Silicon photonics and Co-Packaged Optics testing Report Interpretation
Morgan Stanley’s SEMICON Taiwan seminar takeaways highlight rapidly rising silicon-photonics testing demand and a shift toward earlier-stage testing. Large-scale CPO production remains some distance away, with qualification targeted around mid-2027 and broader migration expected from late 2028.
Summary
Morgan Stanley’s SEMICON Taiwan seminar takeaways highlight rapidly rising silicon-photonics testing demand and a shift toward earlier-stage testing. Large-scale CPO production remains some distance away, with qualification targeted around mid-2027 and broader migration expected from late 2028.
- Testing was identified across the supply chain as the principal challenge for CPO mass production.
- SiPh combines optical engines with XPUs and HBM, raising requirements for alignment, automation and data handling.
- ficonTEC plans to ship 1,000 systems in the next 1–1.5 years, versus 2,000 cumulatively over the prior 25 years.
- The industry is emphasizing earlier testing stages to secure Known Good Optical Engines before assembly value and scrap costs rise.
- Advantest targets high-volume-manufacturing qualification by mid-2027; Lightmatter expects NPO to remain in use through 2027.
Report Interpretation
Overview
This conference update summarizes silicon-photonics discussions at SEMICON Taiwan 2026, centering on testing as the main manufacturing bottleneck for Co-Packaged Optics (CPO). The report argues that technical and production hurdles will delay volume deployment, but not the longer-term migration toward CPO.
Core views
Morgan Stanley attended the Silicon Photonics Global Summit on the first day of SEMICON Taiwan 2026, where 15 companies, including Advantest, ficonTEC and Lightmatter, presented. The firm’s central observation was broad supply-chain agreement that testing is the key obstacle to mass production of silicon photonics and CPO. This view came not only from test-equipment vendors but also from foundries and outsourced semiconductor assembly and test providers, indicating that the issue extends across the manufacturing chain. Testing complexity is increasing as optical engines are packaged alongside XPUs and HBM. The report links recent demand growth for Advantest testers to advanced packaging such as CoWoS for XPUs, and argues that SiPh raises the difficulty further. Manufacturability requires high yields at an economical cost, while testing must address tighter alignment, more automation and data beyond conventional electrical testing. In particular, photonic integrated circuit requirements are below one micron, compared with micron-level precision for electronic integrated circuits. The report highlights strong prospective demand for SiPh testing systems. ficonTEC stated that the next five years will require more capacity than it delivered during the preceding 25 years combined. It has shipped 2,000 systems cumulatively over 25 years, but plans to ship 1,000 systems—half that historical total—within the next 1–1.5 years. Morgan Stanley presents this as evidence of a sharp increase in testing demand for silicon-photonics applications. For CPO, the testing flow is expected to span four insertions: separate wafer testing of EIC and PIC wafers; double-sided wafer testing after hybrid bonding of PIC wafers and EIC dies; optoelectrical testing of the optical engine; and final testing after packaging. The report identifies a shift-left trend because scrap costs increase at later assembly stages as more value is added. Securing a Known Good Optical Engine earlier is therefore critical, and Advantest indicated growing emphasis on the first two testing insertions. Insertion 2 is especially challenging because of wafer warpage and alignment with the fiber-array unit. The report also notes that fiber-connector standards remain fragmented across the industry, adding handling complexity. These practical issues help explain why large-scale CPO production remains some way off: Advantest’s objective has moved from an expected high-volume-manufacturing ramp by mid-2027 to completing high-volume-manufacturing qualification by mid-2027. Lightmatter expects Near-Packaged Optics to remain in use through 2027, with migration toward CPO accelerating from late 2028. Despite the testing bottleneck, the report says participants broadly agreed that CPO will become essential because it offers lower power consumption, greater bandwidth and lower latency. Morgan Stanley’s overall impression was that the ecosystem is actively working together to resolve the remaining technical and production challenges.
Analysis framework
The report synthesizes presentations from silicon-photonics participants and traces how testing requirements change through the CPO assembly flow. It combines supply-chain observations, production-process detail, equipment-shipment plans and company timing guidance to assess demand, manufacturability constraints and the likely path from NPO to CPO.
Methodology notes
Supply-chain analysis of silicon-photonics testing
The report compares views from test vendors, foundries and OSATs to show that testing constraints affect the full SiPh and CPO manufacturing chain.
Testing-system capacity and demand comparison
ficonTEC’s historical and planned system shipments are used to illustrate the expected increase in demand for SiPh testing capacity.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- AdvantestTesting-equipment vendor closely connected to the SiPh testing theme and cited as emphasizing early testing insertions.
- Strengths
- Its presentation highlighted increasing emphasis on Insertion 1 and Insertion 2 testing, where earlier screening can reduce downstream scrap risk.
- Risks
- CPO high-volume-manufacturing qualification is targeted only by mid-2027, reflecting unresolved testing and manufacturability challenges.
- ficonTECSupplier of SiPh testing systems cited as evidence of rapidly increasing testing-capacity demand.
- Strengths
- Plans to ship 1,000 systems within 1–1.5 years after shipping 2,000 systems cumulatively over 25 years.
- LightmatterSiPh ecosystem participant providing timing commentary on NPO and CPO adoption.
- Comparison
- Expects NPO to remain in use through 2027 before CPO migration accelerates from late 2028.
- Risks
- Testing remains a key obstacle to large-scale CPO production.
Key data
- SEMICON Taiwan 2026 schedulePre-event forums: August 31–September 1; exhibition: September 2–4The report attended the Silicon Photonics Global Summit on August 31.
- Summit presenters15 companiesIncluded Advantest, ficonTEC and Lightmatter.
- ficonTEC cumulative systems shipped2,000 systems over 25 yearsHistorical shipment base cited by the company.
- ficonTEC planned shipments1,000 systems within 1–1.5 yearsEquivalent to half of its prior 25-year cumulative shipments.
- PIC alignment requirementBelow one micronMore stringent than the micron-level precision cited for EICs.
- CPO HVM qualification targetMid-2027Advantest’s stated objective.
- CPO migration timingAccelerating from late 2028Lightmatter expects NPO to remain in use through 2027.
Impact & implications
The report links the expected migration toward CPO to lower power consumption, greater bandwidth and lower latency, while suggesting that the near- and medium-term opportunity is constrained by the need for more capable testing, earlier defect screening and scalable production processes. It identifies testing as closely connected to Advantest.
Risks
- Testing complexity, including sub-micron PIC alignment, wafer warpage and fiber-array-unit alignment, could constrain CPO manufacturability.
- Fragmented fiber-connector standards add handling complexity.
- High-volume CPO production remains distant, with qualification targeted for mid-2027 and broader migration expected later.