cover
Lasers, photodiodes, modulators, epiwafers and substrates for AI data-centre optics: EML, CW and UHP lasers, DWDM sources, VCSELs, co-packaged optics and optical I/O.
- Published: October 2026
- Pages: 198
- Tables: 53
- Figures: 35
Indium phosphide (InP) and gallium arsenide (GaAs) have moved from specialist telecom materials to strategic infrastructure for artificial intelligence. Silicon cannot emit light efficiently, so every optical link in an AI cluster depends on a compound-semiconductor laser. That holds even when the rest of the link is built on silicon photonics. As each new generation of GPU and accelerator doubles its I/O bandwidth, and copper runs out of reach at 200 Gb/s per lane, demand for InP lasers, photodiodes and modulators has outpaced supply at every level of the chain. Recent developments show how acute the bottleneck has become:
- Shortages: through 2025 and 2026, leading suppliers reported EML laser under-shipment of more than 30% and capacity sold out into 2027–2028.
- Substrate supply: InP substrate prices rose repeatedly after China introduced export licensing for InP and indium compounds in February 2025.
- Customer funding: NVIDIA took strategic stakes in Lumentum and Coherent, and hyperscalers are now prepaying for laser capacity.
- Substrate expansion: AXT, Sumitomo Electric and JX Advanced Metals have announced multi-fold capacity increases.
- Wafer-fab expansion: Coherent, Lumentum, Nokia and an EU-backed consortium in Eindhoven are scaling 6-inch InP fabs. The move from 3-inch and 4-inch wafers has become the industry’s central cost lever.
- Chinese suppliers: Yuanjie, Everbright, Vertilite and Source Photonics are investing heavily to localise high-speed laser chips.
The product mix is also changing quickly. 200G-per-lane EMLs entered volume production, and 400G-per-lane devices were demonstrated at OFC 2026. The fastest growth is shifting to high-power and ultra-high-power CW lasers that feed silicon photonics and co-packaged optics (CPO) through external laser sources. Multi-wavelength DWDM sources for optical I/O in scale-up networks are growing fast as well. In GaAs, 200G VCSELs and dense VCSEL arrays for “slow-and-wide” in-rack links are opening new opportunities. Quantum-dot comb lasers on GaAs are emerging as an alternative light source. Venture investment in CPO, optical I/O and multi-wavelength laser start-ups has reached record levels. Looking ahead, supply is expected to reach approximate balance around 2028–2029, after which competition will shift from capacity to cost. Several factors will decide the winners:
- 6-inch yields: how quickly substrate makers and device fabs master 6-inch production.
- Laser reliability: how reliably ultra-high-power lasers perform inside switches and accelerator packages.
- Integration: how far heterogeneous and transfer-printed lasers are integrated with silicon photonics.
- Competing modulators: thin-film lithium niobate and other new modulators will convert demand for modulated lasers into demand for CW lasers.
- Geopolitics: export controls on gallium, indium and InP will remain a permanent design constraint.
As AI scale-up domains move to optics between 2027 and 2030, InP’s role will grow. Value will move from pluggable transceivers into the rack, the package and, ultimately, the accelerator itself.
InP and GaAs Photonics for AI: Global Market 2027–2037 is a comprehensive analysis of the compound-semiconductor devices, epitaxial wafers and substrates that power optical interconnect in AI data centres. It covers the market from substrate to bare die. It quantifies the lasers, photodiodes, modulators and amplifiers that sit inside pluggable transceivers, linear-drive optics, co-packaged optics and optical I/O, and the InP and GaAs wafers they are built on. The AI build-out has made optical components one of the tightest bottlenecks in computing infrastructure. Every silicon-photonics engine depends on an InP light source. Supply of high-speed EMLs, high-power CW lasers and device-grade InP substrates has fallen well short of demand. The report explains why: crystal-growth yields, the move from 3-inch to 6-inch wafers, MOCVD and test capacity, and Chinese export controls on indium and gallium compounds. It also sets out how the industry is responding, through fab acquisitions, multi-fold substrate expansions, customer prepayments and NVIDIA’s direct investment in laser suppliers.
