| 1. | EXECUTIVE SUMMARY |
| 1.1. | What does a modern high-performance AI data center look like? |
| 1.2. | Switches: Key components in a modern data center |
| 1.3. | Advancements in Switch IC Bandwidth and the Need for Co-Packaged Optics (CPO) Technology |
| 1.4. | Overview of key challenges in data center architectures |
| 1.5. | Key trend of optical transceiver in high-end data centers |
| 1.6. | Design decisions for CPO compared to Pluggables |
| 1.7. | What is an Optical Engine (OE) |
| 1.8. | Heterogeneous integration and Co-Packaged Optics (CPO) |
| 1.9. | Overview of interconnection technique in semiconductor packaging |
| 1.10. | Key CPO applications: Network switch and computing optical I/O |
| 1.11. | EIC/PIC integration by advanced interconnect technique |
| 1.12. | 2D to 3D EIC/PIC integration options |
| 1.13. | Benchmark table of different packaging technologies for EIC/PIC |
| 1.14. | Examples of packaging a 3D optical engine with an IC |
| 1.15. | Three types of CPO + XPU/switch ASIC packaging structures |
| 1.16. | Challenges and future potential of CPO technology |
| 1.17. | Supply Chain Overview |
| 1.18. | NVIDIA vs Broadcom: Strategic Comparison in AI Infrastructure and CPO Technologies |
| 1.19. | CPO product benchmark: NVIDIA vs Broadcom |
| 1.20. | NVIDIA and Broadcom: Divergent CPO Ecosystems |
| 1.21. | Current AI system architecture |
| 1.22. | Future AI architecture (short to mid term) predicted by IDTechEx |
| 1.23. | Future AI architecture (long term) predicted by IDTechEx |
| 1.24. | Forecasting the AI accelerator Market: Changes From the previous Edition |
| 1.25. | Forecasts with table: Server boards, CPUs and GPUs/Accelerators |
| 1.26. | Optical I/O for AI interconnect CPO Forecast (units shipped) |
| 1.27. | Optical I/O for AI interconnect CPO Forecast (revenue/market size) |
| 1.28. | CPO network switches (L2 Switches) for AI accelerators forecast (units shipped) |
| 1.29. | CPO network switches (L2 Switches) for AI accelerators forecast (market size and revenue) |
| 1.30. | Total CPO market |
| 1.31. | Total CPO by different EIC/PIC integration technology (unit shipment, millions) |
| 1.32. | System integration of network switches (L2 Switches) for AI accelerators forecast by packaging technologies (unit shipped) |
| 1.33. | System integration of Optical I/O Forecast by packaging technologies (units shipped) |
| 2. | CHALLENGES AND SOLUTIONS FOR FUTURE AI SYSTEM |
| 2.1.1. | The rise and the challenges of LLM |
| 2.1.2. | What does a modern high-performance AI data center look like? |
| 2.1.3. | Closer look into NVIDIA's state-of-the-art AI system |
| 2.1.4. | Switches: Key components in modern data centers |
| 2.2. | Scale-up, Scale-out, and Scale-across network |
| 2.2.1. | Scale-up and Scale-Out |
| 2.2.2. | Overview: Scale-up, Scale-out, and Scale-across |
| 2.3. | Challenges in Network Switches Interconnect for High-end Data Centers |
| 2.3.1. | Roadmap of interconnect technology for network switches in high-end data centers |
| 2.3.2. | Serdes bottleneck in high-bandwidth systems |
| 2.3.3. | Solutions to Serdes bottlenecks in high-bandwidth systems |
| 2.3.4. | Pluggable optics - what are the bottlenecks? |
| 2.3.5. | On-Board Optics (OBO) |
| 2.3.6. | Co-Packaged Optics (CPO) |
| 2.3.7. | Transmission losses in a pluggable optical transceiver connection |
| 2.3.8. | Pluggable optics vs CPO |
| 2.3.9. | Design decisions for CPO compared to Pluggables |
| 2.3.10. | Advancements in switch IC bandwidth and the need for CPO technology |
