• 亿邦公众号
    • APP下载

开盘暴涨80%!清华博士追“光”15年 今日敲钟

作者:漫地 文章来源:创业邦

2026-06-30 10:57

作者:漫地 文章来源:创业邦

2026-06-30 10:57

邦小白快读

EN
全文速览

本文核心是清华博士胡朝阳创立的硬科技企业海光芯正登陆港交所的相关信息,核心干货整理如下

1. 上市核心信息:海光芯正6月29日港交所敲钟,发行价114港元/股,发行市值102亿港元,开盘价199.9港元/股,开盘暴涨80%,开盘市值超180亿港元。

2. 企业成长核心数据:公司2023-2025年营收从1.75亿元增长至12.21亿元,三年复合增长率达163.9%,按2025年收入计,海光芯正在AI光模块领域全球排名第八、中国排名第七,全球市场份额1.6%。

3. 可参考的创业经验:创业初期不正面对标国际巨头,选择从细分市场切入逐步升级,十年前提前押注尚不成熟的硅光技术,成功踩中AI爆发的行业风口,同时也要注意,硬科技创业前期投入大周期长,海光芯正目前仍处于持续亏损阶段。

本文梳理了AI光模块行业的发展趋势和海光芯正的品牌发展路径,相关干货整理如下

1. 产业发展趋势:生成式AI爆发带动高速光模块需求快速增长,行业核心竞争已经从模块组装转向硅光芯片和光子集成平台能力,市场加速向头部集中,预计2026年全球800G和1.6T高速光模块市场规模将达146亿美元。

2. 产品研发经验:海光芯正走差异化全链条自研路线,打造从硅光芯片设计到光模块制造的端到端能力,通过技术优化将硅光芯片流片成本降低30-40%,整体光模块成本比竞争对手低20-30%,提前布局下一代1.6T、3.2T、CPO等技术,早早建立起技术壁垒。

3. 品牌发展合作经验:采用自有品牌、JDM、ODM三种模式并行,通过JDM模式深度绑定大客户获得稳定订单,同时引入国资和产业资本,国资合计持股超三分之一,阿里小米等产业资本入股既带来资金也带来生态协同,支撑长期研发投入。

本文梳理了AI光模块行业的现状、机会和风险,相关干货整理如下

1. 行业变化与增长机会:AI大模型发展带动高速光模块需求爆发,行业增长确定性强,当前竞争核心已经转向上游硅光芯片自研能力,提前布局3.2T预研、硅光集成、CPO等下一代技术的玩家,能获得明显先发优势,不同定位玩家都有发展空间,第三方供应商可靠规模和客户广度抢占市场,内部制造商可靠垂直整合保障供应链优势。

2. 明确风险提示:当前赛道竞争白热化,价格战压缩了全行业利润空间,头部巨头包圆大量高端芯片产能,中小厂商面临20%-30%的供需缺口,行业头部效应明显,前三名合计占据61.93%的市场份额,缺乏800G/1.6T高速产品交付能力的玩家生存空间被大幅压缩,新玩家前期普遍面临持续亏损问题。

3. 可学习的发展经验:可以参考海光芯正的路径,从细分市场切入,提前布局核心技术,通过JDM模式绑定大客户稳定订单,逐步建立自身技术壁垒。

本文梳理了AI光模块行业对产品生产设计的需求,以及给制造工厂带来的机会和启示,干货整理如下

1. 产品生产设计需求趋势:硅光子技术是未来光互连的必然发展方向,和传统模式相比,硅光技术可以把多个光学器件集成到同一块芯片上,解决传统模式成本高、体积大、功耗高的痛点,已经成为行业竞争核心。海光芯正采用Less change CMOS设计方案,可以共享传统12英寸CMOS产线,流片成本比海外流片降低30-40%,整体光模块成本比对手低20-30%,这个生产设计思路值得制造端参考。

2. 全流程自主布局的价值:海光芯正搭建了Wafer In,Module Out一体化制造平台,硅光晶圆从进厂到成品出货的全流程工序都自主完成,能够更好控制成本、质量和响应速度,适配高速迭代的光模块行业需求。

