Greater China photonics and optical interconnect semiconductor ecosystem Report Interpretation

Overview

This CIOE conference update argues that photonics opportunities are broadening for Greater China semiconductor and equipment companies as optical architectures move beyond conventional pluggable modules toward NPO and ultimately CPO. The report highlights USI's external laser-source path, ASMPT's optical-engine packaging technology, and customer roadmaps that support a progression in NPO bandwidth and deployment.

Core views

Morgan Stanley's central conclusion from CIOE is that photonics is creating a wider opportunity set for semiconductor and semiconductor-equipment companies. The report sees external laser sources, optical-engine packaging, and higher-bandwidth NPO architectures as mutually reinforcing parts of the transition from conventional pluggable optics toward NPO and ultimately CPO. For USI, the report says EugenLight, its photonics subsidiary, demonstrated external laser source modules for photonics (ELSFP). Its eight-channel ELSFP delivers more than 23dBm optical output per channel while consuming about 12W. Morgan Stanley considers power efficiency increasingly important as optical I/O scales and system-level power limits tighten. It also views placing the laser source outside the compute/package environment as a strategically important building block for NPO and CPO because it can improve overall system reliability. Although USI's optical-module business has ramped below Morgan Stanley's earlier expectations, CIOE discussions suggested that USI has not fallen behind in ELSFP and retains another route to participate as architectures evolve. Customer design-ins and production ramp are identified as the key next steps. For ASMPT, the report focuses on manufacturing constraints in optical-engine packaging. It identifies warpage and optical-port contamination as two challenges likely to become more significant as optical engines increase in size and complexity. Compared with conventional mass reflow, thermo-compression bonding offers greater control of bonding and may reduce warpage, improving assembly yield and reliability. ASMPT's fluxless TCB, enabled by Atmospheric Oxide Removal, eliminates flux during bonding; the report says this avoids flux residue and lowers contamination risk at sensitive optical ports. Morgan Stanley therefore views the process as well suited to next-generation optical-engine packaging and supportive of ASMPT's position as NPO and CPO scale. Across other exhibitors, the report observed laser-chip capacity expansion and movement toward higher-power continuous-wave products. Source Photonics outlined significant EML and CW capacity expansion for 2027 and targets 200mW CW products for NPO in 1Q27; it also demonstrated a 6.4T ELSFP using in-house CW lasers with eight outputs of about 112mW per channel. Yunling displayed 200mW and 300mW chips, HiSilicon showed a 400mW device, and Everbright launched a 400mW-at-50°C CW DFB chip. Morgan Stanley notes that as products become more available, customer qualification and sustained volume delivery remain differentiators. The report sees NPO optical-engine bandwidth advancing from 3.2T to 6.4T, with examples from Source Photonics and Accelink, and potentially to 7.2T in HiSilicon's integrated-light-source design. VCSEL remains an active NPO route for short-reach applications, illustrated by HiSilicon's 3.2T VCSEL NPO. Tencent and Alibaba presented roadmaps that Morgan Stanley treats as evidence of near-term scale-up demand: Tencent targets deployment of a 512-accelerator supernode using 3.2T NPO in 1Q27, followed by a feasibility assessment in 2Q27; Alibaba targets 4Q26 mass production of 3.2T NPO modules and QFabric switches, followed by 6.4T UPO production around mid-2027. Tencent and Alibaba highlighted NPO's combination of bandwidth density, power consumption, supplier openness, and delivery readiness for scale-up. The report also notes industry views that NPO could be a durable architecture rather than merely a transition, while pluggables may remain preferred for broad scale-out deployment by some participants.

Analysis framework

Morgan Stanley draws its conclusions from CIOE product demonstrations, supplier technology disclosures, and customer deployment roadmaps. It assesses how component power, laser placement, bonding processes, bandwidth progression, and manufacturing constraints affect reliability, yield, and the adoption path for NPO and CPO.

Methodology notes

• Other

  Conference-product and customer-roadmap analysis

  The report combines CIOE demonstrations with supplier capacity plans and customer deployment schedules to assess the likely development path for optical interconnect architectures.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Optical interconnect value-chain analysis

  The report links laser chips, external laser modules, optical-engine packaging, and cloud customer deployments to explain how architectural adoption could affect component and equipment suppliers.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Universal Scientific Ind. (Shanghai) (601231.SS)

  Covered company positioned to participate in evolving optical architectures through EugenLight's ELSFP technology.

  Strengths

  EugenLight's ELSFP is presented as power-efficient and strategically relevant for the migration toward NPO and CPO.

  Weaknesses

  USI's optical-module business has ramped below Morgan Stanley's earlier expectations.

  Comparison

  The report describes ELSFP as an alternative route beyond conventional pluggable optics.

  Risks

  Customer design-ins and production ramp remain key.

• ASMPT (0522.HK)

  Covered company whose fluxless TCB technology is linked to next-generation optical-engine packaging.

  Strengths

  Fluxless TCB may mitigate warpage and reduce optical-port contamination, supporting yield and reliability.

  Comparison

  TCB is described as offering better bonding-process control than conventional mass reflow.

  Risks

  The opportunity depends on NPO/CPO optical-engine scaling.

Key data

• USI EugenLight ELSFP power consumption~12WEight-channel external laser source module; the exhibit states more than 23dBm optical output per channel.
• Source Photonics CW laser target200mW in 1Q27Targeted to support NPO, alongside significant EML and CW capacity expansion for 2027.
• Source Photonics 6.4T ELSFP output~112mW per channelEight-output module using in-house CW lasers.
• Alibaba roadmap3.2T NPO mass production in 4Q26; 6.4T UPO production around mid-2027Alibaba's 6.4T UPO, developed with Accelink, was set for demonstration at the September 2026 Apsara conference.
• Tencent roadmap512-accelerator supernode using 3.2T NPO in 1Q27Followed by a feasibility assessment in 2Q27.
• HiSilicon architecture examples3.2T VCSEL NPO and 7.2T SiPh NPOThe 7.2T design uses integrated CW sources.

Impact & implications

The report indicates that the move to NPO and CPO can expand addressable opportunities across laser suppliers, optical-module vendors, and packaging-equipment providers. It particularly links external laser sources to reliability and power constraints, and fluxless TCB to yield and contamination control in increasingly complex optical engines.

Risks

• USI's optical-module business has so far ramped below Morgan Stanley's earlier expectations.
• Customer qualification and sustained volume delivery remain key differentiators as high-power laser-chip availability broadens.
• USI's customer design-ins and production ramp remain key to monitor.

What to watch

• Alibaba's targeted 4Q26 mass production of 3.2T NPO modules and QFabric switches.
• Tencent's planned 1Q27 deployment of a 512-accelerator supernode using 3.2T NPO and its 2Q27 feasibility assessment.
• Progress toward Alibaba's mid-2027 6.4T UPO production and the related Accelink demonstration.
• USI customer design-ins and production ramp for ELSFP.
• Supplier qualification and sustained-volume delivery of higher-power laser chips.