---
title: "[Light Core Decoding - Enterprise Breakthrough Chapter] Overview of Core Players in the Track"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/43171076.md"
description: "In the previous two articles, we clarified the computing power value of optical chips, the logic of technological iteration, and the full-chain industrial barriers ranging from wafer materials and epitaxial processes to IDM manufacturing. Overall, the optical chip industry presents a clear tiered structure: mid-to-low-end products have mature processes and fully competitive markets, while high-end high-speed optical chips adapted for AI's ultra-high computing power form a high-barrier, highly concentrated supply landscape due to systematic technical barriers. The continuously exploding demand for AI computing power is constantly elevating the strategic value of high-end optical chips and rewriting the competitive logic of the industry..."
datetime: "2026-08-04T02:12:10.000Z"
locales:
  - [en](https://longbridge.com/en/topics/43171076.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/43171076.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/43171076.md)
author: "[财华社](https://longbridge.com/en/profiles/11651030.md)"
---

# [Light Core Decoding - Enterprise Breakthrough Chapter] Overview of Core Players in the Track

In the previous two articles, we clarified the computing power value of optical chips, the logic of technological iteration, and the full industry chain barriers from wafer materials and epitaxial processes to IDM manufacturing.

Overall, the optical chip industry presents a clear tiered structure: mid-to-low-end products have mature processes and fully competitive markets, while high-end high-speed optical chips adapted for AI ultra-high computing power form a high-barrier, highly concentrated supply landscape 凭借 systematic technical barriers.

The continuously exploding demand for AI computing power is constantly elevating the strategic value of high-end optical chips and rewriting the industry's competitive logic. The traditional development model relying solely on module assembly has reached its growth ceiling. Extending the industry chain upstream to self-develop chips has become an inevitable choice for leading manufacturers in the industry.

Against the backdrop of continuous expansion of AI computing power and a long-term shortage of high-end optical chips, the growth focus and layout direction of industry enterprises are shifting accordingly. This article will systematically sort out the core industry player tiers in the field of optical chips, focusing on core listed and pre-listed enterprises in Hong Kong, and deeply analyze the strategic motivations of leading optical module manufacturers to self-develop optical chips upstream, as well as the realistic dilemmas and breakthrough paths faced by the domestic industry in this "hard battle towards the chip breakout".

**Global Optical Chip Landscape: High-end Oligopoly Monopoly, Mid-tier Catching Up**

From the perspective of the global optical chip industry's competitive landscape, overseas companies currently control the high-end core, while domestic companies are steadily catching up in layers.

In the track of 100G and 200G high-speed EML active optical chips, which are rigid demands for AI computing power, the global market is highly concentrated, with overseas leading oligopolies occupying an absolute dominant position and firmly controlling the supply lifeline of global high-end computing power optical chips.

Among them, the US-based **Lumentum (LITE.US)** is a leading supplier of global high-speed EML optical chips and a core chip supplier for NVIDIA (NVDA.US) AI servers, deeply binding the global top-tier computing power supply chain; Japan's Sumitomo Electric simultaneously controls two core links: high-end Indium Phosphide (InP) substrate raw materials and high-end optical chip manufacturing, mastering the most scarce capacity resources upstream. In addition, international giants such as the US-based Coherent (COHR.US), Mitsubishi Electric, and Broadcom (AVGO.US) divide the remaining high-end market share, forming a relatively stable overseas oligopoly barrier.

With the explosive growth of AI computing power, the global high-end optical chip capacity continues to be tight. Leading computing power giants such as NVIDIA have locked in several years of core capacity from mainstream overseas high-end optical chips through advance payments and long-term procurement agreements. This means that the current surge in optical chip stocks has already reflected the potential strong growth in the next few years to some extent.

According to research firm LightCounting, the industry supply chain is operating at full capacity. At the recent Paris RAISE summit, Lumentum's CEO warned that the next bottleneck for AI infrastructure would be Indium Phosphide (InP), which is more severe than the supply-demand gap currently facing the memory industry. He mentioned that the two current industry giants, Lumentum and Coherent, are simply unable to meet the needs of NVIDIA and other customers. The supply-demand imbalance may exceed 30%—which is larger than his expectations earlier this year.

