holds stakes in more than ten raw-material JVs, including Jingmei's high-purity indium refining. Q2 raw-material JV revenue hit a record $10M. The integration is a real cost-and-supply edge versus a pure substrate polisher. It is not unreplicable process IP: Freiberger also uses VGF, and at least two GaAs competitors already ship VGF-like material.
- China Manufacturing & the Permit Throttle: All wafer production sits in China. That is why AXT can add InP capacity faster than the Japanese incumbents — and why non-China shipments need Ministry of Commerce export permits. InP was $3.6M in Q2 2025 when permits were scarce, then $13.6M in Q1 and $30.7M in Q2 once they started clearing. China laser demand (no permit required) was more than half of Q2 revenue. The same geography that funds the ramp can shut the export door.
Ten Moats Verdict
AXT is a net beneficiary of AI — 800G/1.6T and co-packaged optics need InP lasers, and the merchant substrate book is only three names deep. The AI-resilient piece is transaction embedding inside those qualifications and LTSAs, plus the China operating-license bar. Nothing here is a data, network or system-of-record franchise, and the China export-permit regime is the same fact as the capacity story: it is how the 2025 InP collapse happened and how a 2027 destock would happen. Durability is a qualified-vendor seat in a cyclical materials oligopoly, not a compounder moat.
65.0 resilient · 57.1 vulnerable · 80/20 = 63.4 · = 63
AXT sells polished compound-semiconductor wafers, not a user-facing interface or workflow.
VGF furnace designs, doping recipes and 4-inch/6-inch InP process know-how are real and accumulated over decades, but Freiberger also grows VGF InP and Sumitomo's LEC/VGF book is the quality reference — not vendor-owned logic competitors cannot copy.
The business does not derive a moat from access to or aggregation of a public data source.
Crystal-growth engineers are specialized, but the pool is shared with Sumitomo, JX, Freiberger and Chinese substrate shops, so talent is not a standalone franchise.
AXT sells substrates and some JV raw materials, not a multi-product suite that makes any one wafer stickier.
Yield and process data are operational. There is no compounding, non-replicable dataset that a competitor cannot recreate by growing crystals.
China operating licenses, InP export-permit administration, and multi-quarter customer qualifications raise the bar for a new merchant. The same MOFCOM permit regime is also the throttle on non-China shipments — a barrier, not a Visa-style lock-in.
Adding another laser OEM does not make AXT's wafers more valuable to the others; there is no Metcalfe dynamic.
Once qualified, AXT sits in a 3–5 year optical platform cycle. Q2 LTSAs with Coherent, Lumentum and a second customer, plus $47.7M of wafer prepayments, are that embedment in cash. Customers still dual-source, and the 10-K says a vendor that misses lead times loses share.
AXT is a materials supplier. Downstream epi and laser houses keep the device record.
Growth Analysis
Growth Drivers
Key Risk
Every non-China InP wafer still needs a Ministry of Commerce export permit. Q3's $66M guide is only the already-cleared or unrestricted slice; Fischer said timing and success on any given order cannot be predicted. Falsifiable test: if Q3 prints below $66M, or Q4 InP fails to move toward the ~$60M quarterly capacity claim, or a subsequent permit freeze cuts non-China InP the way Q2 2025's $3.6M did, the step-function is a one-cycle spike and the multiple compresses.
Score Derivation
86.7 base + 2.7 trajectory + 4 margin − 10 risk = 83
Base 87 (20–30% CAGR, midpoint 25%) + 3 trajectory (InP and raw-material JVs accelerating; GaAs stable) + 4 expanding margins − 10 high (China export permits, capacity fill, InP pricing) = 83
Price Scenarios (12–24 Months)
Valuation Analysis
Base fair value $55 assumes FY2027 revenue lands near $350–400M as InP capacity ramps, non-GAAP GM holds the mid-40s, and the market pays ~10× that book — a premium to a historical OSAT/substrate multiple, not a CRDO-style design-win multiple. Live tape ~$75.70 (Sep 22); ~$5.0B equity value on ~66.5M diluted shares after the April $632M secondary. Not a hire screen — coverage valuation only. $55.
Valuation Multiples
| Trailing P/E (GAAP) | n.m. |
| Forward P/E (FY2026) | ~95× |
| PEG Ratio | ~3.8× |
| Price / Sales (FY2026) | ~23× |
| Net cash | ~$0.7B |
The multiple is a claim that InP stays scarce through 2027 and that AXT fills toward the $60M then $130M quarterly capacity marks. At ~23× this year's sales and ~95× this year's earnings the tape has already paid for the inflection. PEG ~3.8× is expensive for a cyclical China-manufactured substrate book; AMKR is ~2.9× FY2026 sales. A single permit freeze or an InP destock re-rates this toward a high-single-digit sales multiple quickly.
Approximate figures as of September 22, 2026.
Where We Are vs Targets
Loading live price…
Export permits stall or InP pricing normalizes as Sumitomo and JX add wafers, the $66M Q3 print is a local peak, and the multiple compresses toward ~8× a stalled ~$250M book.
- Q3 prints below the $66M permitted guide, or Q4 InP fails to approach the ~$60M quarterly capacity claim
- A MOFCOM permit freeze cuts non-China InP the way Q2 2025's $3.6M quarter did, and China laser demand is not enough to fill the new furnaces
- InP wafer prices roll over as Japanese and Chinese incremental capacity arrives, and gross margin falls back out of the 40s
Q3 converts near $66M, FY2026 lands ~$210–230M, FY2027 reaches ~$350–400M as InP capacity ramps, mid-40s gross margin holds, and the market pays ~10× that still-cyclical book.
- Q3 revenue at or above $66M and Q4 InP moves toward the exit-2026 ~$60M quarterly capacity mark
- Coherent, Lumentum and the second optical LTSA convert prepayments into shipped wafers through 2027 without a permit air-pocket
- Non-GAAP GM stays in the mid-40s as InP mix and 4-inch/6-inch migration offset raw-material and GaAs drag
AXT fills toward $130M of quarterly InP by the end of 2027, becomes the largest merchant InP producer, 45%+ margins hold, and the market pays 12–14× a $500M+ run-rate.
- Exit-2027 InP capacity of ~$130M/qtr is substantially filled, putting AXT ahead of Sumitomo on merchant volume
- 6-inch InP reaches volume with Coherent and others, lifting mix and holding GM at or above 45%
- Permits become regular enough that non-China hyperscaler optics (the call said AXT material is already in multiple US hyperscalers) matches the China laser run-rate