3. 商业机会与转型启示:AI带动高速光模块需求持续增长,下一代光电集成技术落地需要大量产业链制造配套,有技术能力的工厂可以提前布局硅光相关制造环节,推进数字化和一体化产能建设,抓住行业增长红利。

本文梳理了AI光模块行业的发展趋势和行业痛点,给相关服务商的干货整理如下

1. 行业发展趋势:生成式AI的爆发让光通信从产业链配角变成核心环节,高速光模块需求保持高速增长,行业技术门槛不断拉高,市场加速向头部集中,核心竞争从模块组装转向硅光芯片和光子集成平台能力,下一代技术围绕1.6T、3.2T、CPO等方向发展,行业整体增长空间较大。

2. 行业现存核心痛点:中小厂商普遍面临高端芯片供应短缺问题,存在20%-30%的供需缺口,硬科技企业创业前期研发投入大、设备成本高、客户认证周期长,很多中小企业缺乏全产业链布局的资金和能力,头部企业在下一代技术研发上也需要产业链配套支持。

3. 业务机会方向:服务商可以针对行业痛点,布局高端芯片国产化供应、中小厂商产能共享对接服务,也可以为企业提供硅光技术研发咨询、产线升级改造、早期融资对接等服务,抓住AI光模块行业增长的红利,挖掘新的业务增长点。

本文梳理了AI光模块行业企业的需求和行业风向,给平台商的干货整理如下

1. 行业企业的核心需求:硬科技光通信企业研发周期长、前期投入大,对长期资金支持、产业链资源对接的需求很高,海光芯正从成立到IPO累计完成9轮融资,累计融资超9亿元,国资合计持股超三分之一,阿里小米等产业资本入股既带来资金也带来客户资源和生态协同,对企业发展至关重要。

2. 平台可落地的运营方向:平台针对AI光通信这类硬科技赛道,可以推出针对性的招商和扶持政策,重点引入拥有核心自研技术的企业落地,对接国资、产业资本等不同类型的投资资源,帮助企业解决早期融资难题,同时对接产业链上下游资源,帮助企业对接客户,缩短认证周期。

3. 需要规避的风向:光模块赛道当前价格战激烈,多数企业前期都处于持续亏损状态,缺乏核心自研能力的组装型企业生存压力大,平台在招商和运营中要做好风险评估,重点布局有核心技术壁垒的企业,规避技术落后缺乏竞争力的项目。

本文梳理了AI光模块行业的最新产业动向和竞争格局,给研究者的干货整理如下

1. 产业最新动向:生成式AI爆发带动高速光模块需求快速增长,光通信企业从产业链配角成为资本市场和产业关注的焦点,行业竞争逻辑发生变化,核心竞争已经从谁能组装光模块转向谁掌握硅光芯片和光子集成平台,预计2026年全球800G和1.6T高速光模块市场规模将达到146亿美元,下一代竞争围绕3.2T预研、硅光集成、CPO等前沿技术展开。

2. 行业竞争新格局:当前AI光模块行业头部效应明显,前三名企业合计占据61.93%的市场份额,行业分化加剧,缺乏高速产品能力的企业生存困难,行业玩家分为第三方供应商和内部制造商两类,第三方供应商是市场主流,规模大但依赖海外高端芯片,内部制造商垂直整合能力强,供应链安全有优势,但规模小创新动力不足,拥有上游核心器件自研能力的企业优势越来越明显。

3. 值得研究的新商业模式:海光芯正长期提前投入核心技术、从细分市场切入逐步升级、自有品牌+JDM+ODM多模式并行绑定大客户的硬科技创业路径,为国产替代领域的硬科技创业提供了新的研究样本。

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article summarizes key updates on HG芯cheng (Haiguang Xzheng), a hard-tech firm founded by Tsinghua PhD Hu Chaoyang, ahead of its Hong Kong IPO, as follows:

1. Core IPO details: Haiguang Xzheng will ring the bell on the Hong Kong Stock Exchange on June 29, priced at HK$114 per share, with an initial market cap of HK$10.2 billion. It opened at HK$199.9 per share, surging 80% on debut, pushing its opening market cap above HK$18 billion.

