Phase A — Understand the business
Lens 1 · Company Overview
AXT (American Xtal Technology, renamed AXT in 2000; Delaware corp, HQ Fremont CA; founded 1986; NASDAQ: AXTI) is a materials-science company that grows and sells high-performance compound and single-element semiconductor substrates ("wafers") — it does not design or fab chips. It sits "at the beginning of the semiconductor food chain". Two product lines:
- Specialty substrates (67% of 2025 revenue): indium phosphide (InP), gallium arsenide (GaAs, semi-insulating + semi-conducting), and germanium (Ge) wafers in 2"–8" diameters.
- Raw materials (33% of 2025 revenue): via consolidated subsidiaries — purified gallium (6N/7N Ga), pyrolytic boron nitride (pBN) crucibles, boron oxide, etc., used internally and sold on the open market.
The product split was 67% / 33% in 2025, 68% / 32% in 2024, 63% / 37% in 2023.
The whole company runs through China. Every substrate and raw material is manufactured in the PRC by Beijing Tongmei Xtal Technology Co., Ltd. ("Tongmei") and a constellation of consolidated PRC subsidiaries and JVs (Baoding/ChaoYang Tongmei, JinMei, BoYu, ChaoYang XinMei/LiMei/KaiMei, etc.). AXT is a Delaware company that owns ~85.5% of Tongmei; it is not a VIE structure.
Contract structure: short-cycle purchase orders, single performance obligation, usually <6 months, recognized on shipment/receipt — no take-or-pay, no recurring contracts, no backlog obligation on the customer side ("customers are not obligated to purchase a specified quantity… may reduce, delay or cancel orders"). So the headline "$100M InP backlog" is demand-signal, not contractually firm.
Customers: a majority specialize in epitaxial growth — they grow chemical layers on AXT's wafers before making devices. End applications: AI/data-center optical transceivers, 5G/telecom, fiber optics, LEDs, lasers, VCSELs (3-D sensing), satellite/space solar (Ge). The most strategic end-use today is InP for high-speed optical data transmission in AI data centers — the explicit growth thesis management repeats in both filings.
Lens 2 · Supply Chain
Map: upstream raw inputs → AXT's vertically-integrated PRC raw-material companies → Tongmei crystal growth + wafering → epi houses / device makers → optical-module & systems OEMs → hyperscaler data centers.
- Upstream / raw materials (AXT owns this): AXT is, on its own telling, "the only compound-semiconductor substrate supplier to have a position in raw materials". Consolidated raw-material companies (JinMei — purified gallium + InP base material; BoYu — pBN crucibles + OLED tooling; ChaoYang XinMei/ShuoMei) plus unconsolidated equity-method JVs (e.g. JiYa, Xiaoyi XingAn) supply gallium, arsenic, germanium, pBN, B2O3. This is the moat's foundation (Lens 3) and a revenue diversifier.
- The chokepoint is the Chinese state, not a supplier. Since Aug 2023 China requires export permits for GaAs and Ge; on Feb 4, 2025 China added InP to its export-control list — so all three substrate families now need Ministry of Commerce permits to leave China. InP permits to Europe and Japan began June 11, 2025; no permit to ship InP (or GaAs) to the U.S. has been granted — U.S. GaAs customers are "dual-use".
- Downstream named buyers: the filings don't name customers (no customer >10%), but the InP end-market is optical transceivers / silicon-photonics for AI; a public chain marker — Nokia began photonic-IC manufacturing on 6-inch InP wafers (AIXTRON G10-AsP) in April 2025, illustrating the 6-inch InP pull AXT is positioning for.
- Single-source / concentration risk: manufacturing is 100% China, single-country. A factory shutdown (air-pollution mandates), a permit freeze, or a tariff escalation hits everything at once. There is no geographic redundancy — the opposite of resilient, by design (it's the low-cost moat, Lens 3).
Names-or-it-didn't-happen suppliers/peers along the chain: AXT (substrate), Sumitomo Electric, JX Nippon Mining & Metals (JX), Freiberger Compound Materials, Umicore, China Crystal Technology (CCTC), Vital Materials, Xiamen Powerway, Visual Photonics Epitaxy (peers); AIXTRON (epi-tool maker enabling 6-inch InP).
Lens 3 · Competitive Advantages (moats)
The moat is narrow, real, and concentrated in InP — and it just got handed a regulatory super-charger.
