HomeArticle

The mid-year reports of optical module companies have all delivered explosive stunning performances. How did this group of the most solid enterprises in the AI market boom make their profits?

BT财经2026-07-23 12:38
Optical module enterprises have witnessed explosive performance growth, and the AI narrative has been translated into real profits.

On the evening of July 19, the performance forecast disclosed by Eoptolink, which was only a few pages long, sent a shockwave across the entire market: the company estimates its 2026 semi-annual net profit attributable to shareholders will reach 7 billion to 8 billion yuan, representing a year-on-year increase of 77.56% to 102.93% (Source: Eoptolink Announcement, July 19, 2026).

It is not an isolated case. Yangtze Optical Fibre and Cable previously projected its 2026 first-half net profit attributable to shareholders would rise 711% to 914% year-on-year; Hengtong Optic-Electric forecasts a year-on-year growth of 86.94% to 121.20%; and Yongding Co., Ltd. expects a 57% to 120% year-on-year increase (Source: Listed Company Announcements, China Finance Online, July 21, 2026).

An industry that was previously relatively obscure to the general public — optical modules and optical communications — is now speaking through its income statements. On July 21, the day of the "A-share Miracle Rally," the CPO concept soared, with stocks such as Tongfu Microelectronics, Accelink Technologies, and Shengyi Technology hitting their daily 10% price limit (Source: China Finance Online, July 21, 2026). These companies represent one of the most solid sample groups in the AI market rally: after two years of telling the growth story, the revenue has actually landed in their accounts.

Where Did the Capital Flow In From

To understand the origin of that 7 billion yuan profit, we first need to understand what an optical module is.

In an AI data center, thousands upon thousands of computing power cards do not operate independently. They need to exchange data with each other at extremely high speeds to collaboratively train large models as if they were a single integrated machine. While chips are responsible for computation, the transmission of computing results between server racks relies on light: the optical module is the very "translator" that converts electrical signals into optical signals and then back again, serving as the "nerve fiber" of the AI cluster.

This business has a fascinating characteristic: the larger the scale of the cluster, the more interconnection channels are required between cards, leading to demand for optical modules growing faster than the expansion of computing power cards themselves. Every time overseas cloud vendors build a cluster with ten thousand cards, they simultaneously purchase massive quantities of high-speed optical modules. In the global shipment rankings of optical module manufacturers, Chinese companies have consistently occupied top positions for consecutive years (according to public rankings from multiple industry institutions). This has made optical modules one of the few sectors where "AI narratives directly translate into Chinese companies' income statements": a definite portion of the capital expenditure from companies like NVIDIA flows through a clear pipeline to the revenue of Chinese enterprises.

The evidence chain on the demand side continued to strengthen in July. According to data from the Korea Customs Service, South Korea's chip exports in the first 20 days of July surged 180.6% year-on-year (Source: China Finance Online citing Korea Customs Service, July 21, 2026); on July 16, TSMC raised its full-year 2026 capital expenditure guidance to $60 billion to $64 billion (Source: Securities Times, July 16, 2026); a J.P. Morgan report argues that demand for AI, server upgrades, and high-bandwidth memory will keep the memory market in a state of tight supply and demand extending to 2028 (Source: China Finance Online, July 21, 2026). Upstream players are expanding production, midstream players are scrambling for inventory, and the order visibility of optical modules is reflected in the actions across the entire industrial chain.

Key Data: Eoptolink estimates its 2026 semi-annual net profit will reach 7 billion to 8 billion yuan, a year-on-year increase of 77.56% to 102.93% (Source: Company Announcement, July 19, 2026); Yangtze Optical Fibre and Cable projects a 711% to 914% year-on-year growth, Hengtong Optic-Electric projects 86.94% to 121.20%, and Yongding Co., Ltd. projects 57% to 120% (Source: Listed Company Announcements, China Finance Online, July 21, 2026).

They Also Got Hammered During the Crash Days

A detail that is easy to forget: in the very same week when performance forecasts were being released intensively, these companies had just experienced a indiscriminate market sell-off.

