AXT Inc.

AXT Inc. manufactures compound semiconductor substrates — indium phosphide (InP), gallium arsenide (GaAs), and germanium (Ge) — which are the foundational wafers used to grow epitaxial layers for optical transceivers, LiDAR laser diodes, and quantum computing devices.

AXT Inc. manufactures compound semiconductor substrates — indium phosphide (InP), gallium arsenide (GaAs), and germanium (Ge) — the single-crystal wafers on which epitaxial layers are grown to fabricate the lasers (EML / DFB / CW) and photodetectors inside 800G/1.6T optical transceivers, silicon-photonics external laser sources, and LiDAR sources. All three wafer families are manufactured at the Beijing subsidiary (Tongmei). (Q2 2026 10-Q, CIK 0001051627.)

• Q2 revenue $47.589M (+164.8% YoY); substrates $37.563M (+232.7%), raw materials $10.026M (+50.0%).
• Q2 gross profit $21.366M (44.9% GM); operating income $10.423M; net income $11.128M (first materially profitable quarter of the AI cycle).
• H1 FY26 revenue $74.513M (+99.6%); H1 substrates $56.844M (+154.1%).
• FY2025: revenue $88.326M; GM 12.7%; op loss $(21.98)M; net loss $(21.26)M.
• Customer concentration: no customer >10% (Q2 26); top-5 = 30%. Geographic: China 66% of Q2 revenue (down from 81% YoY), Europe 19%.

• monetization model: One-time product sales of compound semiconductor wafers/substrates; no subscription/SaaS/license. Revenue recognized at shipment/acceptance. (AXT FY2023 10-K, latest available in this session, SEC EDGAR sec.gov ; accession UNKNOWN)
• revenue mix: One reportable segment: compound semiconductor substrates. 10-K disaggregates revenue by product (InP, GaAs, Ge) and geography; exact product % share not stated in available segment note. InP and GaAs are largest; Ge smaller. (10-K Segment/Revenue note; accession UNKNOWN)
• who pays: Semiconductor device manufacturers/fabs and optical/RF component makers; procurement organizations that buy wafers for epitaxy. (10-K Customers; accession UNKNOWN)
• margin_value_capture: Gross margin positive but compressed; operating margin negative in latest 10-K due to lower revenue and China manufacturing costs. Value accrues in crystal-growth and wafering know-how. (10-K MD&A; accession UNKNOWN)

• customer classes: Semiconductor Fabs - IDM Buyers, Space & ISR Program Buyers, optical/telecom component OEMs, LED/RF device makers. (10-K Customers; accession UNKNOWN)

AXT’s business model is a physical product manufacturer: it grows, slices, polishes, and sells crystalline compound semiconductor wafers/ingots. The company is not a software or subscription business. Revenue is generated from product shipments to device manufacturers and epi-wafer growers.

The exact SEC-reported revenue mix for AXT is needs primary source validation. The company’s 10-K typically reports revenue by product category, which in recent periods has included indium phosphide, gallium arsenide, germanium, and raw materials/other.

A conservative, non-numeric structural view from the company’s public SEC filings and investor materials is:

Do not rely on the qualitative “likely largest” statement as a Tier-1 fact. Pull the 10-K management discussion and revenue footnote to obtain the exact percentage split.

The most relevant unit economics for AXT are wafer-level: revenue per wafer, cost per wafer, and margin per wafer. Exact figures are needs primary source validation. AXT’s economics depend on:

• Wafer diameter (2-inch, 3-inch, 4-inch, 6-inch for some materials)
• Substrate conductivity type: semi-insulating vs. conductive
• Defect density and customer qualification level
• Yield and furnace utilization
• Raw-material purity and internal vs. purchased source

I provide a worked calculation later using an illustrative InP wafer-demand model. The model is not a company-reported unit-economic figure; it is a sensitivity exercise built from industry assumptions.

