CIOE takeaways reinforce a bullish AIDC optical-network demand outlook, led by 1.6T transceivers and selective NPO opportunities

Overview

Based on CIOE 2026 meetings and technology forums in Shenzhen, Nomura argues that AI data-center optical networking remains in an upcycle. The report centers on the extended 1.6T transceiver cycle, persistent supply bottlenecks, the emerging NPO/OCS architectures, and opportunities for Chinese suppliers across the optical-communications supply chain.

Core views

Nomura attended CIOE 2026 in Shenzhen on 9–10 September, meeting dozens of companies across optical transceivers, components, electronic and optical chips, fiber, materials and equipment. Its central conclusion is that demand for AI data-center (AIDC) optical networks remains strong and that market leaders able to remain ahead of the technology curve should benefit. The institution highlights Zhongji InnoLight, Accelink, TFC, YOFC and Shenzhen T&S as favored companies in high-end transceivers, NPO/CPO components, and AIDC fiber and connectors. The report identifies 1.6T transceivers as the mainstream product for scale-out networks. Better automation could improve assembly efficiency and enable faster capacity expansion, but supply availability—not assembly—is viewed as the main constraint. Nomura expects incumbent 1.6T suppliers, particularly InnoLight, to retain leading positions through effective supply-chain management. It also expects the 1.6T cycle to last longer because 2.4T and 3.2T products face substantial challenges at 400G per lane. Moving beyond conventional silicon photonics may require new materials such as thin-film lithium niobate (TFLN) or modulation approaches such as Coherent Lite; Murata and unlisted Liobate are cited as suppliers making progress in these areas. NPO was a central theme at the exhibition. Hisilicon showed 3.2T VCSEL-based CPO and 7.2T CPO solutions with internal light sources; CIG displayed 6.4T silicon-photonics CPO using an external light source; and Accelink showed a 6.4T NPO product for scale-up networks. Nomura stresses that no industry-wide NPO standards have yet emerged. In its view, early movers can potentially earn higher value content and margins by supplying customized solutions, but only if they secure orders from large global AI customers. The report therefore sees customer proximity and first-mover execution as especially important during initial commercialization. Supply constraints remain material. While 70mW CW lasers and 100G EML/PD chips are already in mass production and support robust 800G transceiver shipments, 100mW-plus CW lasers and 200G EML/PD suppliers still need higher yields to close the demand-supply gap for 1.6T and NPO products. Nomura also flags 3nm DSPs, mSAP-based advanced PCBs, multicore fibers and high-precision ferrules as potential new chokepoints. These shortages could constrain high-end-product shipments, but the report also views them as an opening for domestic suppliers to gain global market share. The next development areas are domestic substitution in materials, lasers and equipment, together with OCS and CPO penetration. Nomura notes progress by Chinese suppliers in InP and SiGe substrates; high-end EML, CW laser and PD chips; multicore and hollow-core fiber; and transceiver automation, coupling and testing equipment. It argues that Chinese companies already hold meaningful positions in transceivers and components, while upstream materials, laser and equipment markets offer further room to expand. OCS is another potential growth avenue. Nomura cites Cigna AI's projection that OCS revenue will exceed USD2bn this year and USD8bn by 2030E, with scale-up OCS linked to GPUs potentially exceeding scale-out OCS applications by 2030E. Solutions displayed included Coherent's liquid-crystal OCS, Accelink's MEMS-based OCS and Taclink's silicon-photonics OCS. However, the institution concludes that OCS architecture remains fragmented and that the supply chain must make further progress before broad commercialization. NVIDIA's discussion of a scale-in network, described as a fifth AI-networking pillar, could become another future demand driver if AI factories are broadly adopted by enterprises.

Analysis framework

Nomura combines observations from CIOE 2026 forums and company meetings with a supply-chain assessment. It traces AI-network demand through transceiver speed migration, component yields and shortages, architecture choices for NPO and OCS, and the positioning of Chinese suppliers across materials, chips, fiber and equipment.