Forecasts run from 2021 to 2037 and are segmented as follows:
- Device technology: EML, DML, standard CW and ultra-high-power CW lasers, DWDM and multi-wavelength sources, tunable lasers, InP IQ modulators and coherent PICs, SOAs, InP and GaAs photodiodes, VCSELs and quantum-dot sources.
- Material: InP and GaAs.
- Network architecture: scale-out, scale-up, and scale-across/DCI.
- Value-chain level: bare die, captive and open-market epiwafers, and substrates by wafer diameter.
- Region: region of manufacture.
Units, average selling prices and revenues are given for each segment, along with base, upside and downside scenarios. The report assesses technology roadmaps from 100G to 400G per lane and beyond. It covers integration routes with silicon photonics (external laser sources, hybrid flip-chip, heterogeneous bonding, micro-transfer printing and monolithic quantum-dot growth) and the competitive threat from thin-film lithium niobate, ferroelectric and microLED alternatives. Supply-chain chapters map substrate, epitaxy, device and foundry capacity, compare net sellable capacity with demand, and track lead times, pricing, investment, M&A, government programmes and venture funding.
The report closes with 65 detailed company profiles spanning substrate suppliers, epiwafer houses, equipment makers, device manufacturers, foundries, integration-platform developers, start-ups and the system companies driving demand. A companion Excel workbook provides the full forecast data with an interactive dashboard.
Report contents include:
- Executive summary with key findings, scenario summary and recent developments for 2025–2026
- Scope, methodology, and material properties of InP, GaAs, silicon and GaN
- AI demand drivers: scale-up, scale-out and scale-across networks, reach versus technology, the lane-rate roadmap, and laser content per link
- InP device technologies: EML, DML, CW and UHP CW lasers, DWDM sources, tunable lasers, IQ modulators, SOAs, photodiodes and InP PICs
- GaAs device technologies: 850 nm VCSELs, VCSEL arrays for scale-up, GaAs photodiodes and quantum-dot lasers
- Integration with silicon photonics and competing modulator platforms
- Substrate, wafer-diameter and epitaxy technology, including VGF/VB growth, the 6-inch transition and MOCVD
- Market forecasts for 2021–2037 by device, material, architecture, epiwafer, substrate and region, with scenarios
- Supply chain and capacity: suppliers, capacity versus demand, lead times, pricing and geographic footprint
- Investment, M&A, government programmes and start-up venture funding
- China, critical minerals and export controls
- Competitive landscape and market shares
- Technology roadmap, opportunity assessment and outlook
- 65 company profiles including Accelink Technologies, Aixtron, ams OSRAM, Applied Optoelectronics, Avicena, AXT Inc., Ayar Labs, Broadcom, Coherent Corp., DOWA Electronics Materials, EFFECT Photonics, Enlightra, Eoptolink, Everbright Photonics, Freiberger Compound Materials, Furukawa FITEL Optical Components (formerly Fujitsu Optical Components), GCS (Global Communication Semiconductors), HiSilicon, HyperLight, Innolume, Intel, IntelliEPI, iPronics, IQE, JX Advanced Metals, Korea Zinc, LandMark Optoelectronics, Lightmatter, Lumentum, Lumilens and more.....Each company profile includes a key-facts panel (founding, headquarters, ownership, leadership, revenue or funding, and AI photonics products), a company description, products and technology, dated recent developments for 2025–2026, a financial, funding or product table where available, a Future Markets view of the company’s strategic position, competitive context and key partners, and the company website.
Purchasers will receive the following:
- PDF report download/by email.