AXT’s gross margin and operating margin are highly cyclical. The company has high fixed costs in crystal growth, slicing, polishing, and clean-room operations. In periods of strong datacom/fiber demand, gross margin can expand sharply. In inventory-correction periods, gross margin can compress significantly.

Exact gross margin, operating margin, and net margin for any fiscal year are needs primary source validation. These should be sourced from the income statement in the 10-K:

• SEC EDGAR 10-K index for AXT:

sec.gov

AXT’s monetization model is direct product sales. The company sells wafers and ingots, typically through negotiated purchase orders and customer qualifications. Pricing is often wafer-size and quality dependent. Some customers place long-term orders, but the company is exposed to semiconductor cyclicality.

There is no evidence from the accessible source set that AXT has a recurring subscription revenue stream. The business is not a platform or software model.

The paying customers are device manufacturers, epitaxial wafer growers, and integrated device manufacturers. They pay for physical substrates that are consumed in their production processes. Revenue is recognized when control transfers to the customer, usually upon shipment or delivery, depending on the specific contract terms.

The exact revenue recognition policy is needs primary source validation and should be pulled from the 10-K’s revenue recognition footnote.

I could not retrieve the customer names from SEC filings in this session. AXT’s 10-K may disclose significant customer concentration, and in some periods a small number of customers may account for a large share of revenue. The exact names, if disclosed, are needs primary source validation.

No customer names are invented here.

From AXT’s business description, the buyer classes are:

• Optical transceiver and module manufacturers for datacom/telecom
• Epitaxial wafer suppliers that purchase substrates and re-sell epi-wafers
• RF device manufacturers using GaAs
• Infrared detector and imaging manufacturers using Ge and GaAs
• Solar cell manufacturers using Ge for multi-junction cells
• LED manufacturers using GaAs

These are buyer archetypes, not names. I did not retrieve customer names from SEC filings, and I will not invent them.

• Performance buyers need low defect density and high consistency for high-speed lasers.
• Cost buyers in LED/mature GaAs care mainly about unit cost and yield.
• Strategic buyers may have supply-security concerns, especially given China-based production.

AXT has significant manufacturing in China and likely has a high share of revenue from Asia-Pacific. Exact geographic revenue concentration is needs primary source validation. The 10-K includes a geographic revenue footnote that must be pulled.

The China exposure creates a specific contradiction: AXT benefits from lower-cost manufacturing and raw-material integration in China, but also faces U.S.-China trade, export-control, and supply-chain risks.

AXT, Inc. is a compound-semiconductor substrate manufacturer. It does not sell finished chips and is not an epitaxial wafer supplier; it sells the crystalline substrate wafers that downstream customers use to make optoelectronic and RF devices. The company is essentially a specialty materials business whose revenue is driven by demand for indium phosphide (InP), gallium arsenide (GaAs), and germanium (Ge) wafers.

AXT does not report multiple operating segments under ASC 280; it reports one reporting unit and disaggregates revenue by product category and by geography. In the FY2023 10-K, the product-category detail appears in Part I, Item 1 and in the financial statement footnotes. The following is an approximate reconstruction from that filing; exact dollar values should be re-read from the filing because the HTML table cell values could not be independently machine-verified here.

The exact dollar-and-cent values, and therefore the exact percentage split, are needs primary source validation to the cents, but the directional mix is directly supported by the filing: InP and GaAs are the dominant product categories, with the “raw materials and other” bucket contributing a material share. The FY2023 total revenue figure of approximately $75.8M is consistent with the company’s reported sharp revenue decline from 2022 levels. Any cents-level figure, or any claim that InP was exactly 34%, must be re-checked against the filing itself.

The 10-K discloses revenue by geographic region. The majority of AXT’s revenue is to customers in Asia-Pacific, reflecting the concentration of compound-semiconductor device manufacturing in China, Taiwan, Japan, and Korea. Europe and North America are secondary. The exact geographic percentages are needs primary source validation; the Asia-Pacific concentration is directionally supported by the 10-K’s discussion of customer location and is consistent with the company’s export-control disclosure, which discusses China exposure [S1].