- Best-in-class InP in a 3-supplier oligopoly. AXT claims its InP has the lowest defect densities / lowest stress & slip lines on the market, and that "there are only three primary suppliers, including AXT". Independent market data corroborates the oligopoly: the top five InP suppliers (Sumitomo, AXT, Freiberger, JX, Visual Photonics) held ~70% of 2024 revenue.
- The 6-inch InP frontier. AXT believes it "can be the dominant supplier in the emerging 6-inch diameter market," where its quality edge widens with diameter — and that this converges with AI-driven data-center demand.
- Vertical raw-material integration — unique among substrate peers; pricing/lead-time advantage and a second revenue stream.
- Low-cost China manufacturing vs. competitors based in Germany (Freiberger) / Japan (Sumitomo, JX) — also the single greatest source of risk (Lens 13).
- Proprietary VGF crystal-growth process technology (barrier to entry, especially InP).
- The accidental moat — China's export regime. Because AXT manufactures inside China and has received InP permits for Europe/Japan, it is, for now, one of the few suppliers that can both make InP at scale and (partly) ship it. The same control that blocks the U.S. channel blocks Western rivals from sourcing Chinese InP — pushing 6-inch InP wafer prices up ~250% to ~$5,000. AXT is simultaneously a victim (lost its U.S. channel) and a beneficiary (scarcity pricing + strategic status) of the same policy.
Bargaining power: weak on customers historically (commoditizing, ASP declines of 5–10%/yr in normal times ) — but temporarily strong in InP while supply is short. Power over suppliers is high because it owns much of its own raw-material chain.
Durability check: moats 1–5 are genuine but modest (this was a sub-$100M-cap microcap for a reason). Moat 6 is policy-dependent and reversible — it can evaporate the moment Beijing widens permits, or invert if Beijing freezes them. The market is currently capitalizing moat 6 as if it were permanent.
Lens 4 · Segments
AXT reports one operating segment, two product lines; it does not publish segment EBIT. Revenue by product and geography (all ``, $ thousands):
By product line:
| Product | 2025 | 2024 | 2023 | 2024→2025 |
|---|
| Substrates | 58,900 | 67,748 | 47,466 | −13.1% |
| Raw materials & other | 29,426 | 31,613 | 28,329 | −6.9% |
| Total revenue | 88,326 | 99,361 | 75,795 | −11.1% |
The 2025 substrate decline was driven by a deliberate cut to germanium (raw-material cost spiked, AXT chose to walk) and the InP export-permit hit — InP demand rose but recognizable revenue fell because permits gated shipments. 2024 was the boom (InP for 5G/silicon-photonics + GaAs LED).
By geography (2025 / 2024 / 2023 share):
| Region | 2025 $k | 2025 % | 2024 % | 2023 % |
|---|
| China | 55,076 | 62% | 56% | 53% |
| Taiwan | 13,039 | 15% | 14% | 11% |
| Europe (mostly Germany) | 11,156 | 13% | 14% | 16% |
| Japan | 4,906 | 5% | 5% | 6% |
| Asia-Pac (ex-CN/TW/JP) | 2,441 | 3% | 3% | 5% |
| North America (mostly US) | 1,708 | 2% | 8% | 9% |
The single loudest segment fact: North America collapsed −77.5% to just 2% of revenue in 2025 (and ~1% in Q1-2026 ) — a direct consequence of China's InP export controls cutting off the U.S. channel. 98% of revenue is now ex-North-America. Trend: decelerating then inflecting — FY2025 was the trough; Q1-2026 reaccelerated (Lens 5).
Phase B — Measure performance
Lens 5 · Earnings Result
Two prints matter: the FY2025 trough and the Q1-2026 inflection.