Indium Phosphide (InP) is a light-emitting compound semiconductor. As mentioned in our industry barrier article, InP is the core material supporting AI high-speed computing power optical modules and is also the mainstream base material for 100G and 200G high-speed EML lasers. Every laser in the optical modules of AI data centers, every pump laser powering the optical amplifiers, and every co-packaged optics (CPO) engine next to the GPU cannot do without it.

This situation has directly led to domestic optical module manufacturers being in a passive position for a long time, often facing problems such as shortages of high-end chips, price increases, and delivery delays. Supply chain stability is difficult to guarantee, becoming a core bottleneck restricting the development of the domestic high-end computing power hardware industry.

In fact, NVIDIA invested $2 billion each in Lumentum and Coherent to help the two suppliers with R&D and capacity.

In March this year, Lumentum announced the purchase of a factory from semiconductor chip manufacturer Qorvo in North Carolina, USA, to produce Indium Phosphide (InP) based optical devices to serve NVIDIA's orders, significantly expanding its capacity for 6-inch Indium Phosphide wafers. However, mass production ramp-up is not expected until mid-2028.

Coherent is also expected to double its internal Indium Phosphide capacity by the end of 2026 and more than double it again in 2027.

Representative enterprise AXT (AXTI.US) in the substrate link recently completed a financing of $632.5 million to expand its Indium Phosphide capacity.

Although these leading suppliers are all expanding production, they cannot immediately solve the supply bottleneck problem. Moreover, they prioritize fulfilling contracts with major North American clients and may not be able to immediately meet the needs of domestic optical module manufacturers. Coupled with the potential risks of trade restrictions, this is estimated to be one of the important reasons for the recent correction of optical module manufacturers that had risen significantly earlier.

**Domestic Player Panoramic Tier: Hong Kong Stock Market Becomes Core Breakout 阵地**

Compared to the full-chain advantages of overseas mature oligopolies, the domestic optical chip industry shows obvious tiered development characteristics, with enterprises having different focuses in layout and significant differences in technical capabilities. The Hong Kong stock market has gathered a batch of distinctive differentiated enterprises, while many A-share leaders are accelerating their rush to A+H listings, leveraging overseas capital and global channels to sprint in the high-end optical chip track, becoming the core main force for domestic substitution.

**1) Core Targets Already Listed on HKEX: Differentiated Layout, Guarding Sub-track Niches**

$XIZHI TECH-P(01879.HK) As the first photonics computing listed enterprise on the Hong Kong stock market, it has carved out a differentiated route distinct from traditional optical chip manufacturers. The company does not deeply cultivate traditional laser emitting optical chips but focuses on optoelectronic hybrid computing power, building a dual-wheel drive business system of optical interconnects and optical computing. Its optical interconnect products are mainly deployed in AI data centers to build high-performance GPU super-nodes.

Interestingly, Lightelligence listed on the Hong Kong Stock Exchange on April 28 this year with an issue price of HK$183.20. It opened 380.35% higher on the listing day, then soared to a high of HK$1,050.00 in the following days, but has now fallen back to HK$273.00. Although it has not broken below the issue price, a drawdown of 74% in less than four months is still quite thought-provoking.

We believe that although the reasons for the stock price decline include poor performance and high valuation corrections, the more main reason might be: The company occupies an 88.3% market share in the domestic independent Scale-up optical interconnect market—usually combining silicon photonic integrated chips with external continuous wave (CW) laser sources, possessing a first-mover advantage. However, Scale-up is short-distance optical interconnect, which does not belong to the currently most 紧缺 EML high-speed transmitting optical chip track for AI computing power. Although Lightelligence dominates the Scale-up optical interconnect market, the overall revenue scale of this sub-track is still small, and there is still a distance to large-scale volume release. Moreover, many giants at home and abroad are competing on the same stage, and there is great uncertainty in the commercialization rhythm.