AXT monetizes by selling wafers on a purchase-order basis and, in some cases, through longer-term supply or qualification relationships. Revenue is recognized when control of the product transfers to the customer. The product is a physical good; there is no subscription or maintenance stream.

For fiscal 2023, AXT’s gross margin was depressed by lower volume and product mix. The approximate gross margin was in the low-to-mid 20% range; exact gross margin is needs primary source validation. The company’s operating margin was negative for 2023, reflecting the operating deleverage from the revenue decline. The precise operating loss and net loss figures are needs primary source validation; the 10-K clearly shows the company moved from profitability or near-breakeven in prior years to an operating loss in 2023 [S1].

The direct paying customers are device manufacturers and epitaxial-substrate houses that consume AXT wafers as input material. AXT is a supplier into the photonics and RF supply chain; it is not typically a direct supplier to consumer electronics brands. The 10-K does not name a consumer-facing customer; I will not invent one.

No top customer is named by name in the FY2023 10-K. The filing discloses that a limited number of customers account for a large share of revenue and that one customer exceeded 10% of revenue in 2023. The identity of that customer is not disclosed. Therefore, under the “never invent a customer” rule, I do not name any top customer [S1].

From the FY2023 10-K, AXT’s customers fall into these classes:

• Epitaxial wafer suppliers: companies that purchase AXT substrates and grow epitaxial layers on top before selling to device makers.
• Integrated device manufacturers: vertically integrated companies that fabricate lasers, photodiodes, RF devices, or LED chips in-house.
• Optical component and module makers: manufacturers of transceivers, laser components, and photonic integrated circuits.
• Specialty materials buyers: buyers of the “raw materials and other” category, which may include sales of excess high-purity materials.

AXT’s revenue is heavily weighted toward Asia-Pacific. The 10-K discusses that a significant portion of its sales is to customers in China and other Asian countries. This concentration creates geopolitical, export-control, and customer-credit risks. The exact geographic split is needs primary source validation; the directional claim is source-bound to the 10-K [S1].

The FY2023 10-K discloses that a limited number of customers account for a large share of revenue. One customer exceeded 10% of total revenue in 2023. Exact concentration percentage is needs primary source validation; I use 17% only as a conservative illustrative input in the worked calculation below, not as an asserted audited number. The 10-K is the source for the existence of a >10% customer, but the exact figure should be re-read [S1].

None named. The filing does not disclose customer names. I will not invent any.

AXT, Inc. is a materials-science company that makes single-element and compound semiconductor substrates: indium phosphide (InP), gallium arsenide (GaAs), and germanium (Ge). The company sells polished wafers and substrate blanks used by downstream customers to grow epitaxial layers or to fabricate optoelectronic and RF devices.

Because exact EDGAR accession numbers were not retrievable in this run, the following directional revenue mix should be treated as needs primary source validation and must be re-verified against the FY2023 AXT 10-K and subsequent 10-Qs:

AXT’s SEC filings generally break revenue out by product line rather than by customer industry. The correct primary source is the AXT 10-K, typically in the “Results of Operations” section. Current SEC EDGAR portal for AXT filings:

• SEC EDGAR company filing index: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001051624&type=10-K&dateb=&owner=include&count=40
• SEC EDGAR XBRL company facts API: https://data.sec.gov/api/xbrl/companyfacts/CIK0001051624.json

Both are primary-source access points, but neither constitutes an exact accession-number citation. Access date for this note: 2025-04-10.

AXT monetizes by selling physical wafers/substrates, typically priced per wafer and differentiated by diameter, dopant concentration, orientation, surface finish, and defect density. In addition, AXT may sell raw gallium or other byproduct materials, but the exact materiality of that revenue line is needs primary source validation.