FY2025 (10-K) — all ``, $ thousands:
| Line | 2025 | 2024 | 2023 |
|---|
| Revenue | 88,326 | 99,361 | 75,795 |
| Cost of revenue | 77,084 | 75,525 | 62,477 |
| Gross profit | 11,242 | 23,836 | 13,318 |
| Gross margin | 12.7% | 24.0% | 17.6% |
| SG&A | 24,169 | 24,096 | 22,806 |
| R&D | 9,049 | 14,543 | 12,081 |
| Loss from operations | (21,976) | (14,803) | (21,569) |
| Equity in income of JVs | 765 | 3,439 | 1,884 |
| Net loss | (23,202) | (11,791) | (19,193) |
| Net loss attrib. to AXT | (21,260) | (11,624) | (17,881) |
| EPS (basic & diluted) | (0.49) | (0.27) | (0.42) |
The structural tell: operating expenses ($33.2M) dwarf gross profit ($11.2M) — at 2025 revenue, AXT loses money at the operating line by design; it has lost money from operations for three straight years (−$22.0M / −$14.8M / −$21.6M). GM halved in 2025 on under-absorption ("fixed costs spread over fewer units… manufacturing variances, particularly in the first half"). R&D was cut 37.8% — a stress signal in a tech-moat business.
Balance-sheet flags (FY2025):
- Cash & equivalents $120.3M + restricted $8.1M = $128.4M (up from $33.8M) — funded by the Dec 2025 raise: $93.9M net at $12.25/share. ~$27M of cash sits offshore in China.
- Inventories $81.7M against $77.1M COGS — roughly 12 months of inventory, with a $28.4M excess-&-obsolete reserve (up from $24.1M). This is the loudest balance-sheet number in the file (Lens 10).
- Short-term China bank loans $62.8M (current); total liabilities only $99.1M; total stockholders' equity $296.6M; accumulated deficit −$64.9M.
- Redeemable noncontrolling interest $38.1M — the Tongmei PE redemption overhang (~$49M at original price), in temporary equity.
Q1-2026 (10-Q) — the inflection, $ thousands:
| Line | Q1-2026 | Q1-2025 |
|---|
| Revenue | 26,924 | 19,356 |
| Cost of revenue | 18,946 | 20,597 |
| Gross profit (loss) | 7,978 | (1,241) |
| GAAP gross margin | ~29.6% | negative |
| Loss from operations | (1,585) | (10,275) |
| Net loss | (1,485) | (10,016) |
Revenue +39% YoY, gross margin swung from negative to ~30%, operating loss narrowed to near-breakeven (−$1.6M), EPS ~−$0.01 vs consensus −$0.04 (a beat). InP revenue $13.6M = 50.6% of total, driven by AI data-center optics; Q2 guided to record InP revenue >$17M. GaAs fell to $5.4M, license-constrained. Market reaction: the stock had already run ~155% since late Feb 2026 into and after this print — the tape is pricing the inflection as the start of a hyper-growth ramp.
Lens 6 · Earnings Calls (sentiment trend)
No transcripts on the shelf (transcripts/ empty); this is ``-grounded from release commentary across the last ~4 quarters.
- Tone arc: FY2024 call → optimism on InP/5G/silicon-photonics (record-ish year). Through 2025 → defensive/anxious: the recurring phrase in the filings is that permit fluidity "creates anxiety for us and for our customers"; management cut R&D, walked Ge, and watched the U.S. channel vanish. Q1-2026 → emphatic bullishness: CEO Morris Young, "InP substrates are a key ingredient in high-speed optical data transmission required in AI-focused data centers"; "China's InP-based laser market more than doubled QoQ"; record $100M backlog; doubling capacity.
- Phrases that arrived: "AI-focused data centers," "record InP backlog," "doubling capacity," "best in class," "6-inch." Phrases that faded: germanium growth, U.S. demand, 5G-handset.
- Net: sentiment has flipped from survival-mode to growth-narrative inside three quarters — and management is now actively selling the AI-InP story into capital markets (two large equity raises, below). Read this with a forensic eye (Lens 10): the same team is issuing stock aggressively into its own bullish narrative.
Lens 7 · Comps
| Company | Mkt cap | EV/Sales | EV/EBIT | P/E | Div yld | 5-yr avg ROE |
|---|
| AXT (AXTI) | ~$5.88B | ~60–67× TTM sales | n/a — negative EBIT | n/a — loss-making | 0% | negative (accumulated deficit; valuation allowance on DTAs) |
| Sumitomo Electric | n/a as pure-play | n/a | n/a | n/a | n/a | n/a |
| JX (ENEOS) | n/a as pure-play | n/a | n/a | n/a | n/a | n/a |
| Freiberger | private — not disclosed | n/a | n/a | n/a | n/a | n/a |
| Umicore | n/a as pure-play | n/a | n/a | n/a | n/a | n/a |
The only comp that matters is AXT vs. its own history and its own forward ramp. A ~$5.9B market cap on $88M trailing revenue ($96M annualizing Q1) is ~60–67× sales for a business at 13–30% gross margin with negative operating income. That multiple is only defensible if you believe the capacity-doubling roadmap (to $35M/qtr InP by end-2026, $65–70M/qtr by 2027 ) lands and InP scarcity pricing holds and China keeps the permits flowing. It is priced as a structural-growth compounder, valued like AI infrastructure, on the financials of a turnaround microcap. The comp set, properly labeled, can't justify the price — only the narrative can.