| 2.3.11. | L2 frontside network architecture diagram CPO versus non-CPO |
| 2.4. | Challenges in Compute Switches Interconnect (i.e. Optical I/O) for High-end Data Centers |
| 2.4.1. | Number of Cu wires in current AI system Interconnects |
| 2.4.2. | Limitations in current copper systems in AI |
| 2.4.3. | Nvidia's connectivity choices: Copper vs optical for high-bandwidth systems |
| 2.4.4. | Copper vs. optical for high-bandwidth systems: Benchmark |
| 2.4.5. | Moving from Cu to optical interconnects for a high-end AI system |
| 2.4.6. | Current AI system architecture |
| 2.4.7. | L1 backside compute architecture with Cu systems |
| 2.4.8. | L1 backside compute architecture with optical interconnect: Co-Packaged Optics (CPO) |
| 2.4.9. | Opportunities for swapping copper interconnects to optical connects - what did the leader say? |
| 2.5. | Future AI system in high-end data center |
| 2.5.1. | Power efficiency comparison: CPO vs pluggable optics vs copper interconnects |
| 2.5.2. | Latency of 60cm data transmission technology benchmark |
| 2.5.3. | Future AI architecture (short to mid term) predicted by IDTechEx |
| 2.5.4. | Future AI architecture (long term) predicted by IDTechEx |
| 3. | INTRODUCTION TO CO-PACKAGED OPTICS (CPO) |
| 3.1.1. | What's covered in this chapter |
| 3.2. | PICs Key Concepts |
| 3.2.1. | What are Photonic Integrated Circuits (PICs)? |
| 3.2.2. | PICs vs Silicon Photonics - what are the differences |
| 3.2.3. | PIC architecture |
| 3.2.4. | Advantages and challenges of PIC |
| 3.3. | Optical Engine (OE) |
| 3.3.1. | What is an optical engine? |
| 3.3.2. | How an optical engine works |
| 3.3.3. | Optical power supplies |
| 3.4. | Co-Packaged Optics |
| 3.4.1. | Three key concepts in co-packaged optics (CPO) |
| 3.4.2. | Key technology building blocks for CPO |
| 3.4.3. | Benefits of CPO: Latency |
| 3.4.4. | Benefits of CPO: Power consumption |
| 3.4.5. | Benefits of CPO: Data rate |
| 3.4.6. | Overview of value proposition of CPO |
| 3.4.7. | Future challenges in CPO |
| 4. | PACKAGING FOR CO-PACKAGED OPTICS (CPO) |
| 4.1.1. | Key components to be packaged in an optical transceiver |
| 4.1.2. | Heterogeneous integration and Co-Packaged Photonics |
| 4.1.3. | CPO for network switch - packaging concept |
| 4.1.4. | Example: 1.6 Tbps Co-packaged optics for network switch |
| 4.1.5. | CPO as optical I/O for XPUs - packaging concept |
| 4.1.6. | CPO as optical I/O for XPUs - packaging concept (follow) |
| 4.1.7. | Example: CPO integration for compute silicon |
| 4.1.8. | Overview of CPO packaging technologies |
| 4.2. | Overview and development roadmap of 2.5D and 3D advanced semiconductor packaging technologies |
| 4.2.1. | Evolution roadmap of semiconductor packaging |
| 4.2.2. | Semiconductor packaging - an overview of technology |
| 4.2.3. | Key metrics for advanced semiconductor packaging performance |
| 4.2.4. | Overview of interconnection technique in semiconductor packaging |
| 4.2.5. | Overview of 2.5D packaging structure |
| 4.3. | 2.5D Si-based Packaging Technologies |
| 4.3.1. | 2.5D packaging that involves Si as interconnect |
| 4.3.2. | Through Si Via (TSV) - now and the future |
| 4.3.3. | Developing trend for 2.5D Si-based packaging |
| 4.3.4. | Si interposer vs Si bridge benchmark |
| 4.4. | 2.5D Organic-based Packaging technologies |
| 4.4.1. | 2.5D packaging - high density fan-out (FO) packaging |
| 4.4.2. | Redistribution Layer (RDL) |
| 4.4.3. | Electronic interconnects: SiO2 vs organic dielectric |