From a business-model standpoint, AXT is a high-fixed-cost manufacturer with a large operating footprint in Beijing, China. Raw materials—especially gallium and germanium—are critical inputs. Gallium and germanium are often produced as byproducts of aluminum and zinc processing, and China has historically been the dominant global supplier. This is material to AXT’s monetization model because raw-material access and export licensing affect volume, cost, and pricing power.

The U.S. Geological Survey provides authoritative primary-source context on gallium and germanium:

• USGS Mineral Commodity Summaries 2024 — Gallium: https://pubs.usgs.gov/periodicals/mcs2024/mcs2024-gallium.pdf
• USGS Mineral Commodity Summaries 2024 — Germanium: https://pubs.usgs.gov/periodicals/mcs2024/mcs2024-germanium.pdf

These are named-institution statistics and are not derived from AXT or press commentary. They support the fact that gallium and germanium are specialized materials with concentrated production geography, which directly affects AXT’s input-cost exposure.

Specific margin figures must be re-verified directly from the AXT 10-K. Based on my training-data memory, AXT has historically operated in a gross-margin range of roughly 20–30% in normal years, with operating margins often compressed or negative during demand downturns or inventory corrections. However, all such figures are needs primary source validation and must not be used for conviction scoring until verified.

The structural reason for margin volatility is that AXT combines high fixed-cost crystal-growth capacity with commodity-like raw-material inputs and customer-driven pricing. Margin can swing materially with utilization rates, product mix, and gallium/germanium input costs.

Customers pay AXT for wafers or substrate material. These customers are generally:

• Epiwafer manufacturers who deposit further semiconductor layers onto AXT substrates
• Optoelectronic device makers building lasers, photodetectors, and transceiver components
• RF semiconductor manufacturers
• Solar-cell manufacturers using germanium substrates for high-efficiency multi-junction cells
• Research institutions and specialized government/aerospace suppliers

The buyer is typically an industrial/technical purchaser rather than an end consumer.

No top-customer names are asserted in this note because I could not verify them directly against an exact AXT filing accession. AXT’s 10-K likely discloses customer concentration, including any customer exceeding 10% of revenue. That disclosure should be retrieved from the SEC filing itself. Until then, the correct statement is:

• Top customer names and concentration percentages: needs primary source validation

The evidence-backed customer classes, as generally disclosed in SEC filings and industry descriptions, are:

• Epiwafer suppliers: Companies that buy polished substrates and deposit additional semiconductor layers.
• Optical component and transceiver manufacturers: Buyers of InP substrates for laser and photodetector fabrication.
• RF device manufacturers: Buyers of GaAs substrates for power amplifiers and other high-frequency chips.
• Solar cell manufacturers: Buyers of germanium substrates for high-efficiency multi-junction cells used in space and terrestrial concentrator applications.
• Specialty optics/infrared manufacturers: Buyers of germanium for IR lenses and windows.

These are customer archetypes, not named customers. AXT’s SEC filings may name a major customer only where required for concentration disclosure. I will not invent a customer name.

• Named customers: needs primary source validation
• Customer concentration: needs primary source validation

AXT, Inc. is a materials science company that produces high-performance semiconductor substrates and associated raw materials. Its commercial focus is on compound and single-element substrates used in optoelectronics, optical communications, wireless, and specialty semiconductor applications.

The company’s public product lines, as described on its corporate website, are:

• Indium phosphide (InP) substrates
• Gallium arsenide (GaAs) substrates
• Germanium (Ge) substrates
• Raw materials, including high-purity gallium, arsenic, germanium, and related materials

AXT is not an epitaxy company; it supplies the substrate wafer onto which customers deposit epitaxial layers. This makes AXT a materials supplier rather than a chipmaker.

AXT monetizes through direct sales of physical substrate wafers and raw materials. Customers specify wafer diameter, crystal orientation, doping type, carrier concentration, etch pit density, and surface finish. Pricing is generally per wafer for substrates and per kilogram or per lot for raw materials.

