Home/Reports/Light / Photonics Value Chain — T-Glass Method Applied

Light / Photonics Value Chain — T-Glass Method Applied

2026-05-13 19:16 · 13.2 KB

Date: 2026-05-13

Type: Explore — value chain tracing on new Light/Photonics theme

Method: T-Glass method (trace from global leader backwards, find Japan monopolies/dominant positions)

Result: 7 companies covering 5 photonics value-chain layers; 1 MONOPOLY position, 2 DOMINANT positions, 4 STRONG positions


Value Chain Map — Light Theme Has 5 Core Layers

LayerWhatJapan's PositionMoat
1. Photon DetectionPMT / SiPM / SPAD / sCMOSHamamatsu 40% global PMT, 100% Western 20-inch (Hamamatsu IR)★★★★ (★★★★★ in 20-inch sub-layer)
2. EUV Mask InspectionActinic patterned mask defect inspectionLasertec ~100% global (senrigan.tech)★★★★★ MONOPOLY
3. Light SourcesUV / EUV-metrology / Xenon cinemaUshio specialty position (NOT EUV litho — Cymer/Gigaphoton own that)★★★ STRONG
4. Optical Fiber & SplicersHyperscale data center fiber + fusion splicers + InP wafersFujikura #1 fusion splicer, Sumitomo Electric 60% InP, Furukawa Lightera 13,824-count (Furukawa PR Mar 2026)★★★ (★★★★ for InP)
5. Spectroscopy / Process OpticsRaman + mass flow controllers for semi processHORIBA world leader in Raman + mass flow★★★★ DOMINANT

7 Companies in Coverage

MONOPOLY Position (Must Hold)

TickerCompanyMkt CapLayerWhy
6920.TLasertec Corporation¥3.90TEUV mask inspection~100% global market share in EUV mask defect inspection. TSMC, Intel, Samsung named customers per senrigan.tech. Without Lasertec, ASML EUV scanners ship "blind".

DOMINANT Positions (Strong Add)

TickerCompanyMkt CapLayerWhy
6965.THamamatsu Photonics¥627BPhoton detectionSole Western 20-inch PMT supplier; 20,000 R12860 units for Hyper-Kamiokande per Wikipedia. 40% global PMT share. Cheapest in theme on 3M return (+29% vs Fujikura +713%).
6856.THORIBA¥1.01TSpectroscopy / processWorld leader Raman + mass flow controllers for advanced semi fabs per HORIBA IR. Cheapest valuation (Fwd PE 23.7) in theme.

STRONG Positions (Good Add)

TickerCompanyMkt CapLayerWhy
5803.TFujikura¥12.6TOptical fiber + fusion splicers#1 globally in fusion splicers via Mordor Intelligence. Stock +713% 52w — primary AI hyperscale fiber rally beneficiary.
5801.TFurukawa Electric¥3.06TOptical fiber + HTS REBCOLightera 13,824-count fiber mass production Mar 2026 — "world's highest fiber density" per Furukawa PR. SuperPower HTS for fusion (Tokamak Energy). Stock +949% 52w; NI +413% FY25.
5802.TSumitomo Electric¥8.74TOptical fiber + InP compound semi60% global InP share — substrate for laser sources in CPO transceivers per Sumitomo Electric. Top-3 optical fiber globally. Cheapest cross-theme bridge to CPO.

MEDIUM Positions (Selective Add)

TickerCompanyMkt CapLayerWhy
6925.TUSHIO INC.¥279BUV / Cinema / EUV metrologySpecialty light source maker. NOT a direct EUV litho supplier (Cymer/ASML own that). Fwd PE 29.6 vs TTM 59.6 = large guidance recovery priced in; need to verify on next earnings.

What's NOT in the Light Theme (Important Negatives)

ExcludedWhy
Topcon (7732.T)DELISTED 2025-12-02 — KKR completed MBO tender offer at ¥3,300/share per KKR/Topcon PR.
Sony Semi SPAD / IMX479Not Japan-listed pure-play (Sony 6758 is conglomerate). But flag as competitive threat to Hamamatsu MPPC — stacked dToF SPAD released June 2025 per Sony press release.
8035 Tokyo Electron + 6857 AdvantestAlready in semi/AI themes — covered there. Light theme references them via cross-theme supply chain links (Lasertec — ASML — TEL — TSMC).

Cross-Theme Linkages

The Light theme is the infrastructure layer beneath Semiconductor and AI/aigpu. Explicit supply chain references in light_companies.json:

Light companyLinks toHow
Lasertec (6920)semi, ai, aigpuTSMC/Intel/Samsung EUV mask inspection enables every advanced AI chip
Fujikura (5803)aigpu, aiHyperscale AI cluster fiber + 400G/800G transceiver fusion splicing
Furukawa Electric (5801)aigpu, ai, nuclearLightera 13,824-count fiber for AI data centers; SuperPower HTS for Tokamak fusion
Sumitomo Electric (5802)ev, aigpu60% InP share (CPO laser substrate); EV wire harness (existing EV theme)
Hamamatsu (6965)robotics, aiLiDAR SiPM for autonomous ADAS; scientific cameras for quantum/medical
HORIBA (6856)semiconductorMass flow controllers at every advanced fab; Raman for 2D semi research
USHIO (6925)semiconductorEUV metrology light sources (NOT EUV litho)

Japan's 3 TRUE Light/Photonics Chokepoints

1. Lasertec Corporation (6920.T) — The EUV Inspection Monopoly

Lasertec's ACTIS family is the only commercial system for actinic patterned EUV mask defect inspection — using the same 13.5 nm wavelength as ASML's EUV scanners. Named customers: TSMC, Intel, Samsung — the only three companies running EUV at scale globally. As Intel 18A, Samsung SF2, and TSMC N2 ramp through 2026-2027, every new EUV mask must pass through a Lasertec system. The de facto pair to ASML — without Lasertec inspection, ASML EUV scanners ship "blind". Stock +257% 52w with PE 44× = priced for monopoly. Reference: senrigan.tech 6920 analysis, Lasertec EUV IR page, ACTIS A300 release.

