Type: thesis-shift (subtype: hardens; sub-event continues 2026-05-06 structural-verifications entry)

event_id: 2026-05-06-cohr-q3-print-axti-triple-confirmed

Sources:

• Coherent Q3 FY2026 press release (GlobeNewswire, 2026-05-06): record revenue $1.81B (+21% YoY), non-GAAP EPS $1.41 (beat $1.39 est, +55% YoY), non-GAAP gross margin 39.6%, Q4 guide $1.91B–$2.05B, backlog visibility extends to calendar 2028, NVDA $2B equity stake + multiyear CPO supply agreement confirmed in the print itself.
• Coherent Q3 FY2026 earnings transcript (Motley Fool, 2026-05-06 4:30 PM ET) — full prepared remarks from CEO James Robert Anderson + CFO Sherri R. Luther.
• Coherent CHIPS Act PMT (Dec 2024): $33M for "advanced wafer fabrication equipment to produce InP devices at scale" — no crystal growers, no LEC/VGF named.
• Coherent press release "World's First 6-inch InP Scalable Wafer Fabs" (2024) — "wafer fabrication," "device manufacturing" language; zero LEC/VGF references.
• Yole / Compound Semiconductor "$24B CS substrate market by 2027" — top-5 InP wafer suppliers (~70% revenue 2024): Sumitomo, AXT, Freiberger, JX Nippon Mining, VPEC. COHR explicitly NOT in the substrate-supplier list.
• Oxford Instruments press release (2024): delivered automated 6-inch InP wafer processing tools to Coherent — etch equipment, not crystal-growing equipment.
• 36kr "Indium Phosphide Takes the Spotlight" (Apr 2026): Sumitomo ~60% / 4-inch dominant / VB-method; AXT ~35% / 6-inch in mass production / VGF-method; AXT "significant cost advantages" + lower etch-pit density.
• Mordor Intelligence InP wafer market 2026 update: top-5 = ~70% revenue.
• Sumitomo Electric FY2024 results & FY2025 forecasts (2025-05).
• Coherent press release "PMT for Expansion of World's First 150mm Indium Phosphide Mfg Line" (Dec 2024).
• Compound Semiconductor "World's first 6-inch InP scalable wafer fabs" (2024).
• Wikipedia Coherent Corp (acquisition history: Finisar 2019, EpiWorks 2016, Integrated Photonics 2017, Photop 2010, Anadigics 2016 — all epi/device, none crystal-growing).

Shift: Three open questions from yesterday's structural-verifications entry now resolved.

(1) COHR Q3 FY2026 timing — DEFINITIVE 5/6, not 5/13. The original earnings calendar entry (5/13) was wrong; the Q3 print landed today 4:30 PM ET. Confirmation closes the date-discrepancy flag in yesterday's entry. The Q3 transcript color we logged (scale-out CPO 2H 2026, scale-up 2H 2027, Sherman TX named as NVDA-partnership capacity node) is verified. New color from the Q3 print to fold in: (a) backlog visibility extends to calendar 2028 — multi-year cycle thesis hardens beyond the previous "mid-2028 ramp" date; (b) stock declined post-beat despite $1.81B record revenue + Q4 guide $1.91B–$2.05B + NVDA $2B confirmed — consensus was already pricing it in. Worth tracking as a recurring optical-supercycle dynamic: the cycle is becoming consensus, so beats stop generating upside surprise. Position-sizing implication: Wave 1 names trade like late-cycle in their own narrative.

(2) COHR Sherman TX captive-vs-merchant — MERCHANT-BUY, with high confidence. Five convergent evidence lines: (a) CHIPS Act PMT language describes "wafer fabrication equipment," not crystal growers; (b) every COHR press release on Sherman TX uses "wafer fabrication" / "device manufacturing" language with zero LEC/VGF references; (c) II-VI's pre-merger acquisitions (Finisar/EpiWorks/Integrated Photonics/Photop/Anadigics) brought epitaxial growth + device fab capability but never InP crystal-growing capability; (d) Yole's top-5 InP wafer-substrate suppliers explicitly EXCLUDE COHR (top-5 = Sumitomo, AXT, Freiberger, JX Nippon Mining, VPEC, ~70% of 2024 revenue); (e) Oxford Instruments delivered automated 6-inch InP wafer processing tools (etch is downstream of substrate, this is device fab equipment).