- Comprehensive Excel spreadsheet of all data including data on all investments & deals segmented by technology, sub-sector, government, region, and company
- Mid-year Update
1 EXECUTIVE SUMMARY 14
- 1.1 Key Findings at a Glance 14
- 1.2 Market Size and Growth, 2027–2037 14
- 1.3 Where the Value Goes: Devices and Architectures 15
- 1.4 Supply Constraints in Brief 16
- 1.5 Recent Developments, 2025–2026 17
- 1.6 Scenario Summary 17
- 1.7 Strategic Imperatives 18
2 INTRODUCTION, SCOPE AND METHODOLOGY 19
- 2.1 Report Scope 19
- 2.2 Methodology 20
- 2.3 Comparison With Other Future Markets Forecasts 20
- 2.4 Why Compound Semiconductors? 20
- 2.5 Device Technologies Covered 20
3 AI INFRASTRUCTURE DEMAND DRIVERS 22
- 3.1 Bandwidth per Accelerator and the Interconnect Wall 22
- 3.2 Network Tiers: Scale-Up, Scale-Out and Scale-Across 22
- 3.3 Reach Decides the Technology 23
- 3.4 Lane-Rate Roadmap 23
- 3.5 Laser Content per Link: Pluggables, LPO, CPO and Optical I/O 24
- 3.6 Drivers and Restraints 25
4 INP DEVICE TECHNOLOGIES 26
- 4.1 The InP Photonics Platform 26
- 4.2 Electro-Absorption Modulated Lasers (EML) 26
- 4.3 Directly Modulated Lasers (DML) 26
- 4.4 CW-DFB Lasers for Silicon Photonics 26
- 4.5 High-Power and Ultra-High-Power CW Lasers for NPO and CPO 27
- 4.6 DWDM Laser Arrays and Multi-Wavelength Sources for Optical I/O 27
- 4.7 Coherent Components: Tunable Lasers, IQ Modulators and Receivers 28
- 4.8 Semiconductor Optical Amplifiers 28
- 4.9 InP Photodiodes 28
- 4.10 Monolithic InP PICs 28
- 4.11 InP Device Benchmark 29
5 GAAS DEVICE TECHNOLOGIES 30
- 5.1 The GaAs Photonics Platform 30
- 5.2 850 nm VCSELs for Short-Reach AI Links 30
- 5.3 VCSEL Arrays for Slow-and-Wide Scale-Up and NPO 31
- 5.4 GaAs Photodiodes 31
- 5.5 Quantum-Dot Lasers on GaAs 31
- 5.6 GaAs vs InP vs Silicon Photonics for Short Reach 31
6 INTEGRATION WITH SILICON PHOTONICS AND COMPETING PLATFORMS 32
- 6.1 Why Integration Matters for InP Demand 32
- 6.2 Hybrid Flip-Chip Integration 32
- 6.3 Heterogeneous III-V-on-Silicon 32
- 6.4 Micro-Transfer Printing and Monolithic Growth 32
- 6.5 Competing Modulator Platforms 33
7 SUBSTRATE, WAFER AND EPITAXY TECHNOLOGY 34
- 7.1 InP Crystal Growth: VGF, VB and LEC 34
- 7.2 Wafer Diameter Roadmap 34
- 7.3 Epitaxy: MOCVD and MBE 36
- 7.4 Device Fab and Test Bottlenecks 36
8 MARKET FORECASTS 2021–2037 37
- 8.1 Forecast Assumptions 37
- 8.2 Total Bare-Die Market by Device Technology 37
- 8.3 Bare-Die Volumes 38
- 8.4 Average Selling Prices 39
- 8.5 Market by Network Architecture 40
- 8.6 InP vs GaAs 41
- 8.7 Epiwafer Market 41
- 8.8 Substrate Market 42
- 8.9 Total Value Chain 45
- 8.10 Market by Region of Manufacture 45
- 8.11 Scenario Analysis 47
9 SUPPLY CHAIN AND CAPACITY 48
- 9.1 InP Laser Supply Chain 48
- 9.2 GaAs Laser Supply Chain 48
- 9.3 Substrate Suppliers 48
- 9.4 Epiwafer Suppliers 49
- 9.5 Device Makers and Foundries 49
- 9.6 Capacity vs Demand: InP Substrates 50
- 9.7 Capacity vs Demand: InP Laser Die 51
- 9.8 Lead Times and Pricing Indicators 51
- 9.9 Geographic Footprint 51
10 INVESTMENT, M&A AND STRATEGIC PARTNERSHIPS 53
- 10.1 NVIDIA and Hyperscaler Investments in Laser Supply 53
- 10.2 Fab and Capacity Investments 53
- 10.3 Venture Funding for Photonics Start-ups 55
- 10.4 Government Support Programmes 56
- 10.5 Mergers and Acquisitions 56
11 CHINA, CRITICAL MINERALS AND EXPORT CONTROLS 58
- 11.1 Indium and Gallium Supply 58
- 11.2 Export Control Timeline 58
- 11.3 Impact on the Substrate Market 59
- 11.4 China’s Domestic Photonics Chip Industry 59
- 11.5 Geopolitical Scenarios 59
12 COMPETITIVE LANDSCAPE 61
- 12.1 Bare-Die Market Shares 61
- 12.2 InP Substrate Market Shares 61
- 12.3 Strategic Positioning 62
13 TECHNOLOGY ROADMAP AND OUTLOOK 63