2. Key growth metrics: The company's revenue grew from RMB 175 million in 2023 to RMB 1.221 billion in 2025, representing a three-year compound annual growth rate (CAGR) of 163.9%. Based on 2025 revenue, Haiguang Xzheng ranks 8th globally and 7th in China in the AI optical module market, holding a 1.6% global market share.

3. Actionable startup takeaways: Instead of competing head-to-head with global giants in its early days, the company started in a niche market and upgraded its capabilities gradually. It bet on the still-maturing silicon photonics technology a decade ago, successfully catching the boom of the AI industry. It is worth noting that hard-tech ventures require large upfront investment and long R&D cycles, and Haiguang Xzheng remains unprofitable to date.

This article summarizes AI optical module industry development trends and Haiguang Xzheng's brand growth path, as follows:

1. Industry development trends: The generative AI boom has driven rapid growth in demand for high-speed optical modules. Industry competition has shifted from module assembly to silicon photonic chips and photonic integration platform capabilities, and the market is increasingly concentrated among top players. The global 800G and 1.6T high-speed optical module market is projected to reach $146 billion by 2026.

2. Product R&D insights: Haiguang Xzheng follows a differentiated full-stack in-house R&D strategy, building end-to-end capabilities from silicon photonic chip design to optical module manufacturing. Through technological optimization, it has cut silicon photonic chip tape-out costs by 30-40%, and its overall optical module costs are 20-30% lower than competitors. It has pre-positioned R&D for next-generation technologies including 1.6T, 3.2T and CPO, building solid technological barriers early on.

3. Brand development and partnership insights: It operates three parallel models: proprietary brand, JDM and ODM. The JDM model helps it deeply tie up with key clients to secure stable orders. It has also onboarded state-owned capital and industrial capital: state-owned investors hold a combined stake of over one-third, while industrial investors including Alibaba and Xiaomi provide not only capital but also ecosystem synergy, supporting its long-term R&D investment.

This article summarizes the current status, opportunities and risks of the AI optical module industry, as follows:

1. Industry changes and growth opportunities: The rise of large AI models has driven a boom in demand for high-speed optical modules, and the industry's growth trajectory is highly certain. Competition now centers on in-house upstream silicon photonic chip capabilities, and players that pre-position next-generation technologies including 3.2T R&D, silicon photonic integration and CPO will gain clear first-mover advantages. Players with different positioning all have room for growth: third-party suppliers can capture market share through scale and broad client coverage, while in-house manufacturers can secure supply chain advantages via vertical integration.

2. Clear risk warnings: The sector is now seeing cutthroat competition, and price wars have compressed profit margins across the industry. Top giants have locked up most high-end chip capacity, leaving small and medium-sized (SME) players facing a 20-30% supply gap. The industry is heavily concentrated: the top three players control a combined 61.93% of the market. The survival space for players without 800G/1.6T high-speed product delivery capabilities has shrunk sharply, and new entrants almost always face sustained losses in early stages.

3. Actionable growth lessons: Sellers can follow Haiguang Xzheng's path: start from a niche market, pre-invest in core technologies, secure stable orders by tying up with key clients via the JDM model, and gradually build own technological barriers.

This article summarizes product manufacturing and design requirements for the AI optical module industry, as well as opportunities and insights for manufacturing plants, as follows:

1. Product manufacturing and design trends: Silicon photonics is the inevitable future development direction for optical interconnection. Compared with traditional approaches, silicon photonics integrates multiple optical components on a single chip, solving the pain points of high cost, large size and high power consumption of traditional solutions, and has become the core of industry competition. Haiguang Xzheng adopts a "Less Change" CMOS design scheme that can share existing traditional 12-inch CMOS production lines, cutting tape-out costs by 30-40% compared with overseas tape-out, and lowering overall optical module costs by 20-30% compared to competitors. This manufacturing and design approach is a valuable reference for manufacturing players.

2. Value of full-process in-house layout: Haiguang Xzheng has built an integrated "Wafer In, Module Out" manufacturing platform, handling all processes from silicon photonic wafer incoming to finished module shipment in-house. This allows better control over cost, quality and response speed, adapting to the fast iteration needs of the optical module industry.