On July 17, the A-share ChiNext Index fell 7.15% in a single day, marking its largest decline of the year, with over 5,000 stocks dropping. Optical communications and semiconductors were also among the hardest-hit sectors (Source: National Business Daily, 21st Century Business Herald, July 17, 2026). When the market is driven by panic selling, it never distinguishes between companies with solid performance and those that only have stories to tell — a chaotic, all-encompassing decline is the inevitable outcome of market panic.

This precisely explains the divergence that occurred on July 21: as the panic receded and capital reaccumulated positions, the first stocks to rebound were exactly those that had been "sold off incorrectly." The performance forecasts served as their proof of strength: with 7 billion yuan in net profit on the books, the valuation gap created by the crash became the first area to be filled when the industrial logic recovered. Between this rise and fall, the market completed a simple screening process: stories and performance, which seemed identical before the crash, became two distinct entities after the crash.

Of course, solid performance does not equal a permanent safety guarantee for stock prices. The optical module industry itself has clear underlying variables: the product iteration window from 800G to 1.6T is opening, and if the co-packaged optics (CPO) technology route accelerates its penetration, the existing market landscape could be reshuffled; the pace of capital expenditure by downstream cloud vendors also directly determines the slope of the order curve.

Takeaway Framework · Three Questions to Evaluate Performance Quality: First, question the source of growth (is it from core business orders or one-time gains?), second, question the demand evidence chain (can upstream and downstream data mutually validate each other?), third, question sustainability variables (technological iteration and the capital expenditure rhythm of major clients). In short, strong performance must be verified within the context of the entire industrial chain — no matter how large an isolated number is, it remains just a number.

The following is a line of reasoning for readers' reference. If overseas cloud vendors maintain their first-half capital expenditure intensity in the second half of 2026, the performance delivery of optical module companies may continue to exceed market expectations; conversely, if the pace of AI infrastructure investment slows down, the current high growth rates will face dual pressures from high base effects and order fluctuations. In addition, discrepancies may still exist between the performance forecast and the official semi-annual report, and the final figures shall be subject to the official financial report. The above is logical deduction and does not constitute any judgment on individual stocks.

What Does This Have to Do With You

First, for investors focused on the AI and technology tracks: optical modules provide an identification sample. Companies that all tell AI stories can have vastly different performances in extreme market conditions depending on whether they have the backing of real income statements. Learning to assess the "performance evidence chain" is more important than chasing after concepts.

Second, for practitioners in the optical communications and electronics manufacturing industrial chain: the doubling of profits at leading companies means the expansion cycle of production capacity, recruitment, and supply chain orders is still ongoing. The prosperity of this industrial chain can be directly perceived through job openings and order volumes.

Third, for general readers: the optical module story is a microcosm of "Chinese manufacturing integrating into the global AI wave." In every round of capital expenditure driven by global technological revolutions, countries and companies that can secure a definite share will eventually convert it into employment, tax revenue, and industrial upgrading — which is closely related to the economic environment everyone lives in.

That few-page announcement on the evening of July 19 is more convincing than all the eloquent speeches about AI over the past two years combined: when the tide rises, everyone is telling stories, but when the tide recedes, only the income statements remain standing. What do you think about this? Feel free to share your views in the comment section. 

This article is for information sharing and industry analysis only, and does not constitute any investment advice, investment analysis opinion, or transaction invitation. The data in the article comes from public sources including Eoptolink's announcement (July 19, 2026), announcements from Yangtze Optical Fibre and Cable, Hengtong Optic-Electric, and Yongding Co., Ltd. (compiled by China Finance Online, July 21, 2026), Securities Times (July 16, 2026), National Business Daily and 21st Century Business Herald (July 17, 2026), and China Finance Online citing Korea Customs Service (July 21, 2026). All data shall be subject to the original sources; performance forecast data shall be subject to each company's official semi-annual report. Institutional views represent only their own positions. The market carries risks, and decisions should be made with caution. Content marked as "reasoning" in the text is logical deduction based on public information and does not represent an official position.

This article is from the WeChat Official Account "BT Business and Tech", authored by Yining, and published with authorization from 36Kr.