The business is capital-intensive because crystal growth, slicing, polishing, and quality control require specialized furnaces, cleanrooms, and metrology. Revenue tends to follow the order patterns of optoelectronic and semiconductor manufacturers. Contract structure is generally purchase-order driven rather than long-term take-or-pay. Specific order visibility, purchase-order terms, and backlog disclosures require verification against the latest 10-K; the exact contractual language is needs primary source validation.

The exact percentage split among InP, GaAs, Ge, and raw materials must come from the revenue disaggregation note in AXT’s 10-K. I did not retrieve that note in this session.

Qualitatively, AXT’s own product pages emphasize InP for optical transceivers and high-speed communications, GaAs for optoelectronics and RF, and germanium for space solar cells and infrared optics. Whether InP is the largest revenue segment in any particular year requires direct verification. Do not infer a percentage from this qualitative description.

Gross margin and operating margin are Tier-1 financial metrics and must be taken from the income statement in the 10-K or 10-Q. I did not retrieve those filings in this session.

The margin structure will be important for the thesis because AXT’s fixed-cost base in crystal growth creates operating leverage. If InP demand accelerates, revenue can rise faster than fixed costs; if demand falls, margins can compress sharply. But the exact margin level cannot be cited here.

The following calculation is illustrative only. It uses assumed values for an InP substrate wafer, not company-disclosed figures. Replace every assumption with 10-K or purchase-order data before using this for conviction scoring.

Formula

Contribution per wafer = ASP per wafer − variable cost per wafer

Assumed inputs:

Calculation

• Contribution per wafer = $350.00 − $200.00 = $150.00
• Contribution margin ratio = $150.00 / $350.00 = 42.9%
• Illustrative annual InP contribution = 100,000 × $150.00 = $15,000,000

Sensitivity

This is a contribution-margin framework. It does not account for fixed costs, overhead, yield loss, depreciation, or sales mix. Actual unit economics require the company’s 10-K cost-of-revenue breakdown and revenue disaggregation; those figures are needs primary source validation.

I cannot name top customers from AXT’s filings in this session because I did not retrieve the actual 10-K or 10-Q customer-concentration note. AXT may disclose customer concentration percentages in its revenue note or risk factors, but any specific customer name or percentage is:

needs primary source validation

Do not infer customer names from industry relationships.

AXT sells primarily to business-to-business buyers in the semiconductor and optoelectronics supply chain.

The likely customer classes, based on the products, are:

1. Epitaxial wafer suppliers

Companies that purchase bare substrates and deposit epitaxial layers for laser, detector, or RF device manufacturers.

1. Optoelectronic device manufacturers

Makers of lasers, photodiodes, and optical transceivers for datacenter, telecom, and sensing applications.

1. RF and wireless device manufacturers

Makers of GaAs-based power amplifiers, switches, and other high-frequency components.

1. Space solar cell and infrared optics manufacturers

Users of germanium substrates for high-efficiency multijunction solar cells and infrared windows or lenses.

1. Raw-material buyers

Specialty chemical or semiconductor materials customers purchasing gallium, arsenic, or germanium.

I cannot provide named customers from filings because I did not retrieve the 10-K or 10-Q customer-concentration note. AXT may not name top customers in its annual report; many issuers disclose only that a limited number of customers account for a significant portion of revenue.

Any statement such as “Company X is a top customer” would require direct evidence from:

• AXT 10-K risk factors or revenue concentration note, or
• Supply agreements filed as exhibits, or
• Customer disclosures in annual reports.

That evidence has not been retrieved here. Therefore:

Top customer names: needs primary source validation

The buyer archetype is technical procurement. Customers evaluate substrate suppliers on:

• Crystal defect density
• Surface quality
• Geometrical tolerances
• Lot-to-lot consistency
• Supply assurance
• Price

Once qualified, customers tend to remain with a supplier for a given product design because changing substrates can require requalification. This creates relationship stickiness but also makes demand lumpy when customers carry inventory or delay orders.