3. Business opportunities and transformation insights: AI has driven sustained growth in demand for high-speed optical modules, and the commercialization of next-generation optoelectronic integration technology requires extensive supporting manufacturing from the industrial chain. Manufacturing plants with technical capabilities can pre-position silicon photonics-related manufacturing links, advance digital and integrated capacity building, and capture the industry's growth dividend.

This article summarizes development trends and core pain points of the AI optical module industry for relevant service providers, as follows:

1. Industry development trends: The generative AI boom has elevated optical communication from a supporting role in the industrial chain to a core link. Demand for high-speed optical modules continues growing rapidly, technical barriers are rising, the market is increasingly concentrated in top hands, and core competition has shifted from module assembly to silicon photonic chips and photonic integration platform capabilities. Next-generation technology development focuses on 1.6T, 3.2T, CPO and other directions, and the industry as a whole has large growth room.

2. Existing core industry pain points: Small and medium-sized enterprises (SMEs) generally face shortages of high-end chip supply, with a 20-30% supply gap. Hard-tech startups require large upfront R&D investment, high equipment costs and long client certification cycles, and most SMEs lack the capital and capacity for full industrial chain layout. Even leading companies need supporting industrial chain support for next-generation technology R&D.

3. Business opportunity directions: Targeting these industry pain points, service providers can lay out localized supply of high-end chips and production capacity matching and sharing services for SMEs. They can also offer silicon photonics R&D consulting, production line upgrade and transformation, and early-stage financing matching services, to capture the growth dividend of the AI optical module industry and unlock new business growth points.

This article summarizes corporate needs and industry trends in the AI optical module sector for platform operators, as follows:

1. Core demands of industry players: Hard-tech optical communication companies have long R&D cycles and large upfront investment, so they have strong demand for long-term capital support and industrial chain resource matching. From founding to IPO, Haiguang Xzheng completed 9 rounds of financing, raising over RMB 900 million in total. State-owned investors hold a combined stake of over one-third, while industrial investors including Alibaba and Xiaomi provide not just capital but also client resources and ecosystem synergy, which was critical to the company's growth.

2. Actionable operational directions for platforms: For hard-tech tracks like AI optical communication, platforms can roll out targeted investment attraction and support policies, prioritize introducing companies with core in-house R&D technologies, match them with different types of investment resources including state-owned capital and industrial capital to solve their early-stage financing difficulties, and connect them with upstream and downstream industrial chain resources to help them access clients and shorten certification cycles.

3. Risk warnings to note: The optical module track is currently experiencing fierce price wars, and most companies remain in sustained losses in early stages. Assembly-focused companies without core in-house capabilities face huge survival pressure. Platforms should conduct thorough risk assessment during investment attraction and operation, prioritize projects with core technological barriers, and avoid technically backward, uncompetitive projects.

This article summarizes the latest industry developments and competitive landscape of the AI optical module industry for researchers, as follows:

1. Latest industry developments: The generative AI boom has driven rapid growth in demand for high-speed optical modules, and optical communication companies have shifted from supporting players in the industrial chain to the focus of capital and industry attention. The competitive logic of the industry has changed: core competition has shifted from who can assemble optical modules to who controls silicon photonic chips and photonic integration platforms. The global 800G and 1.6T high-speed optical module market is projected to reach $146 billion by 2026, and next-generation competition centers on frontier technologies including 3.2T pre-development, silicon photonic integration and CPO.

2. New industry competitive landscape: The current AI optical module industry is heavily concentrated, with the top three players holding a combined 61.93% of the market share. Industry polarization is intensifying, and players without high-speed product capabilities face growing survival difficulties. Industry players fall into two categories: third-party suppliers and in-house manufacturers. Third-party suppliers are the mainstream of the market: they have large scale but rely on overseas high-end chips. In-house manufacturers have strong vertical integration capabilities and advantages in supply chain security, but are smaller in scale and have weaker innovation incentives. Companies with in-house R&D capabilities for core upstream components are gaining increasingly clear competitive advantages.