On the other hand, Lightelligence's core narrative of photonics computing chips—using light to directly complete AI large model calculations, with theoretical power consumption far lower than GPUs—is in a very early commercialization stage given the overall small size of the global optical computing market. Although the company has launched the PACE series of products and cooperated with large model manufacturers such as StepFun and DeepSeek, the short-term contribution may be negligible. Furthermore, major companies like Huawei and NVIDIA are all self-developing optical computing or optical interconnect architectures. Once mainstream major companies choose other solutions, Lightelligence's technology route may be replaced.

In addition, the recent retreat of the overall market sentiment in the Hong Kong stock market's AI and optical chip sectors, along with the downward adjustment of valuations driven by the macro environment, has also dragged down its performance.

Different from the industry mainstream model of purchasing chips externally and focusing on backend packaging, $CREALIGHTS(01191.HK) practices a full-stack silicon photonic integration development path: Starting from the independent R&D of the underlying silicon photonic PDK process kit, the company has completely connected the entire chain of silicon photonic chip architecture design, wafer testing, wafer packaging coupling, and high-speed optical module mass production; self-developed silicon photonic chips integrate passive optical waveguides, silicon photonic modulators, photodetectors, and other core optical components on a single chip; relying on the self-developed "Wafer-In, Module-Out" integrated production platform, all cutting, processing, packaging, and whole-machine optical and electrical testing after the silicon photonic wafer enters the factory are completed internally, forming a closed-loop manufacturing capability.

However, the current main demand for AI computing power is still cross-rack 800G/1.6T pluggable optical modules, and the optimal solution is the traditional architecture of Indium Phosphide EML + DSP. The Scale-up rack internal short-distance silicon photonic interconnects deeply cultivated by Haiguang Xinzhen belong to frontier technologies, with limited short-term order volumes. Moreover, its self-developed full chain has increased short-term costs, its customer concentration is high, and it cannot enter the supply chain of top North American cloud providers. Additionally, when it listed in June, it coincided with the period of the optical interconnect frenzy. The current weak market sentiment has caused its current price to fall below the issue price at the time of listing.

$YOFC(06869.HK) As the global leader in fiber optic cables with A+H listings, its fiber optic preforms and high-end trunk line fiber transmission technology have long maintained a global leading position. The company has also launched AOC active optical cables and high-speed data communication interconnection products through its subsidiaries, successfully entering the supply chains of high-end AI data centers at home and abroad. As the largest shareholder of Changxin Bochuang (300548.SZ), an A-share listed company, the latter has a small layout in passive optical chips and low-speed optical chips.

**Cambridge Technology (06166.HK)** stands out in competitiveness in the field of high-speed computing power optical modules. The company's strategy is deeply bound to its participating enterprise Nanjing Leixin Optoelectronics. Its CW-DFB continuous optical chips have achieved mass production and can meet the needs of external light sources for silicon photonic architectures. However, EML modulator laser chips for ultra-high-speed computing scenarios are still in the R&D sampling stage and have not yet landed in batch production. The overall self-sufficiency rate of high-end optical chips still has considerable room for improvement.

**ZTE (00763.HK)**, listed synchronously in A+H, has an overall layout of optical chips mainly for internal equipment matching: Self-developed optical chips cover two scenarios: fiber access and backbone optical transmission. In the field of long-distance optical transmission, it has achieved full-stack self-developed and mass-produced 100G/400G/800G coherent optical modules and coherent DSP chips; among them, access-side PON chips cater to both self-use and external sales. High-end coherent DSP and coherent optical chips are mainly supplied to its own base stations, optical transmission equipment, and telecom switches, with only a small amount supplied externally. It is worth noting that ZTE's subsidiary, ZTE Photonics, focuses on the R&D of optical devices and optical chips for high-speed interconnection scenarios in intelligent computing centers, achieving substantial progress in high-end optical interconnection layouts.