Lens 8 · Stock-Price Catalysts
The >5% moves over the relevant window (all ``) — and the pattern is unusually clean: InP supply-shock + China-permit headlines + capacity/raise news are the only things that matter.
- 52-week range: $1.80 → $143.16; now ~$92. An ~80x move and a ~36% pullback from the peak — extreme regime change for a name that traded at $2.09 in June 2025.
- Feb 4, 2025 — China adds InP to export-control list. The double-edged catalyst: near-term revenue hit, long-term scarcity thesis seeded.
- June 11, 2025 — first InP export permits (Europe/Japan) — de-risking signal.
- Dec 29–30, 2025 — prices a public offering at $12.25/share, ~$100M gross / $93.9M net. Stock had already re-rated ~6x off the June low into this.
- Late Feb 2026 onward — InP-shortage reports (6-inch InP +~250% to ~$5,000) ignite the move; stock +~155% from 2/28/2026.
- Apr 30, 2026 — Q1-2026 beat (rev +39%, GM ~30%, record InP backlog).
- ~Apr 22, 2026 — prices a $550M offering at $64.25/share (8.56M shares; up to $632.5M w/ greenshoe). Raising at 5x the December price — capital-markets opportunism on the InP/AI narrative.
What the tape reveals: AXTI is now a pure thematic / supply-scarcity vehicle — it trades on InP-price headlines, China policy, and capacity news, not on diversified fundamentals. That cuts both ways: the same sensitivity that drove $1.80→$143 will drive the drawdown if InP pricing softens or Beijing changes the permit posture.
Phase C — Judge people & books
Lens 9 · Management
- Dr. Morris S. Young — co-founder (1986), Chairman (since Aug 2021), CEO (1989–2004, then CTO, reappointed CEO July 2009). PhD metallurgy (Polytechnic Inst. of NY); ex-AT&T Bell Labs, ex-Lawrence Livermore. 40-year founder — deep materials-science credibility, which is the right archetype for a crystal-growth moat.
- Skin in the game: owns ~3.46%, worth ~$256M at the current price. Total comp ~$1.67M (≈29% salary / 71% bonus+equity). Genuine alignment — but note that the bulk of that $256M paper wealth was minted by the 2025–26 re-rating, which sharpens the incentive to keep the narrative aloft.
- Gary L. Fischer — CFO (named, with Young, in the securities litigation — Lens 10).
- Capital-allocation history: mixed, and now pivotal. For years AXT plowed capital into China capacity and JV stakes while generating cumulative operating losses (accumulated deficit −$64.9M) and negative operating cash flow in 2024–25. No buybacks (program dormant since 2015), no dividend. The defining capital-allocation act is the 2025–26 equity issuance — ~$650M raised across two offerings ($100M @ $12.25, $550M @ $64.25) to fund Tongmei's InP capacity doubling/quadrupling. Read generously: brilliantly-timed, fortress-building, sells stock when the market gifts a price — exactly what a shareholder-minded operator should do. Read skeptically: serial dilution at the top of a thematic spike, deploying the proceeds into single-country China capex whose output is hostage to an export-permit regime.
- Red flags: (i) active securities class action naming CEO+CFO over FY21–24 disclosures; (ii) the China corporate structure (transfer pricing reviewed only "annually with our auditor"; dividends need SAFE approval; PCAOB-inspection caveats); (iii) the Tongmei STAR-Market saga is now in its 5th year with no listing — a long-promised value-unlock that keeps slipping.
- Founder vs. professional: decisively founder-operator — high conviction, technical, long-tenured, now richly incentivized to the upside. Good for moat stewardship; a reason to scrutinize the bullish narrative he's monetizing.
Lens 10 · Forensic Red Flags
Forensic lens — every figure `` unless noted.