3. A new business model worth studying: Haiguang Xzheng's hard-tech startup path—pre-investing in core technologies long in advance, starting from a niche market to upgrade gradually, and tying up with key clients via a parallel model of proprietary brand + JDM + ODM—provides a new research sample for hard-tech entrepreneurship in China's import substitution space.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

作者丨漫地

编辑丨关雎

今日(6月29日)海光芯正敲钟港交所,发行价114港元/股,发行市值102亿港元。开盘价199.9港元/股,开盘市值超180亿港元。

海光芯正的创始人胡朝阳是清华博士,曾在美国加州大学及多家光通信企业深耕十余年。2011年回国创立海光芯正,旨在填补国内中高端光器件空白。创业初期主攻低速产品,后转向硅光子技术。

根据招股书披露,公司营收在三年内实现了惊人增长——从2023年的1.75亿元升至2024年的8.62亿元,再到2025年的12.21亿元,三年复合增长率高达163.9%。按2025年收入计,公司在全球专业光模块提供商中排名第十二、市场份额0.8%,在AI光模块领域全球排名第八、中国排名第七、市场份额1.6%。

海光芯正自2011年成立至IPO前,累计完成了9轮融资,累计融资金额超过9亿元,在最后一轮(F+轮)融资的投后估值为26.6亿元人民币。

上市后,海光芯正创始人胡朝阳直接持有公司9.44% 的股份。此外,他通过员工激励平台苏州海怡和苏州海旭,合计控制公司17.95% 的投票权。

在外部机构股东中,国资背景股东占据重要地位,主要包括苏州高新创投(通过苏州融联创投、苏州汇琪、聚升创投、金谷源鑫合计持有8.72%)、江苏交通控股(通过苏州协立创投、苏州君实协立、君鼎协立合计持有7.70%)、天津泰达(5.61%)、北京信息发展基金(4.79%)、北京经开区产业升级基金二期(4.79%)、苏州工业园区管理委员会(通过元禾、苏州启纳合计持股2.62%)等。国资机构合计持股比例超过三分之一。

产业资本及战略投资者方面,主要包括博华资本(4.79%)、阿里巴巴中国(4.02%)、小米智造(2.3%)等,这些机构的入股主要基于产业链上下游协同及生态布局的考量。另外,财务投资机构中邦盛投资(通过江苏高投邦盛、邦盛赢新、邦盛聚源合计持有5.89%)持有份额较大。

苏州高新创投是公司最大外部股东,累计投资5700万元,持股8.72%。按照开盘市值180亿港元(156亿元人民币)计算,其持股价值13.6亿元,回报倍数为23.9倍。

清华博士回国追“光”

2011年的北京,一间不足300平方米的办公室里,只有5个人的创业团队正在研发一种大多数中国企业尚未真正掌握的技术——高速光通信芯片与器件。

彼时,中国已经是全球最大的互联网市场之一,但在高速光模块、光芯片等核心领域,仍严重依赖国外供应商。对于刚从美国回国的胡朝阳来说,这既是一种遗憾,也是一种机会。

胡朝阳并非典型的连续创业者。他本科毕业于北京航空航天大学,后在清华大学获得博士学位,长期从事光电子和光通信技术研究。随后赴美工作多年,参与光通信器件和集成光电子技术研发。胡朝阳在加州大学圣芭芭拉分校等顶尖机构从事研究,并在多家美国知名光通信企业担任技术负责人。

在美国期间,互联网流量的爆发式增长让他敏锐意识到一个趋势:未来的信息社会,真正决定算力和数据流动效率的,不只是计算芯片,还有负责数据传输的光通信技术。

彼时业内流传着一句话:CPU负责计算,光模块负责让计算发生。这意味着服务器之间、数据中心之间乃至全球互联网的数据交换,都需要依靠高速光通信完成。

然而,中国市场巨大的需求背后,高端产品却长期依赖进口。胡朝阳后来回忆,自己创业的初衷非常明确:“希望做出中国自己的核心光通信器件和芯片。”

2011年,他放弃海外稳定的职业发展机会,回到国内创立海光芯正。

与典型的硬科技企业相似,创业初期的海光芯正面临诸多问题:技术研发周期长、设备投入巨大、市场验证慢、客户认证周期往往长达一年甚至更久。

在创业早期,海光芯正并没有直接挑战国际巨头。胡朝阳深知:在资金、规模和客户资源远远不如国外企业的情况下,正面竞争几乎没有胜算。因此公司采取了一条更务实的路线——从细分市场切入。先做高速光器件,再向光模块延伸,最后布局光芯片。