| 4.4.4. | Two types of fan-out: Panel level |
| 4.4.5. | Two types of fan-out: Wafer level |
| 4.4.6. | Wafer level fan-out vs panel level fan-out: The differences |
| 4.4.7. | Key trends in fan-out packaging |
| 4.4.8. | Challenges in future fan-out process |
| 4.5. | 2.5D Glass-based Packaging Technologies |
| 4.5.1. | Roles of glass in semiconductor packaging |
| 4.5.2. | Glass core as interposer for advanced semiconductor packaging |
| 4.5.3. | Overcoming limitations of Si interposers with glass |
| 4.5.4. | Glass vs molding compound |
| 4.5.5. | Glass core (interposer) package - process flow |
| 4.5.6. | Challenges of glass packaging |
| 4.6. | 3D Advanced Semiconductor Packaging Technologies |
| 4.6.1. | Evolution of bumping technologies |
| 4.6.2. | Challenges in scaling bumps |
| 4.6.3. | µ bump for advanced semiconductor packaging |
| 4.6.4. | Bumpless Cu-Cu hybrid bonding |
| 4.6.5. | Three ways of Cu-Cu hybrid bonding: Benchmark |
| 4.6.6. | Challenges in Cu-Cu hybrid bonding manufacturing process |
| 4.7. | CPO Packaging: EIC and PIC Integration |
| 4.7.1. | EIC/PIC integration - by conventional interconnect technique |
| 4.7.2. | EIC/PIC integration by emerging interconnect technique |
| 4.7.3. | 2D to 3D EIC/PIC integration options |
| 4.7.4. | Benchmark table of different packaging technologies for EIC/PIC |
| 4.7.5. | Pros and Cons of 2D integration of EIC/PIC |
| 4.7.6. | Pros and Cons of 2.5D integration of EIC/PIC |
| 4.7.7. | Pros and Cons of 3D hybrid integration of EIC/PIC |
| 4.7.8. | Pros and Cons of 3D monolithic integration of EIC/PIC |
| 4.8. | TSV for EIC/PIC Integration |
| 4.8.1. | TSV for EIC/PIC integration in CPO |
| 4.8.2. | Why using TSV for PIC and EIC integration |
| 4.8.3. | Cisco packaging architectures of optical engine over generations |
| 4.8.4. | Cisco: 2.5D Chip-on-Chip (CoC) packaging architecture for EIC/PIC integration |
| 4.8.5. | Cisco: 3D TSV for PIC/EIC integration |
| 4.8.6. | Key TSV Fabrication steps and challenges in CPO - 1 |
| 4.8.7. | Key TSV fabrication steps and challenges in CPO - 2 |
| 4.8.8. | Packaging options for silicon photonics - w/ or w/o TSV? |
| 4.8.9. | Pros and Cons of 2.5D Si interposer for EIC/PIC integration |
| 4.9. | Fan-out for EIC/PIC Integration |
| 4.9.1. | ASE's proposed fan-out solution for CPO packaging |
| 4.9.2. | FOPOP from ASE - process |
| 4.9.3. | Detailed analysis of FOPOP vs WB packaging for CPO |
| 4.9.4. | Optical packaging process considerations for silicon photonics - ASE |
| 4.9.5. | SPIL's Fan-Out Embedded Bridge (FOEB) structure for PIC/EIC integration in CPO |
| 4.9.6. | Process flow of integrating PIC and EIC in a FOEB structure |
| 4.9.7. | Process challenges in packaging OE |
| 4.9.8. | Rockley Photonics proposes FOWLP for CPO packaging structure |
| 4.9.9. | Rockley Photonics' FOWLP CPO packaging process flow - 1 |
| 4.9.10. | Rockley Photonics' FOWLP CPO packaging process flow - 2 |
| 4.9.11. | Challenges of using fan-out for EIC/PIC integration |
| 4.10. | Glass-based CPO Packaging Technologies |
| 4.10.1. | Glass-based Co-packaged optics - Corning's vision |
| 4.10.2. | Glass-based Co-packaged optics - packaging structure |
| 4.10.3. | Glass-based Co-packaged optics - process development |
| 4.10.4. | Corning's 102.4 Tb/s test vehicle |
| 4.11. | Hybrid bonding for EIC/PIC integration |
| 4.11.1. | TSMC: Integrated HPC technology platform for AI |
| 4.11.2. | Optical engine roadmap from TSMC - 1 |
| 4.11.3. | Optical engine roadmap from TSMC - 2 |