Methodology notes

• Industry AnalysisSupply-demand framework

  Supply-demand assessment of AIDC optical-network components

  The report compares demand for 1.6T and NPO products with available component supply and manufacturing yields to identify bottlenecks and possible shipment constraints.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Optical-communications supply-chain analysis

  Nomura links AI-networking demand to transceivers, components, lasers, chips, substrates, fiber, connectors and equipment, identifying where shortages and domestic substitution could affect participants.

Asset mapping & comparison

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

• Zhongji InnoLight (300308 CH / 3308 HK)

  Favored high-end optical-transceiver and NPO-solution provider; cited as an incumbent 1.6T leader.

  Strengths

  Effective supply-chain management and leading position in 1.6T transceivers.

  Comparison

  Positioned among leading Chinese AIDC optical-network suppliers.

  Risks

  High-end component shortages could constrain 1.6T shipments.

• Accelink (002281 CH)

  Favored supplier with NPO, components and MEMS-based OCS exposure.

  Strengths

  Showcased 6.4T NPO for scale-up networks and MEMS-based OCS.

  Weaknesses

  NPO and OCS architectures remain fragmented.

  Comparison

  Competes with other NPO/CPO and OCS technology approaches.

  Risks

  Commercial success depends on customer orders and broader architecture standardization.

• TFC (300394 CH)

  Favored NPO/CPO component maker.

  Strengths

  Exposure to high-value optical components for emerging architectures.

  Comparison

  Part of the group of Chinese suppliers with leading component positions.

  Risks

  Component and architecture bottlenecks may delay adoption.

• YOFC (6869 HK)

  Favored high-end AIDC fiber and fiber-connector maker.

  Strengths

  Exposure to multicore and hollow-core fiber development.

  Comparison

  Cited alongside other Chinese fiber suppliers.

  Risks

  Multicore fiber and high-precision ferrules are potential supply-chain chokepoints.

• Shenzhen T&S (300570 CH)

  Favored high-end AIDC fiber and fiber-connector maker.

  Strengths

  Exposure to AIDC optical-network infrastructure demand.

  Risks

  High-end supply-chain constraints could affect deployment.

Key data

• CIOE 2026 attendance dates9–10 September 2026Nomura attended technology forums and met dozens of optical-communications supply-chain companies in Shenzhen.
• Mainstream transceiver speed1.6TDescribed as the mainstream product for scale-out networks.
• Next-generation transceiver speeds2.4T / 3.2TStill face technology challenges in the 400G-per-lane era.
• OCS market revenueMore than USD2bn this year; more than USD8bn by 2030ECigna AI presentation cited by Nomura.
• Mass-production components70mW CW laser and 100G EML/PD chipsReportedly supporting robust 800G transceiver shipment growth.
• Constrained components100mW+ CW laser and 200G EML/PD chipsYield improvement is needed to close the supply-demand gap for 1.6T and NPO.

Impact & implications

Nomura believes the AIDC optical-network upcycle favors suppliers that can lead technology transitions and manage constrained component supply. The report sees potential share gains for Chinese companies in upstream materials, lasers and equipment, while NPO and OCS represent longer-term opportunities whose benefits depend on customer wins, architecture evolution and commercialization progress.

Risks

• Shortages and insufficient yields in high-end lasers and 200G EML/PD chips could result in lower-than-expected 1.6T transceiver shipments.
• NPO and OCS architectures lack settled industry-wide standards, which may delay commercialization.
• OCS commercialization requires further progress across the supply chain.

What to watch

• Yield improvement for 100mW-plus CW lasers and 200G EML/PD chips.
• Availability of 3nm DSPs, mSAP-based PCBs, multicore fiber and high-precision ferrules.
• Orders from large global AI customers for early NPO suppliers.
• Progress in domestic substitution across materials, lasers and equipment.
• OCS and CPO penetration, including the development of scale-up and scale-in networks.