随着中国云计算产业兴起,阿里云、腾讯云等大型数据中心开始建设,高速光通信需求迅速增长。

而海光芯正在这个时候开始对硅光技术进行深度布局。在十年前,这是一项投入巨大、商业前景尚未完全明确的技术。但一旦成功,将给企业建立极高的技术壁垒。

因为在传统光模块中,激光器、调制器、探测器等器件往往需要分别制造,再进行复杂组装。这种模式成本高、体积大、功耗高。而硅光技术则试图把这些光学器件像集成电路一样集成到同一块芯片上,胡朝阳判断这是未来光互连的必然方向。

正当许多企业选择观望时,海光芯正却决定提前下注。此后多年,公司持续投入研发,逐步形成从硅光芯片设计、封装测试到高速模块制造的完整能力。

2022年以后,生成式AI迅速崛起。大模型训练需要成千上万块GPU协同工作。而GPU之间的数据交换速度,成为决定训练效率的关键因素。这时,光模块的重要性被重新认识,过去属于产业链配角的光通信企业,也开始成为资本市场关注的焦点。

海光芯正此前十余年积累的硅光技术能力,在这时迎来了需求爆发。公司产品逐步覆盖100G、200G、400G、800G高速光模块,并进入了AI数据中心市场。

营收爆发式增长

2023年至2025年,海光芯正的营收呈现出爆发式增长。公司收入从1.75亿元增至12.21亿元,三年复合增长率高达163.9%。

其中,光模块是公司的绝对核心收入来源,2025年光模块收入达9.24亿元,占总收入的75.7%;AOC(有源光缆)收入2.48亿元,占比20.3%。

凭借这一增长势头,按2025年AI光模块收入计,海光芯正在全球专业AI光模块提供商中排名第八、中国排名第七,全球市场份额为1.6%。2022年至2024年间,公司更是全球前十大专业光模块供应商中收入增长最快的企业。

与多数光模块厂商“外购芯片+模组拼装”的模式不同,海光芯正最大的差异化在于从芯片设计到光模块制造的端到端硅光子技术能力。目前,海光芯正已商业化四款硅光子光模块产品,主要集中在400G及以上速率——所有400G及以上规格的单模光模块均采用硅光子技术。

在成本控制方面,海光芯正把成本压缩到极致。公司硅光子芯片采用“Less change CMOS”设计,可与传统CMOS集成电路共享12英寸产线产能,相比海外流片成本降低30-40%,光模块整体成本较可比竞争对手低20-30%。

此外,公司还搭建了“Wafer In,Module Out”一体化制造平台,硅光晶圆进厂后,测试、切割、封装、成品检测等全流程工序均自主完成。在这背后,公司持续加大研发投入,2025年研发费用达1.04亿元,同比增长63.0%。

目前,海光芯正三种业务模式并行:自有品牌模式(2025年收入占比50.7%)、联合设计制造(JDM)模式(占比45.3%)和ODM模式(占比4.0%)。其中JDM模式通过与国内大客户深度绑定、联合开发定制产品以换取稳定的订单量和产能利用率。虽然海光芯正未公开披露其JDM大客户的具体身份,但市场普遍认为其核心JDM客户正是阿里巴巴、小米等产业资本股东。同时,海光芯正也与国内头部独立半导体存储解决方案与封测厂商佰维存储(688525.SH),建立了深度的战略合作关系。

在不同的业务模式下,海光芯正当前已实现AI数据中心长、中、短距离互连的产品全覆盖。其中光模块用于服务器与交换机之间的较长距离数据传输,AOC用于机柜内或相邻设备间的短距互连需求。公司还于2025年12月商业化推出AEC(有源铜缆)产品,进一步补充超短距离传输场景。

而对于当前光通信领域最具战略意义的产品之一1.6T光模块,海光芯正的1.6T硅光模块已进入客户验证阶段,3.2T/6.4T光电芯片同步研发推进。同时全面布局NPO(近封装光学)、CPO(共封装光学)等下一代光电集成技术。