**2) High-Quality Targets Submitted for Listing on HKEX: High-end Breakthrough, Sprinting to Global Supply Chain**

**Source Star Technology (688498.SH)** is the absolute leader in pure high-speed active EML chips in China and a core benchmark for the breakout of domestic high-end optical chips. The company's 100G EML has completed customer verification and achieved batch shipments, supplying major domestic mainstream optical module manufacturers in batches; its 200G EML chips have entered the verification stage with overseas top customers such as NVIDIA and may achieve mass production by the end of 2026. Meanwhile, its CW continuous light source required for the CPO architecture ranks first in domestic shipment volume, deeply positioning itself in the next-generation optoelectronic integration track. The company has submitted documents to the Hong Kong Stock Exchange, planning to raise funds for the expansion of 200G high-end optical chip capacity, helping the enterprise connect with the overseas computing power supply chain and accelerate the process of high-end domestic substitution.

Also submitted to the Hong Kong Stock Exchange are **Nazhen Technology** and A-share listed **LianTe Technology (301205.SZ),** both of which are leading domestic AI optical module manufacturers. Relying on their main optical module business to generate stable cash flow, they provide continuous funding support for upstream chip R&D. Both enterprises focus on laying out core products such as high-speed EML lasers and silicon photonic integrated chips, extending upstream in the industry chain by leveraging customer resources and process experience accumulated from module businesses.

A-share listed company **Dongshan Precision (002384.SZ)**, which has already submitted documents to the Hong Kong Stock Exchange, has quickly 补齐 high-end optical chip capacity by acquiring mature Indium Phosphide IDM assets from overseas Solectron Optoelectronics. It has become one of the few domestic enterprises capable of mass-producing 200G high-end EML chips and passing certification from leading cloud providers. Self-developed chips are prioritized to match its own optical module business, achieving industry chain synergy.

**Huagong Tech (000988.SZ)** has also submitted documents to the Hong Kong Stock Exchange. Its participating company, Yunling Optoelectronics, serves as an optical chip platform, laying out Indium Phosphide active chips. 25G DFB/EML has achieved mass production, and 100G, 200G EMLs for AI computing scenarios continue to iterate. Relying on Huagong Zhengyuan, it lays out CW light sources and silicon photonic chip R&D, but high-end high-speed chips are still supplemented by external purchases at this stage.

**Barriers and Opportunities for Domestic Optical Chip Breakout**

By sorting out the global landscape and domestic tiers, it is not difficult to find that the domestic optical chip industry has achieved comprehensive autonomous controllability in mid-to-low-end tracks, completing technical breakthroughs and sample breakthroughs in high-end tracks. However, there are still multiple structural barriers to large-scale substitution and global supply.

The main barriers to the current breakout of domestic optical chips may include:

1) Core materials and capacity are locked by overseas entities, lacking supply and demand discourse power. Indium Phosphide substrates and epitaxial wafers are the core cornerstone of high-speed EML optical chips. Currently, global high-end Indium Phosphide capacity and 6-inch and above wafer manufacturing capabilities are highly concentrated in the Japanese and American overseas giants mentioned above. Overseas leading manufacturers have monopolized the core capacity of the next 2-3 years through advance capacity expansion, binding top North American computing power customers, and long-term locking orders. Domestic enterprises not only face problems such as premium prices for raw material procurement and unstable delivery cycles but also find it difficult to obtain quality capacity quotas. Even if chip R&D is completed, large-scale mass production is difficult to achieve.

2) There is a generational gap in underlying processes and toolchains, highlighting the short board of yield rates. High-end 100G/200G EML chips have extremely high requirements for wafer epitaxial precision, etching processes, and packaging coupling precision. Overseas oligopolies have formed mature PDK process kits, standardized production processes, and ultra-high mass production yields through decades of technical accumulation. In contrast, domestic enterprises have weak underlying process accumulation, and their independent PDK systems are not yet fully mature. Some core production equipment and testing instruments still rely on imports, resulting in low yield rates for high-end chip mass production and high production costs. Compared to the scaled cost advantages of mature overseas manufacturers, domestic chips lack short-term cost-effectiveness and cannot quickly cut into the high-end supply chain.