- Inventory is the headline risk. $81.7M inventory vs. $88.3M revenue (~0.92×) and $77.1M COGS (~12 months on hand), carrying a $28.4M excess-&-obsolete reserve (up from $24.1M; +$4.3M YoY). For a company whose key product (InP) was export-blocked to its highest-value market (US) and whose Ge line was deliberately curtailed, a year's worth of inventory is a material obsolescence/write-down risk. If demand for any line disappoints, "additional inventory adjustments… could have a material impact" — management's own words.
- Cash flow diverges from the narrative. Operating cash flow was −$12.8M (2025) and −$12.1M (2024) — two straight years of cash burn from operations, masked at the cash line only by the $107.1M financing inflow. The business does not yet self-fund.
- Equity-method JV income flatters the P&L and is volatile — $0.8M (2025) vs $3.4M (2024); it sits below the operating line and swings with Chinese gallium pricing.
- Full valuation allowance on deferred tax assets — $27.2M (up from $20.7M) — management's own statement that future U.S. profitability is uncertain.
- Redeemable NCI / $49M redemption overhang — if the STAR IPO fails or is cancelled, the 10 PE funds can force AXT to redeem ~$49M (carried $38.1M in temporary equity). Classified as not-currently-probable, so issuance costs aren't amortized — a judgement that flips if IPO odds fade.
- China-structure opacity: transfer-pricing between AXT and PRC subs (reviewed annually), SAFE-gated cash repatriation, 10% PRC withholding on dividends, and PCAOB-inspection risk on PRC-headquartered audit firms (auditor BPM is US-based, mitigating this).
- SBC $3.3M (modest, ~3.7% of revenue) — not a major non-GAAP flatter here, but note Q1-2026's ~30% GAAP GM is the real number, and the "non-GAAP gross margin 29.9% vs −6.1%" framing is doing narrative work.
- Nasdaq sub-$1.00 delisting risk is explicitly disclosed (the stock has closed below $1 before) — moot at ~$92 today, but a reminder of how recently this was a sub-$1 candidate.
Regulatory findings (required sub-section) — from regulatory/regulatory-findings.md + filings + web:
- SEC EDGAR EFTS (LR + AAER), 2021–2026: No SEC Litigation Releases and no AAERs naming AXT.
- 10-K Item 3 (the company's own disclosure): A putative securities class action is live. Filed May 6, 2024 (E.D.N.Y., transferred to N.D. Cal.) against AXT, CEO Morris Young and CFO Gary Fischer, alleging §10(b)/§20(a)/Rule 10b-5 violations over a class period March 24, 2021 – April 3, 2024; defendants' motion to dismiss is fully briefed and pending. A related derivative suit (filed Aug 22, 2024) naming officers/directors was dismissed March 17, 2025 (no pre-suit demand; failure to plead) and is on appeal to the 9th Circuit. AXT calls the claims "meritless" and cannot estimate a loss range.
- Non-SEC enforcement (web search): no material FTC/DOJ/FDA/CFPB enforcement actions surfaced. The operative external constraints are China's export-control regime and US Section 301/232 tariffs, not US agency enforcement.
- Net: clean on SEC accounting enforcement; one unresolved securities class action over historical disclosures is the live legal overhang. Verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-18.
Phase D — Project & stress-test
Lens 11 · Forward Projection
Bottom-up from the Q1-2026 run-rate + management's capacity roadmap. Output ``; this is a loss-making turnaround, so the right unit is revenue/GM trajectory and path-to-profitability, not a precise EPS line.
Base inputs:
- Q1-2026 revenue $26.9M (~$108M annualized) with InP $13.6M/qtr; Q2 InP guided >$17M.
- Capacity roadmap: InP to ~$35M/qtr by end-2026 ($30–40M capex), to
$65–70M/qtr by 2027 ($100M capex). Total revenue is InP + GaAs + Ge + raw materials.
- GM trajectory: ~30% at current InP mix/scarcity pricing, with operating leverage as volume absorbs the fixed PRC cost base.
FY2026 (base): revenue ~$120–135M, GM ~28–31%, operating result roughly breakeven to slightly positive as gross profit ($35–40M) approaches opex ($38–40M). EPS ~breakeven, call −$0.05 to +$0.05 on ~63M shares.