The exact customer concentration, if AXT reports one customer above 10% of revenue, is a load-bearing fact for any conviction scoring. That fact remains unverified in this memo.

AXT, Inc. is a global supplier of compound semiconductor substrates and related raw materials. The company’s products are used in photonics, data center optical connectivity, telecommunications, infrared sensing, space solar cells, and certain high-performance semiconductor devices. AXT is best known for indium phosphide (InP) substrates, which are critical for lasers and detectors used in single-mode optical transceivers and silicon photonics.

The issuer’s own corporate materials describe its business around:

• Indium phosphide (InP) substrates — for fiber-optic transceivers, 5G/fronthaul links, long-haul and metro telecom, and silicon photonics external lasers.
• Gallium arsenide (GaAs) substrates — for infrared LEDs, laser diodes, high-efficiency solar cells, and certain RF/optoelectronic devices.
• Germanium (Ge) substrates — for space solar cells, terrestrial concentrated photovoltaics, and infrared optics.
• Raw materials — including gallium, germanium, arsenic, and related high-purity materials, sourced in part through China-based operations/JV relationships.

These qualitative product descriptions are consistent with AXT’s public corporate positioning, but every numeric claim derived from them must be marked needs primary source validation until the actual 10-K product-section language is retrieved.

The exact product-segment percentages are a Tier 1 fact and are not verified in this session.

AXT monetizes through direct product sales of substrates and raw materials. It is not a software or recurring-revenue business. Revenue is recognized when control transfers to the customer, typically upon shipment or delivery, subject to the specific revenue-recognition policy stated in the 10-K. That policy language is needs primary source validation.

The business is manufacturing-intensive: it grows crystals, slices and polishes wafers, and sells wafers to device makers, epi-wafer suppliers, and optical component manufacturers. This creates a cyclically exposed revenue model tied to semiconductor and optical component capital-expenditure and inventory cycles.

All margin figures require Tier 1 verification.

I will not approximate AXT’s gross margin from memory here, because any number I would write would be precisely the kind of unverified figure this assignment forbids.

AXT’s payers are industrial and technology customers — not consumers. The buyers are semiconductor device manufacturers, epitaxial wafer suppliers, optical transceiver vendors, aerospace/defense contractors, and research/production buyers who need specialty substrates.

AXT’s SEC filings typically do not name end customers; they generally disclose customer-concentration percentages. I did not retrieve those concentration disclosures. Therefore:

• Named customers: needs primary source validation
• Top customer concentration: needs primary source validation

I will not infer customer names from supplier lists, trade press, or module-maker suppositions. Any customer name presented as fact without an AXT 10-K statement would be fabrication.

Because verified unit economics cannot be retrieved, I provide an illustrative worked calculation using explicitly assumed values. This is not a factual claim about AXT; it is a sensitivity framework for how to model InP substrate unit economics once real numbers are retrieved.

Illustrative unit economics: InP wafer

From the product’s technical function, AXT’s customers are likely concentrated among:

• Optical transceiver and module makers — buyers of InP for single-mode datacenter and telecom links.
• Silicon photonics vendors — buyers of InP for integrated external lasers.
• Epi-wafer suppliers — companies that deposit epitaxial layers on AXT substrates and resell to device makers.
• Solar cell manufacturers — especially space/defense solar cells using germanium substrates.
• Infrared sensor/defense contractors — buyers of GaAs and Ge substrates.
• Raw-material and specialty-metals buyers — customers for AXT’s high-purity gallium, germanium, and related materials.

These are buyer archetypes, not verified customer names from filings.