3) The barrier to high-end customer certification is extremely high, making global breakthroughs difficult. AI data centers and overseas cloud providers have extreme requirements for the stability, reliability, and consistency of high-speed optical chips. The customer certification cycle for high-end optical chips generally lasts 1-2 years, and the certification threshold increases year by year. Currently, most domestic enterprises have only achieved batch supply to domestic small and medium-sized customers, and very few have passed the certification of global top computing power customers such as NVIDIA, Meta (META.US), and Amazon (AMZN.US). At the same time, the supply chain ecosystem barriers long formed by overseas giants further hinder the export of domestic chips, causing domestic high-end products to be limited to niche domestic markets, making it difficult to open up incremental space.

4) Diversified iteration of technology routes leads to high R&D error tolerance costs. The current industry is in a critical stage of parallel iteration of traditional pluggable optical modules, NPO near-package optics, and CPO co-package optics. Technologies such as silicon photonic integration, thin-film lithium niobate, and indium phosphide discrete chips have their own advantages and disadvantages. Leading overseas enterprises comprehensively lay out multiple tracks to hedge against iteration risks. In contrast, domestic enterprises have limited funds and R&D resources. If they bet on a single technology route, they are extremely likely to face the risk of elimination due to technological iteration; multi-route layouts will also disperse R&D efforts and delay breakthrough rhythms, putting enormous pressure on commercialization error tolerance.

5) Insufficient industrial synergy and fragmented development of upstream and downstream links. There is an obvious fragmentation problem in the domestic optical chip industry chain: insufficient linkage between upstream substrate materials and epitaxial process enterprises and midstream chip design and manufacturing manufacturers. Downstream optical module enterprises and equipment manufacturers lag behind in adapting to upstream chip R&D. Most module manufacturers are still accustomed to the overseas chip procurement model and lack enthusiasm for adapting, verifying, and introducing domestic chips, which may lead to a lack of large-scale application scenarios for domestic chips, and technical iteration and yield optimization lack data support.

But at the same time, domestic optical chips may be ushering in a valuable substitution window because:

1) Overseas supply continues to be tight. As mentioned earlier, the current EML supply-demand gap is significant. Although overseas leading manufacturers continue to expand capacity, the construction of Indium Phosphide production lines and the capacity ramp-up cycle are long. New capacity from Lumentum and Coherent will be released in 2027-2028, and the short-term supply-demand mismatch pattern cannot be reversed. At the same time, overseas capacity prioritizes fulfilling orders for top North American computing power customers. Domestic manufacturers are always at the end of the supply priority list, and problems such as shortages, price hikes, and delays have become normalized, forcing the downstream industry chain to accelerate the introduction of domestic substitution solutions.

2) Domestic technology achieves group breakthroughs, and product landing accelerates. Domestic manufacturers have formed differentiated breakthroughs in key tracks. The process maturity and controllable yield rates in the CW continuous light source field have become the core breakthrough point for domestic substitution. Head enterprises such as Source Star Technology and Dongshan Precision rank at the forefront of domestic shipment volumes. In the core track of high-end EML, domestic technology iteration continues to accelerate. Multiple manufacturers have achieved stable mass production and batch supply of 100G EMLs. 200G EML chips have successively completed sampling and head customer verification, gradually breaking the exclusive monopoly pattern of overseas high-end products.

3) Silicon photonic technology route iteration brings structural overtaking opportunities. Industry reports predict that in 2026, the sales volume of optical modules based on silicon photonic architecture will account for more than 50% of the global market for the first time, becoming the mainstream solution for computing power optical modules. The silicon photonic architecture no longer relies on traditional high-speed EML chips but adopts a combination scheme of CW external light sources + silicon-based modulators, transferring the technical difficulties from laser epitaxial manufacturing to modulator design and optical coupling packaging. For the domestic industry, this is a differentiated path that can partially reuse mature CMOS capacity, effectively avoiding the short-term shortcomings of domestic Indium Phosphide EMLs, and providing an excellent era opportunity for the large-scale release of domestic optical chips and rapid occupation of the mid-to-high-end market.