FY2027 (base): if the second capacity double lands and scarcity pricing holds, revenue ~$180–230M, GM ~30–33%, first full-year operating profit plausible — EPS ~+$0.30 to +$0.60.
FY2028 (base): revenue ~$230–300M, EPS ~+$0.60 to +$1.10 if InP secular demand + permits + pricing all cooperate.
- Bull path: InP scarcity persists, 6-inch wins ramp, US permits eventually granted (re-opening 8–9% of TAM), capacity fills → FY2027 revenue >$250M, EPS >$0.80, and the multiple is "justified" retroactively.
- Bear path: InP pricing normalizes as Chinese/Western supply responds, Beijing widens permits (commoditizing AXT's scarcity edge) or freezes US/EU permits, capacity comes on into softening price → revenue stalls ~$110–130M, GM reverts toward the teens, back to operating losses and a brutal de-rate from ~60× sales.
The reframe that matters: even on the base path, FY2027 EPS of ~$0.30–0.60 against a ~$92 stock is a ~150–300× forward P/E. The bull case isn't "will AXT grow" — it clearly is, right now — it's "is 60× sales / 150×+ forward earnings already paying for several years of flawless execution in a China-gated, policy-dependent business." Forecast not logged (forecast.ts create skipped per --watchlist breadth rules).
Lens 12 · Bull vs Bear
Bull case. AXT owns a genuinely scarce asset — best-in-class InP, one of three Western-accessible suppliers — exactly as AI data centers discover they need vast quantities of high-speed optical interconnect, and exactly as China's export controls choke off competing supply (6-inch InP +250% to ~$5,000). The Q1-2026 inflection is real and verifiable: revenue +39%, gross margin from negative to ~30%, near-breakeven operating, record $100M InP backlog, InP now >50% of revenue. Management raised $650M at escalating prices — the balance sheet is now a fortress ($650M+ cash post-April raise, minimal leverage) fully funding a capacity quadrupling. The vertical raw-material integration and 40-year founder-operator are durable edges. If InP becomes the picks-and-shovels of AI optics and AXT executes the ramp, today's revenue is a fraction of the 2027–28 base, and the stock is an AI-infrastructure compounder still early in its S-curve.
Bear case (permanent-impairment risks). (1) Single-country China manufacturing under an export-permit regime AXT does not control — Beijing can freeze permits (it already blocks the US channel) or, conversely, widen them and dissolve the scarcity that is the bull thesis. The moat and the risk are the same policy. (2) The price has lapped the business — ~$5.9B cap on ~$88–108M revenue, loss-making, is ~60× sales and ~150×+ base-case forward earnings; even flawless execution may not earn into it, and any wobble (InP price, a permit headline, a soft quarter) invites a 40–60% de-rate (the stock already fell ~36% from its $143 peak). (3) Inventory/obsolescence — $81.7M (12 months) with a $28.4M reserve is a latent write-down. (4) Commodity reversion — substrates normally see 5–10% annual ASP declines; scarcity pricing is temporary by nature.
Pre-mortem (18 months out, thesis broke): It's late 2027. China quietly broadened InP export permits and domestic Chinese suppliers (CCTC, Vital, Powerway) ramped; 6-inch InP prices round-tripped toward pre-shock levels. AXT's freshly-doubled capacity came on into a softer price; GM slid back to the high-teens; the $650M war-chest is now half-spent on China capex that's under-utilized. Revenue plateaued ~$130M, operating losses returned, and the stock is back in the teens — the $1.80→$143→$92→$15 round-trip of a thematic supply-shock vehicle. The securities class action settled, a reminder the market had over-trusted the narrative.
Are multiples too high? Yes, on any sober reading — ~60× sales prices in multiple years of perfect execution in a business whose key advantage is a reversible policy. Contrarian view (what the market refuses to see): the bears' "it's a China policy bet" is right, but the market is under-weighting how strategically China may want a captive, permitted national champion in compound semiconductors — Tongmei's STAR listing could yet unlock a Chinese strategic/sovereign bid that re-rates the asset on Chinese, not US, terms. That optionality is real — and unpriceable.
Lens 13 · Devil's Advocate (short-seller)
Dismantling the bull case.