- 13.1 Technology Roadmap 2025–2037 63
- 13.2 Opportunity Assessment 64
- 13.3 Outlook and Conclusions 65
14 COMPANY PROFILES 66 (65 company profiles)
15 APPENDIX 197
- 15.1 Annual Forecast Data, 2021–2037 197
- 15.2 Glossary 198
- 15.3 Principal Public Sources 199
LIST OF TABLES
- Table 1. InP and GaAs photonics for AI: key market metrics, 2025–2037 (US$ million) 15
- Table 2. Selected developments in InP and GaAs photonics for AI, 2025–2026 17
- Table 3. Bare-die market scenarios, 2027–2037 (US$ million) 17
- Table 4. Definition of value-chain levels used in this report 19
- Table 5. Key properties of photonic semiconductor platforms 20
- Table 6. InP and GaAs device technologies covered in this report 20
- Table 7. AI network tiers and their compound-semiconductor content 22
- Table 8. Lane-rate generations and compound-semiconductor implications 24
- Table 9. Laser configuration by optical architecture 25
- Table 10. Market drivers and restraints, 2027–2037 25
- Table 11. EML generations for AI optics 26
- Table 12. CW laser power classes and applications 27
- Table 13. Multi-wavelength source approaches for optical I/O 27
- Table 14. Benchmark of InP device categories for AI optics (2026) 29
- Table 15. VCSEL generations for data-centre optics 30
- Table 16. Short-reach technology comparison (≤100 m) 31
- Table 17. Laser integration approaches compared 33
- Table 18. Modulator platforms and impact on InP content 33
- Table 19. InP and GaAs substrate growth methods 34
- Table 20. Wafer diameter transitions for InP and GaAs photonics 35
- Table 21. Epitaxy and process equipment for InP and GaAs photonics 36
- Table 22. Bare-die market by device technology, 2025–2037 (US$ million) 37
- Table 23. Bare-die shipments by device technology, 2025–2037 (million units) 38
- Table 24. Average selling price by device technology, 2025–2037 (US$ per die) 39
- Table 25. Bare-die market by network architecture, 2025–2037 (US$ million) 41
- Table 26. Bare-die market by substrate material, 2025–2037 (US$ million) 41
- Table 27. Epiwafer volume and open-market value, 2025–2037 42
- Table 28. Substrate demand and revenue from AI photonics, 2025–2037 43
- Table 29. Bare-die revenue by region of manufacture, 2025–2037 (US$ million) 46
- Table 30. Bare-die market scenarios, 2025–2037 (US$ million) 47
- Table 31. InP substrate suppliers: capacity and expansion 48
- Table 32. GaAs substrate suppliers relevant to AI photonics 48
- Table 33. Open-market epiwafer suppliers for InP and GaAs photonics 49
- Table 34. InP and GaAs photonic device makers for AI 49
- Table 35. InP supply-chain lead times and pricing indicators, 2026 51
- Table 36. Geographic footprint of InP and GaAs photonics manufacturing 51
- Table 37. Strategic investments by GPU vendors and hyperscalers in photonics supply, 2025–2026 53
- Table 38. Device fab expansions for AI photonics, 2024–2026 54
- Table 39. Substrate capacity investments, 2025–2026 54
- Table 40. Venture funding rounds for photonics start-ups relevant to InP/GaAs photonics for AI, Q2 2025 – Oct 2026 55
- Table 41. Government programmes supporting InP/GaAs photonics 56
- Table 42. Selected M&A in compound-semiconductor photonics, 2022–2026 56
- Table 43. Chinese export controls relevant to InP and GaAs photonics 59
- Table 44. Selected Chinese InP and GaAs photonic chip suppliers 59
- Table 45. Estimated bare-die market shares by supplier, 2025 61
- Table 46. Strategic positioning of leading players 62
- Table 47. Technology roadmap milestones 63
- Table 48. Opportunity and risk assessment by segment 64
- Table 49. Bare-die revenue by device technology, 2021–2029 (US$ million) 197
- Table 50. Bare-die revenue by device technology, 2030–2037 (US$ million) 197