| 4.11.4. | iOIS - Integrated Optical Interconnection System from TSMC |
| 4.11.5. | Combining EIC and PIC with 3D SoIC bond |
| 4.11.6. | Roadmap of bond pitch scaling |
| 4.12. | System integration of OE and ASIC/XPU, etc |
| 4.12.1. | Co-Packaging vs Co-Packaged Optics (CPO) |
| 4.12.2. | Three types of CPO + XPU/switch ASIC packaging structures |
| 4.12.3. | Examples of packaging an optical engine with an integrated circuit (IC) in a 2D or 2.5D configuration |
| 4.12.4. | OE integration with ASIC in a 2.5D configuration |
| 4.12.5. | Examples of packaging an optical engine with an integrated circuit (IC) in a 3D configuration |
| 4.12.6. | Future 3D-CPO structure |
| 4.12.7. | Nvidia's 3D integration of SoC, HBM, EIC and PIC on co-packaged substrates (TSV interposer) |
| 4.12.8. | Example of a 51.2 Tb/s switch module based on 3D integration of EIC/PIC |
| 4.12.9. | Process in fabrication of the 3D heterogeneous integration of EIC and PIC on a glass interposer |
| 4.12.10. | Example of a switch module based on 3D integration of EIC/PIC on glass interposer |
| 4.12.11. | Challenges and Future Potential of CPO Technology |
| 4.13. | Optical alignment and Laser Integration |
| 4.13.1. | How CPO is Built and the Bottleneck |
| 4.13.2. | Interface between Coupler and FAU - Key to the success of CPO packaging |
| 4.13.3. | Grating vs Edge Couplers: Challenges in high-density optical I/O for silicon photonics |
| 4.14. | Fiber Array Unit (FAU) |
| 4.14.1. | Optical alignment challenges and solutions - 1 |
| 4.14.2. | Optical alignment challenges and solutions - 2 |
| 4.14.3. | Optical alignment challenges and solutions - 3 |
| 4.14.4. | Two alignment approaches |
| 4.14.5. | Reducing optical fiber packaging complexity |
| 4.14.6. | Key technical challenge: The size mismatch between silicon waveguides and planar optical fibers |
| 4.14.7. | Fiber Array Unit |
| 4.14.8. | Fiber Attach Methods |
| 4.14.9. | Key players in FAU for CPO |
| 4.14.10. | FOCI (Fiber Optical Communication Inc.) |
| 4.14.11. | FOCI's SiPh and CPO product roadmap |
| 4.14.12. | Baseline Optical Fiber Alignment Structure from FOCI |
| 4.14.13. | Benchmark of Optical Fiber Alignment Structure Variations - 1 |
| 4.14.14. | Benchmark of Optical Fiber Alignment Structure Variations - 2 |
| 4.14.15. | Benchmark of Optical Fiber Alignment Structure Variations - 3 |
| 4.14.16. | Senko Advanced Components |
| 4.14.17. | Senko Advanced Components - Key CPO solution |
| 4.14.18. | Senko Advanced Components - Partnership |
| 4.14.19. | Suppliers of Other Optical Components in CPO (1) |
| 4.14.20. | Suppliers of Other Optical Components in CPO (2) |
| 4.15. | Laser Integration |
| 4.15.1. | On-chip light source integration methods |
| 4.15.2. | External lasers for CPO (1) |
| 4.15.3. | External lasers for CPO (2) |
| 4.15.4. | Laser attach technology benchmark - 1 |
| 4.15.5. | Laser attach technology benchmark - 2 |
| 4.15.6. | Benchmark of different laser integration technology |
| 5. | KEY CPO COMPANIES: DESIGN ROADMAP, STRATEGIES, ECOSYSTEM |
| 5.1. | Supply Chain Overview |
| 5.2. | Broadcom |
| 5.3. | Broadcom CPO Products Benchmark |
| 5.4. | Broadcom's CPO strategy |
| 5.5. | Advanced Semiconductor Packaging Approach for Broadcom's CPO |
| 5.6. | Broadcom's XPU-CPO |
| 5.7. | Detachable fiber connector |
| 5.8. | Broadcom's CPO roadmap |
| 5.9. | Broadcom CPO ecosystem |
| 5.10. | Nvidia: Opportunities for Co-Packaged Optics |
| 5.11. | Nvidia's CPO strategy |