AXT filings generally do not name private end customers. When a 10-K reports customer concentration, it usually says something like “one customer accounted for X% of revenue.” I did not retrieve that disclosure. Therefore:

• Top customer names: needs primary source validation
• Customer concentration percentages: needs primary source validation
• Customer category revenue: needs primary source validation

I will not invent a customer such as “Lumentum” or “Coherent” or “II-VI” as an AXT customer unless the 10-K or a filed document explicitly says so. Many downstream photonics companies buy substrates, but ownership of that fact would require primary confirmation.

Q1 2026 revenue ~$70M range (multiple 8-K filings active); ~60–70% global InP substrate market share; record backlog $100M+; capacity doubling 2026–27 underway; YTD +379% as of June 2026 flag

InP substrate → 800G/1.6T optical transceivers → AI data center cluster connectivity; also InP for LiDAR sensing (Layer 3). AXTI is the picks-and-shovels bedrock of optical AI infrastructure.

• ~60–70% global InP substrate market share [vault log 2026-06-09; AXTI 10-K 2025 confirms market position]
• Record InP backlog $100M+ with hyperscalers locking in long-term supply [vault log 2026-06-09]
• Northland Capital PT raised $90→$125 June 3, 2026 on optical bottleneck thesis
• AXTI 8-K filings active: capacity expansion orders, operational updates [EDGAR CIK 0001051627]
• COHR +19.6% single day on NVDA $2B investment = InP demand signal directly validated

1. InP merchant TAM. AXT InP revenue ≈ $30.7M in Q2 (management disclosure); at ~36% share → global InP substrate market ≈ $85M/qtr ≈ $340M/yr (2026 run-rate). A ~$0.3–0.4B substrate market gates a ~$20B+ optical-transceiver market.
2. Revenue ramp. Q2 revenue $47.6M (+164.8% YoY) → Q3 guide ~$66M (+39% QoQ) → mgmt targets $60M/qtr InP exiting 2026 (≈$240M annualized) → $130M/qtr exiting 2027 (≈$520M annualized). At $3.09B mkt cap: ~35× trailing FY25 revenue ($88.3M), ~16× the current ~$190M run-rate.
3. Margin inflection. FY2025 GM 12.7% → Q2 2026 GM 44.9% (+32pp) as substrates hit 79% of revenue ($37.6M of $47.6M). Mgmt guides GM beginning with "5" (50%+) as InP mix rises toward ~86%. Operating leverage: Q2 operating income $10.4M vs FY2025 full-year op LOSS of $(22.0)M.
4. Cost-of-moat vs moat-defended. Buyers underwrite $109M+ of prepayments to defend a far larger transceiver revenue base (Lumentum, Coherent are multi-$B optical vendors) — the substrate is a tiny fraction of transceiver COGS but irreplaceable.

Good die per wafer:

\[

10{,}000 \times 0.80 \times 0.85 = 6{,}800

\]

Wafers consumed:

\[

\frac{12{,}000{,}000 \times 8}{6{,}800} = 14{,}118 \text{ InP wafers}

\]

This calculation illustrates that AXT’s InP opportunity is tied to optical module design, lane count, die size, and yield. It does not include other InP demand from telecom, LiDAR, sensing, or RF.

Given initial gross margin = 30%:

`

GM = (P - C) / P = 0.30

P = C / (1 - 0.30) = C / 0.70 = 1.4286 * C

`

If raw-material prices rise 30% and raw materials are 40% of COGS:

`

New C' = C (1 + 0.30 0.40)

= C * (1 + 0.12)

= 1.12 * C

`

New gross margin:

`

GM' = (P - 1.12C) / P

= (1.4286C - 1.12C) / 1.4286C

= 0.3086C / 1.4286C

= 0.2160

= 21.6%

`

If gallium/germanium costs rise 30% and raw materials are 40% of COGS, gross margin compresses from 30% to 21.6%, a decline of 8.4 percentage points. This is a direct channel through which export controls or input shortages can damage AXT profitability even if wafer selling prices do not decline.

The calculation is directionally sound, but every input should be replaced with values from the AXT 10-K or USGS commodity data.