- The moat is rented from Beijing. AXT's scarcity edge exists only because China restricts InP exports. The same authority can (a) freeze AXT's permits (it already blocks US/GaAs shipments — "dual-use"), instantly impairing revenue, or (b) flood permits/back domestic rivals, collapsing the ~$5,000 InP price that the entire bull thesis capitalizes. You are long a company whose competitive advantage is a Chinese policy lever it cannot influence.
- Revenue concentration risk is geographic and political, not customer. No customer >10% (good), but 98% of revenue is ex-North-America and 62% is inside China — a single-country manufacturing base selling largely into a politically fraught corridor. One factory shutdown (air-pollution mandate), one permit freeze, one tariff jump hits the whole P&L.
- Most dangerous competitor bulls underrate: domestic Chinese substrate makers (CCTC, Vital Materials, Xiamen Powerway) — China's explicit drive for semiconductor self-sufficiency plus the same export controls that help AXT today incentivize Beijing to grow Chinese InP champions; AXT's own 10-K flags "a company established by a former employee in China" already competing in GaAs.
- Capital allocation: serial equity issuance into a thematic spike (~$650M, $12.25→$64.25) is shareholder-friendly if the capex earns a return — but it's deploying into single-country China capacity hostage to permits, while the company has never generated sustained operating profit (accumulated deficit −$64.9M, three years of operating losses). The Tongmei STAR listing has been "coming months" away for five years.
- What must hold for ~$92: InP scarcity pricing persists for years; permits keep flowing to EU/Japan (and ideally the US); the capacity quadrupling fills at ~30%+ GM; no Chinese policy reversal; the securities suit doesn't surprise. That's a long conjunction of must-trues.
- Growth-disappoints-20–30% scenario: at ~60× sales, a 20–30% revenue miss or a GM reversion to the mid-teens doesn't trim the stock — it halves or worse. There is no valuation floor; this is a momentum/narrative quote.
- Single permanent-impairment scenario, and its plausibility: China freezes or politicizes InP export permits (retaliation in a US-China escalation) → AXT's shippable InP collapses, inventory writes down, losses return. Plausibility: moderate and structural — it's the central, ever-present risk the bull case waves away.
Lens 14 · Management Questions (ordered by information value)
- What is your realistic, probability-weighted timeline for InP export permits to the United States, and what specifically must change in Beijing for them to be granted?
- If China broadened InP export permits industry-wide tomorrow, what happens to 6-inch InP pricing and to your gross margin — i.e., how much of your current ~30% GM is scarcity rent vs. structural?
- You've raised ~$650M at $12.25 then $64.25. Walk through the return on that capex — what InP revenue and GM does the doubled/quadrupled capacity need to generate to clear your cost of capital, and by when?
- Of the $100M InP backlog, how much is firm purchase orders vs. forecast/intent, and what cancellation experience have you seen historically when demand softened?
- After five years, what is the honest status and remaining gating items for the Tongmei STAR-Market listing — and what is your plan if the CSRC never approves it (the ~$49M redemption)?
- Your inventory is ~12 months of COGS with a $28.4M obsolescence reserve. What is the composition by product, and what's the write-down exposure if Ge or GaAs demand stays soft?
- Who are your most credible domestic-Chinese InP/GaAs competitors, how fast are they closing the quality gap, and how does China's self-sufficiency drive change your 3-year competitive position?
- What share of InP demand is 6-inch today vs. 2-/3-/4-inch, how fast is the mix shifting, and where is your 6-inch yield relative to Sumitomo/Freiberger?
- How exposed is the InP-for-AI thesis to a shift toward co-packaged optics / silicon photonics / alternative materials that could reduce InP content per port over time?
- You cut R&D 37.8% in 2025. How do you sustain a technology moat (especially 6-inch InP) while cutting R&D, and what's the go-forward R&D plan now that you're cash-rich?
- How should investors think about cash trapped in China (~$27M offshore, SAFE-gated repatriation, 10% withholding) given most of the $650M is destined for PRC capex?
- What is the path and timeline to sustained positive operating cash flow — at what quarterly revenue and GM does AXT self-fund?
- What is the status of the securities class action, and what disclosure changes have you made since the FY21–24 class period?
- How do US Section 301/232 tariffs (now up to 70%, post-IEEPA) factor into your decision not to build any non-China capacity to serve Western customers?
- If you could permanently de-risk one thing — China permits, InP price durability, or execution on the capacity ramp — which matters most to the equity value, and why?