2. Hamamatsu Photonics (6965.T) — The 20-Inch PMT Monopoly

Hamamatsu's R12860 50-cm PMT is the only Western-aligned large-aperture PMT at scale. Hyper-Kamiokande inner detector needs 20,000 units (with main cavern excavation completed 2025-07-31), data taking starts 2028 per Wikipedia. NNVT (China) makes MCP-PMT alternatives but only Chinese projects (JUNO 75% NNVT share) select it. The bottleneck is the 12-stage glass-vacuum-dynode-photocathode process at Toyooka — not replicable in <5 years. Reference: Hamamatsu 20-inch PMT page.

3. Sumitomo Electric (5802.T) — The InP Compound Semi Choke

Sumitomo Electric controls ~60% of the global indium phosphide (InP) wafer market — InP is the substrate material for laser sources in silicon-photonics transceivers and Co-Packaged Optics (CPO). NVIDIA and Broadcom started CPO product shipments in 2026 (per DigiTimes) — every CPO transceiver consumes InP. Reference: Sumitomo Electric fiber optic interconnect.


What Japan LACKS in Light/Photonics

GapWhyImpact
EUV lithography light sourceCymer (ASML-owned, San Diego) + Gigaphoton (Komatsu) duopoly; Ushio exited LPP-litho in 2013No domestic pure-play; cross-listed exposure via Komatsu
Silicon photonics PIC for CPOTSMC COUPE, GlobalFoundries (post-AMF acq), Coherent dominateLight theme is photodetector-adjacent only; concept-stock risk for any "Japan CPO" thesis
Stacked CMOS SPAD (LiDAR)Sony Semi IMX459/IMX479 leads consumer + emerging autoHamamatsu MPPC moat being challenged from CMOS direction (June 2025 IMX479 release)
High-volume PET-TOF SiPMSiemens (Biograph Vision) and Philips (Vereos) vertically insourced their SiPM detectors per JNM 2019Hamamatsu shut out of 2 of top 4 PET OEMs

Pricing Inefficiency (T-Glass Observation)

The market has bid up optical fiber and EUV inspection hardest. The pricing dispersion is the inefficiency:

Ticker52w returnTTM PEFwd PE (company guide)Notes
Furukawa Electric (5801)+949%65.765.8Hit ストップ高 limit-up 2026-05-12 +16% on FY26 beat. 10-for-1 split announced.
Fujikura (5803)+713%87.784.26-for-1 split effective 2026-04-01. Consensus Fwd PE 59x is 43% above company guidance EPS JPY 90.60 — gap is the rally fuel.
Sumitomo Electric (5802)+377%34.023.9Cheaper InP optionality
Lasertec (6920)+257%44.141.1Priced monopoly, watch peak-earnings
HORIBA (6856)+134%27.523.7Cheapest semi-process play
Ushio (6925)+98%59.629.6Mixed business — laggard
Hamamatsu (6965)+29% (3M only)50.442.5LAGGARD — concept-stock risk on CPO; real thesis is PMT/Hyper-K

Errata note (2026-05-13): initial DD pass quoted Fujikura Fwd PE 59x and Furukawa TTM 58 / Fwd 49 from StockAnalysis. Cross-validation against Yahoo Finance JP + manual computation revealed both names had recent stock splits (Fujikura 6:1 Apr 2026, Furukawa 10:1 pending) and StockAnalysis had not propagated company-guidance EPS revisions. The corrected company-guidance numbers are above. Per data quality rule in analysis_workflow.md:252-257, this is a known yfinance/StockAnalysis failure mode after splits. User-prompted catch.

Trade construct: trim Furukawa + Fujikura (peak-earnings risk on +713%/+949% rally, both still PE 65-85x on company guidance); add Hamamatsu + HORIBA + Sumitomo Electric (cheaper exposure to the same secular theme). Hold Lasertec (monopoly priced but moat real).


Quality Comparison vs Nuclear Theme

This Light DD aims to match the bar set by nuclear_value_chain_tglass_method_2026-04-21.md and japan_nuclear_technology_opus_dd_2026-04-18.md. Comparison:

MetricNuclearLight
Companies covered22 (4 tiers)7 (3 conviction tiers)
Value chain layers65
Monopoly positions named4 (JSW, Toyo Tanso, Okano, Stella Pharma)1 (Lasertec)
Cross-theme tickersYes (5801 in fusion)Yes (5801, 5802, 5803, 6920, 6856 cross to AI/semi/nuclear)
Source URLs per supply chain entryYesYes
Specific bottleneck identificationYes (Muroran forging press)Yes (Toyooka PMT, ACTIS, Lightera)
Macro context tied to global driversYes (US-Japan $40B, BWRX-300)Yes (NVIDIA/Broadcom CPO, Hyper-K, ASML EUV)

Gap to close in future passes:

1. Add Tier 4 unlisted reference companies (e.g., NTT Innovative Devices for InP, Sumitomo Osaka Cement for optical materials).

2. Build per-stock deep DD files for Lasertec (6920), Fujikura (5803), Furukawa Electric (5801) — currently only Hamamatsu has a dedicated DD.

3. Errata table tracking any corrections vs initial entries.

4. Recent updates log entries on each subsequent evolve / explore round.


Sources