Optical interconnects and optical computing are seen as important beneficiaries of AI cluster expansion

Overview

This report is a meeting note following Goldman Sachs' visit with Lightelligence's chairman during its China tech tour. The discussion centers on opportunities in optical interconnects and optical computing under the expansion of AI infrastructure. The report argues that as AI clusters move toward higher compute density, greater bandwidth, and lower latency, the potential market space for optical interconnect and optical computing solutions is expanding.

Core views

The report's core view is: first, the upgrade of AI cluster scale-up architectures will create demand for optical interconnects, and the company's solutions can cover electrical packet switching, optical circuit switching, and higher-integration forms such as NPO/CPO; second, the company has been positioning early in optical computing and optical interconnects and has capabilities in complex opto-electronic chips, PIC/EIC design, and co-packaging; third, PACE2 and the in-development PACE3 show progress in optical computing productization, with PACE3 targeting AI inference and LLM decoding and emphasizing low latency and high energy efficiency.

Analysis framework

The report mainly uses management interviews and industry read-throughs, combined with changes in AI infrastructure network architecture, to analyze Lightelligence's potential positioning in scale-up, scale-out, optical interconnects, and optical computing, and maps the company's opportunities to global optical network and AI data center buildout trends.

Methodology notes

• Industry chain read-throughAI infrastructure network upgrade analysis

  Deriving optical interconnect demand from AI cluster architecture upgrades

  The report uses super nodes, scale-up, NPO/CPO, and GPU cluster interconnects as the main thread to judge that higher bandwidth, lower latency, and lower power requirements will expand the optical networking market.

• Company competitiveness analysisAssessment of technical and commercialization capabilities

  Opto-electronic chip design, PIC/EIC, and co-packaging capabilities

  The report focuses on the company's accumulated expertise in complex opto-electronic chips, optical computing processors, packaging, and partnerships with industry players to assess its commercialization capabilities.

• Goldman Sachs methodology disclosureGS Factor Profile

  Growth, Financial Returns, Multiple, and Integrated factor profile

  The disclosure states that Goldman Sachs uses growth, financial returns, valuation multiples, and integrated factors to provide an investment backdrop for stocks, but Lightelligence is Not Covered in this report and no specific factor scores are provided.

Asset mapping & comparison

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

• 1879.HK / Lightelligence

  Research target

  Strengths

  Entered optical computing and optical interconnects early; has complex opto-electronic chip design, PIC/EIC design, and co-packaging capabilities; already has PACE2 and is advancing PACE3; collaborates with foundry, OSAT, and EDA partners to support commercialization.

  Weaknesses

  The report does not disclose key financial or commercialization metrics such as revenue, profit, orders, customer concentration, or mass-production scale; the stock is Not Covered.

  Comparison

  Compared with traditional electrical packet switching, optical circuit switching and higher-integration NPO/CPO solutions are expected to reduce optical-electrical conversions, lower power consumption, and improve signal integrity.

  Risks

  Changes in the pace of AI infrastructure buildout, commercialization progress in optical interconnects and optical computing falling short of expectations, technology-route competition, customer adoption, and uncertainty around large-scale shipments.

• Optical networking / AI infrastructure industry chain

  Read-through beneficiary

  Strengths

  Rising demand from AI clusters for high-bandwidth, low-latency, low-power networks expands the potential TAM for optical interconnects.

  Weaknesses

  The report does not quantify market size, pricing, margins, or penetration rates.

  Comparison

  The report links Lightelligence's opportunity to the expansion of global optical networking demand and believes all kinds of AI networking configurations are likely to grow.

  Risks

  Volatility in data center capex, changes in GPU cluster architecture, competition from alternative interconnect technologies, and supply chain and yield risks.

Key data

• Stock code1879.HKLightelligence, Hong Kong-listed, marked as Not Covered in the report.
• Visit time and location2026-05-22, ShanghaiGoldman Sachs visited the company's chairman during the China Tech Tour.
• Report time2026-05-23 1:49AM HKTTime disclosed in the report header.
• PACE2 specifications128×128 optical matrixUsed to support optical computing clusters.
• PACE3 target specifications256×256 optical matrixAimed at larger-scale commercialization and AI inference scenarios, including LLM decoding.
• PACE2 chip complexity40k+ photonic devicesUsed by management to illustrate the company's complex opto-electronic chip design capability.
• Current price disclosureHK$575.00The disclosure page shows price information related to Lightelligence, but the report does not provide a target price.

Impact & implications

If AI data centers continue to increase per-rack compute density and cluster scale, optical interconnects may extend from traditional scale-out networks into scale-up architectures, driving demand for optical chips, optical modules, optical switching, and co-packaging solutions. Lightelligence matters because it covers both optical interconnects and optical computing, which may benefit from the expansion of the AI infrastructure networking market; however, investors should note that the company is not formally covered by Goldman Sachs, and the report does not provide earnings forecasts, a target price, or a valuation conclusion.

Risks

• The company is Not Covered, with no formal rating, target price, earnings forecast, or valuation framework.
• The report is mainly based on management discussions, and commercialization scale, customer orders, and financial contributions are not quantified.
• Optical computing and routes such as CPO/NPO are still evolving rapidly, and there is uncertainty around technical standards, costs, and mass-production timing.
• If AI infrastructure capex slows, demand realization for optical interconnects and optical computing may be affected.
• Complex opto-electronic chips and co-packaging solutions carry design, yield, supply chain, and large-scale delivery risks.

What to watch

• PACE3 R&D and commercialization progress, especially the 256×256 optical matrix, latency, and energy-efficiency performance.
• Customer adoption and shipment pace for scale-up EPS, scale-up OCS, and NPO/CPO solutions.
• The speed at which AI data centers evolve from scale-out to scale-up and super node architectures.
• The depth of cooperation and capacity assurance with foundry, OSAT, EDA, and other industry partners.
• Whether verifiable disclosures on revenue, orders, customers, gross margin, or mass production emerge.