**Conclusion**

Standing at the node of rapid expansion of AI computing power, the global optical chip industry is at a crossroads of supply-demand imbalance, technological route alternation, and reshaping of the supply chain landscape. Overseas oligopolies rely on the full-chain capacity accumulation of Indium Phosphide, mature process systems, and long-locked global top computing power customers to firmly guard the moat of high-end high-speed EML chips. The short-term tight supply situation is difficult to fundamentally alleviate, continuously amplifying the realistic pressure of "supply constraints" in the upstream of the domestic computing power industry chain.

Looking at the domestic market, domestic optical chips have completed the initial stage of autonomy in mid-to-low-end products. Different manufacturers choose development paths based on their own endowments: some deeply cultivate Indium Phosphide active chips to tackle mainstream computing power tracks, while others bet on silicon photonic integration to layout short-distance optical interconnects and next-generation optoelectronic integration solutions. The trend of operators and leaders in 细分 tracks of the optical industry chain extending their businesses to self-developed chips is intensifying. But objectively speaking, the journey of domestic high-end optical chips towards large-scale global substitution is still long and arduous. Constraints on core material capacity, gaps in process yields, long customer certification cycles, shortcomings in industry chain synergy, and pressures from multi-route R&D constitute multiple tests lying ahead.

Within the crisis lies a rare time window. The release pace of new overseas Indium Phosphide capacity lags behind the growth rate of computing power demand, and the demand for supply chain autonomy of downstream enterprises continues to increase; 100G EMLs in China have been successfully mass-produced, 200G EMLs are gradually entering the verification cycle, and CW light sources have achieved large-scale shipments; coupled with the rise of silicon photonic solutions bringing structural changes to the track, new competitive tracks have been opened up for domestic enterprises.

In the long run, this "breakout towards the chip" will not be a short-term race that can be accomplished overnight, but a long-term systematic 攻坚战 spanning materials, epitaxy, manufacturing, packaging, and downstream verification. Technological breakthroughs by a single enterprise are not enough to complete industrial leapfrogging. Only by promoting deep synergy between upstream and downstream, continuously polishing processes, stabilizing yields, and connecting domestic and overseas customer certification channels can domestic optical chips seize the dividends of this round of AI computing power. In the future, whoever can continuously consolidate self-research capabilities, match the iterative needs of downstream computing power hardware, and build a stable and controllable capacity system will be able to stand firm in the new round of global optical chip landscape reshuffling, truly realizing the identity leap from a track follower to a core participant.

Author: Wu Yan

### Related Stocks

- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [AXTI.US](https://longbridge.com/en/quote/AXTI.US.md)
- [COHR.US](https://longbridge.com/en/quote/COHR.US.md)
- [01879.HK](https://longbridge.com/en/quote/01879.HK.md)
- [NVDL.US](https://longbridge.com/en/quote/NVDL.US.md)
- [07788.HK](https://longbridge.com/en/quote/07788.HK.md)
- [07388.HK](https://longbridge.com/en/quote/07388.HK.md)
- [NVDY.US](https://longbridge.com/en/quote/NVDY.US.md)
- [NVDD.US](https://longbridge.com/en/quote/NVDD.US.md)
- [NVDX.US](https://longbridge.com/en/quote/NVDX.US.md)
- [NVDQ.US](https://longbridge.com/en/quote/NVDQ.US.md)
- [01191.HK](https://longbridge.com/en/quote/01191.HK.md)
- [06869.HK](https://longbridge.com/en/quote/06869.HK.md)
- [601869.CN](https://longbridge.com/en/quote/601869.CN.md)
- [AVGO.US](https://longbridge.com/en/quote/AVGO.US.md)
- [LITE.US](https://longbridge.com/en/quote/LITE.US.md)