| 5.12. | NVIDIA'S CPO: Spectrum-X and Quantum-X |
| 5.13. | NVIDIA'S CPO - Partnerships |
| 5.14. | Nvidia: Challenges and final thoughts for Co-Packaged Optics |
| 5.15. | Competition between Nvidia and Broadcom |
| 5.16. | NVIDIA vs Broadcom: Strategic Comparison in AI Infrastructure and CPO Technologies |
| 5.17. | CPO product benchmark: NVIDIA vs Broadcom |
| 5.18. | NVIDIA and Broadcom: Divergent CPO Ecosystems |
| 5.19. | Marvell |
| 5.20. | Marvell's XPU-CPO |
| 5.21. | Marvell's CPO reference design |
| 5.22. | Ranovus products and progress - (1) |
| 5.23. | Ranovus products and progress - (2) |
| 5.24. | Ranovus Partnerships |
| 5.25. | MediaTek - CPO strategy/Roadmap/Partnership |
| 5.26. | Lightmatter |
| 5.27. | Lightmatter M1000 |
| 5.28. | Lightmatter - Passage ™: 3D Photonic Interposer (1) |
| 5.29. | Lightmatter - Passage ™: 3D Photonic Interposer (2) |
| 5.30. | Lightmatter: L-Series (3D CPO) |
| 5.31. | Lightmatter: Partners |
| 5.32. | Ayar Labs TeraPHY |
| 5.33. | Ayar Labs Recent Collaboration for AI scale-up - 1 |
| 5.34. | Ayar Labs Recent Collaboration for AI scale-up - 2 |
| 5.35. | Cisco Co-Packaged Optics demo |
| 5.36. | Cisco: CPO power efficiency |
| 5.37. | Cisco: External lasers (ELFPP) |
| 5.38. | Intel Optical Compute interconnect |
| 5.39. | Intel Optical Compute interconnect (2) |
| 6. | MARKET FORECASTS |
| 6.1. | Forecasting the AI accelerator Market: Changes From the previous Edition |
| 6.2. | Forecasting Growth in GPUs and Other Accelerators |
| 6.3. | Forecasts with table: Server boards, CPUs and GPUs/Accelerators |
| 6.4. | Optical I/O for AI interconnect CPO Forecast (units shipped) |
| 6.5. | Optical I/O for AI interconnect CPO Forecast (revenue/market size) |
| 6.6. | CPO network switches (L2 Switches) for AI accelerators forecast (units shipped) |
| 6.7. | CPO network switches (L2 Switches) for AI accelerators data table (units shipped) |
| 6.8. | CPO network switches (L2 Switches) for AI accelerators forecast (market size and revenue) |
| 6.9. | Total CPO market |
| 6.10. | Total CPO by different EIC/PIC integration technology (unit shipment, millions) |
| 6.11. | System integration of network switches (L2 Switches) for AI accelerators forecast by packaging technologies (unit shipped) |
| 6.12. | System integration of Optical I/O Forecast by packaging technologies (units shipped) |
| 6.13. | Table for CPO unit forecast by packaging technologies |
| 7. | COMPANY PROFILES |
| 7.1. | ACCRETECH (Grinding Tool) |
| 7.2. | AEPONYX |
| 7.3. | Amkor — Advanced Semiconductor Packaging |
| 7.4. | ASE — Advanced Semiconductor Packaging |
| 7.5. | Ayar Labs: AI Accelerator Interconnect |
| 7.6. | CEA-Leti (Advanced Semiconductor Packaging) |
| 7.7. | Coherent: Photonic Integrated Circuit-Based Transceivers |
| 7.8. | EFFECT Photonics |
| 7.9. | EVG (3D Hybrid Bonding Tool) |
| 7.10. | GlobalFoundries |
| 7.11. | HD Microsystems |
| 7.12. | Henkel (Semiconductor packaging, Adhesive Technologies division) |
| 7.13. | iPronics: Programmable Photonic Integrated Circuits |
| 7.14. | JCET Group |
| 7.15. | JSR Corporation |
| 7.16. | Lightelligence |
| 7.17. | Lightmatter |
| 7.18. | LioniX |
| 7.19. | LIPAC |
| 7.20. | LPKF |
| 7.21. | Mitsui Mining & Smelting (Advanced Semiconductor Packaging) |
| 7.22. | NanoWired |
| 7.23. | Resonac (RDL Insulation Layer) |
| 7.24. | Scintil Photonics |
| 7.25. | TOK |
| 7.26. | TSMC (Advanced Semiconductor Packaging) |
| 7.27. | Vitron (Through-Glass Via Manufacturing) — A LPKF Trademark |