不过值得注意的是,营收高速增长的另一面是持续亏损。2023年至2025年,海光芯正净亏损分别为1.09亿元、1790万元和1.00亿元。2024年亏损大幅收窄,但2025年亏损同比扩大459%。

亏损的主要原因在于公司高速产品生产尚处于爬升阶段、规模效应尚未充分释放;还有高速光模块赛道竞争白热化,厂商大打价格战,压缩了利润空间。

中国AI光模块行业的

头部效应明显

当前中国AI光模块行业呈现出市场加速向头部集中的趋势。随着技术门槛不断拉高,A股20家光模块核心上市公司业绩出现明显分化,龙头新易盛以187.29%的营收增速领跑,行业前三名市场份额合计占比高达61.93%,但部分缺乏800G/1.6T高速产品交付能力的企业则面临严峻生存考验。

2025年是AI光模块的800G产品主导市场的绝对红利年,也是1.6T高速光模块开启商用的关键分水岭,预计2026年全球800G和1.6T高速光模块市场规模将达146亿美元,下一代竞争将围绕3.2T预研、硅光集成、CPO、LPO等前沿技术展开。

中国AI光模块行业的竞争格局,可以从分为“第三方供应商”与“内部制造商”两大企业定位。

目前,第三方光模块供应商是市场的主导力量,指主要开发及生产AI光模块用于对外销售的公司。从整体对比来看,第三方供应商以对外销售、服务多家客户为特征,规模体量大、客户结构多元、规模效应强,全球Top10中占据7席,但在高端芯片上依赖外部采购。

2025年,包括中际旭创、新易盛、光迅科技等7家中国光模块厂商进入全球前十,第三方供应商凭借规模效应和客户广度占据了市场主流地位。不过,这部分企业的挑战在于对上游高端芯片的依赖,高端DSP芯片仍由博通、Inphi等海外厂商垄断,英伟达等AI巨头甚至直接“包圆”了大量EML产能,导致中小厂商面临20%—30%的供需缺口。

相较而言,内部制造商企业生产光模块主要供内部使用,服务于自身产品和业务体系。内部制造商规模较小、客户单一,通过自研部分解决供应链风险,追求与自有系统的深度协同。

华为海思、中兴光电子等内部制造商的核心优势在于垂直整合与协同效应,芯片与模块的协同设计可以实现更高效率、更低功耗和更好可靠性,同时保障供应链安全。但其挑战在于缺乏外部市场竞争压力,可能导致创新动力不足;同时由于产品主要对内供应,难以通过规模效应摊薄研发和制造成本,且体量相对较小,未独立进入全球Top10榜单。

无论是第三方供应商,比如光迅科技的“芯片—器件—模块”自主体系;还是内部制造商如华为海思的芯片与模块协同设计,拥有上游核心器件自研能力的企业正获得越来越明显的竞争优势。

从全球竞争格局看,目前AI光互连领域真正的核心竞争已经逐渐从“谁能组装光模块”转向“谁掌握硅光芯片和光子集成平台”。而这恰恰是海光芯正过去十多年持续投入的方向。随着800G向1.6T升级,以及共封装光学(CPO)逐步落地,硅光能力的重要性还会进一步提升。

注:文/漫地,文章来源:创业邦(公众号ID:ichuangyebang ),本文为作者独立观点,不代表亿邦动力立场。

【版权提示】亿邦动力倡导尊重与保护知识产权。未经许可,任何人不得复制、转载、或以其他方式使用本网站的内容。如发现本站文章存在版权问题,烦请提供版权疑问、身份证明、版权证明、联系方式等发邮件至run@ebrun.com,我们将及时沟通与处理。

已点赞

+1

{{ praise_text }}

好文章,值得被更多人看到

广告

下一篇

跨境电商资源对接

《全球跨境电商服务商资源手册》第二期手册现已发布,触达更多资源,了解跨境服务商信息,欢迎点击链接下载。

了解详情

“特供”“治病养生”都不能说了!拼多多将加大酒类抽检力度

主播不得口头承诺酒品治病、养生等医疗功效,不得口头提及特供、专供等违规词汇。

电商平台 拼多多 未来零售

2026-06-30 10:16:18