(3) Sumitomo Electric incumbent-share — bear case partially INVERTS. Per 36kr Apr-2026 framing: Sumitomo is the 4-inch incumbent leader (~60% / VB-method / "mature, stable yield"); AXT is the 6-inch first-mover (~35% / VGF-method / "significant cost advantages" + lower etch-pit density). The Lumentum Greensboro mid-2028 ramp is 6-inch territory — exactly where AXT has product, NOT where Sumitomo leads. Sumitomo's 6-inch capacity expansion runs through 2026/2027 (catch-up timeline). Mordor's "Sumitomo ~30%" share figure is total InP merchant including both 4-inch and 6-inch; 36kr's "Sumitomo ~60% / AXT ~35%" is volume-weighted across diameters; both can be true. The risk vector for AXTI shifts:

• PRIOR FRAMING: "Sumitomo eats Greensboro because incumbent + 6-inch sooner"
• NEW FRAMING: "Sumitomo trails AXT on 6-inch by ~12-18 months; AXTI's primary risks are now (i) MOFCOM export-permit cadence for AXT Tongmei subsidiary (~60-day permit gate, first issued June 2025), and (ii) Sumitomo's 6-inch catch-up by 2027 (capacity expansion completion)."

Effect on thesis: AXTI promotion to direct-beneficiary anchor is now triple-confirmed durable — LITE merchant-buy ✓ + COHR merchant-buy ✓ + AXT 6-inch first-mover position ✓. The AXTI thesis now rests on three independent legs: (1) the substrate-layer-below-device-fab picks-and-shovels framing; (2) both top-2 US optical device makers as merchant customers; (3) AXT's 6-inch lead vs Sumitomo's 4-inch dominance, exactly aligned with the 2028 Greensboro ramp diameter. The risk vector resolves cleanly to MOFCOM permit cadence as the primary single point of failure — China export-control dynamics are the real binary here, not Sumitomo capture.

Per-ticker:

• COHR ↑↑↑ UPGRADED from ↑↑ (Q3 print confirms structural symmetry + backlog visibility to calendar 2028 hardens multi-year cycle thesis; stock-decline-on-beat is informational, not bearish — late-cycle consensus pricing is normal at this stage of a supercycle).
• LITE = (no change — already ↑↑↑; reinforced by COHR Q3 confirming the parallel-deal narrative).
• AXTI ↑↑↑ (durability triple-confirmed; primary risk now isolated to MOFCOM permit cadence, not Sumitomo).
• SMTOY ↑ (Sumitomo Electric ADR — newly-watchable; ~$7M dollar-volume/day, OTC Pink, marginally tradable for small positions; tracks the 4-inch InP incumbent leg of the merchant supply chain).

Open questions:

• Post-beat-decline pattern across optical-supercycle Wave 1 names — does the same dynamic show up at LITE Q3 earnings 2026-05-22, or is it COHR-specific (e.g., margin-mix concerns)?
• MOFCOM export-permit cadence for AXT Tongmei — frequency of permit issuance, average lag, any backlog of pending applications. This is now AXTI's primary single-point-of-failure risk.
• Sumitomo 6-inch production volume today (capacity expansion target stated, no shipped-volume number found in primary sources) — would clarify the catch-up timeline more precisely.
• AXT 10-K customer concentration table — Q3 2025 disclosed top-5 at 45.2% concentration with two >10% customers, names redacted in the corporate deck. Worth a direct read of the unredacted SEC filing text — concentration footnotes are typically not redacted in actual EDGAR filings.
• COHR 10-K segment capex breakdown by equipment type — would confirm by line-item that Sherman TX capex is processing/fab tools, not crystal-growing equipment. Closes the Sherman-TX-merchant question by primary-document evidence.

Methodology note: This is a clean example of the picks-and-shovels paradigm advantage — by tracing one layer below the device fab, we ended up at a SMALLER risk surface (MOFCOM single-point-of-failure) than the device fab itself faces (consensus-pricing late-cycle dynamics + competition from captive ramps + customer-mix concentration). The substrate layer is more defensible than the device layer because: (a) it's harder to vertically integrate (15× capex differential = $18M Greensboro device-fab shell vs $220-250M substrate greenfield); (b) supply concentration is structural (top-5 = 70%), not narrative; (c) China export-control dynamics actually protect AXTI's geographic positioning vs the alternative of pure-Asian merchant supply — the MOFCOM gate that looks like a risk is also the moat.