- Table 51. Bare-die shipments by device technology, 2021–2029 (million units) 198
- Table 52. Bare-die shipments by device technology, 2030–2037 (million units) 198
- Table 53. Glossary of terms 198
LIST OF FIGURES
- Figure 1. InP and GaAs photonics for AI: bare-die market, 2021–2037 (US$ million) 15
- Figure 2. Bare-die revenue by device technology, 2027 vs 2037 (US$ million) 16
- Figure 3. Share of bare-die revenue by network architecture, 2025–2037 16
- Figure 4. InP and GaAs photonics value chain and scope of this report 19
- Figure 5. Optical modules in AI scale-up and scale-out networks by data rate, 2026–2036 (million units) 22
- Figure 6. Optical technologies by reach in AI data centres 23
- Figure 7. Lane-rate and laser-architecture roadmap for AI optics, 2024–2038 24
- Figure 8. Typical InP/GaAs laser die per 1.6 Tb/s of link capacity by architecture 25
- Figure 9. Typical maximum output power of CW lasers for AI optics, 2022–2030 27
- Figure 10. Maximum multimode (OM4) reach of 850 nm VCSEL links by lane rate 30
- Figure 11. Laser integration approaches for silicon photonics: maturity timeline 33
- Figure 12. Usable wafer area by diameter relative to 3-inch (5 mm edge exclusion) 35
- Figure 13. Diameter mix of InP wafer starts for AI photonic devices (share of 4-inch-equivalent area), 2021–2037 35
- Figure 14. Bare-die market by device technology, 2021–2037 (US$ million) 37
- Figure 15. Bare-die shipments for selected device technologies, 2025–2037 (million units, log scale) 39
- Figure 16. ASP index for selected device technologies, 2025–2037 (2025 = 100) 40
- Figure 17. Bare-die market by network architecture, 2021–2037 (US$ million) 40
- Figure 18. InP share of bare-die revenue, 2021–2037 41
- Figure 19. InP epiwafer starts for AI devices: captive vs open market, 2021–2037 (thousand 4-inch-equivalent wafers) 42
- Figure 20. InP substrate demand from AI photonic devices by diameter, 2021–2037 (thousand 4-inch-equivalent wafers) 43
- Figure 21. InP substrate revenue: AI photonics vs other applications, 2021–2037 (US$ million) 43
- Figure 22. Blended InP substrate ASP, 2021–2037 (US$ per 4-inch-equivalent wafer) 44
- Figure 23. Total InP and GaAs photonics for AI value chain (bare die + open epiwafers + substrates), 2021–2037 (US$ million) 45
- Figure 24. Bare-die revenue by region of manufacture, 2025, 2031 and 2037 46
- Figure 25. Bare-die market scenarios, 2025–2037 (US$ million) 47
- Figure 26. InP substrates: from nameplate capacity to net sellable supply, 2026 (Future Markets estimate) 50
- Figure 27. Supply as a percentage of demand: InP laser chips and device-grade InP substrates, 2024–2031 51
- Figure 28. Selected announced investments in InP/GaAs photonics capacity, 2024–2026 (US$ million) 54
- Figure 29. Largest venture rounds in optical interconnect start-ups relevant to InP/GaAs photonics, Q2 2025 – Oct 2026 (US$ million) 55
- Figure 30. China share of world indium and gallium supply 58
- Figure 31. Timeline of Chinese export controls affecting InP and GaAs photonics, 2023–2026 58
- Figure 32. Estimated bare-die market shares, 2025 (% of revenue, incl. captive at merchant value) 61
- Figure 33. Estimated InP substrate market shares, 2025 (% of revenue, all applications) 62
- Figure 34. InP and GaAs photonics technology roadmap, 2024–2038 63
- Figure 35. Opportunity map: 2031 market size vs 2027–2037 growth by device technology (bubble size = 2037 revenue)
Purchasers will receive the following:
- PDF report download/by email.
- Comprehensive Excel spreadsheet of all data including data on all investments & deals segmented by technology, sub-sector, government, region, and company
- Mid-year Update
