[Summary]
One of the hottest topics in Chinese AI-related stocks right now is optical modules.
At the center of this are three companies locally known as ``Yizhongten''.
| Stock | Ticker | Main position |
|---|---|---|
| Eoptolink | 300502.SZ | Growth stock for high-speed optical modules |
| Innolight | 300308.SZ | Largest player in optical modules for AI data centers |
| TFC Communication | 300394.SZ | Upstream company in optical components and precision optical devices |
The reason why the three companies are being bought is because they are starting to be seen as companies that hold the bottleneck in AI data centers, rather than just ``Chinese communications parts manufacturers.''
As the number of GPUs increases, the amount of communication between GPUs, racks, and data centers also increases. The AI infrastructure market is not limited to GPUs alone. Without high-speed optical connectivity at 800G, 1.6T, and beyond, large-scale AI clusters will not run.
This is the essence of why the stock prices of the three companies are soaring.
However, Japanese investors need to be very calm when looking at this topic. US-China decoupling, volatility unique to A-shares, price competition after 1.6T mass production, technology shift to CPO and silicon photonics. It's probably dangerous to just read "AI pickaxe stock" without looking at this area.
Why optical modules became an AI core stock
In AI data centers, it is not enough to look at GPU performance alone.
H100, H200, Blackwell, and the further you go to next-generation GPUs, the more computing power you get. However, in order to make full use of that computing power, it is necessary to exchange large amounts of data between servers with low latency.
What is needed is a high-speed optical module.
GPUKey point
↓
AIKey point
↓
Key point
↓
800G / 1.6TupsideKey point
↓
upsideKey pointLooking at NVIDIA's financial results, AI infrastructure investment has not stopped yet. In fact, the importance of the network side has increased even more with the Blackwell generation.
The reason why the market values Yi Zhongtian so highly is because this structure is visible. The main character of AI semiconductors is the GPU, and the optical module is like the blood vessel that connects the GPU. It may sound like a bit of an exaggeration, but in today's AI data centers, ``connection'' is as much a market theme as ``computation.''
The performance of the three companies is already quite strong
His numbers are also strong.
Looking at the January-March period of 2026, all three companies continue to see high growth due to demand for AI data centers.
| Stock | 2026 Q1 Sales | 2026 Q1 Net Income | Market View |
|---|---|---|---|
| Middle Ages of the Year | 19.496 billion yuan | 5.735 billion yuan | 800G/1.6T favorite. Exceptional scale and profit margin |
| Xinyisheng | 8.338 billion yuan | 2.780 billion yuan | Growth rate is high and expectations are high for overseas data centers |
| Tianfu News | 1.33 billion yuan | 492 million yuan | With upstream parts, you can become a winning module manufacturer overall |
What I would like to focus on here is the profit margin.
In a normal manufacturing industry, even if sales increase, profit margins will not rise that much. However, high-speed optical modules are technically more difficult to manufacture as advanced products such as 800G and 1.6T, making it easier to improve the product mix.
In Q1 2026, Chuchu Kyokudan's sales will increase by 192% and net profit will increase by 262% compared to the same period last year. New Yisheng also increased its sales to 24.842 billion yuan and net profit to 9.532 billion yuan in 2025. Although Tianfu Communication is small in scale, it maintains a high profit structure with sales of 5.163 billion yuan and net profit of 2.017 billion yuan for the full year of 2025.
Looking at these numbers, it is natural for the market to evaluate it as an "AI infrastructure stock" rather than a "components stock."
Why is the market premium so high?
There are three main reasons why Yizhongten commands a premium.
The first is the communication bottleneck of the AI cluster.
For generative AI learning and inference, simply lining up a large number of GPUs is not enough. Network bandwidth, power consumption, latency, and heat generation are constraints for large clusters. Optical modules with a capacity of 800G or higher will be a core component for alleviating this constraint.
Second, final demand is not just domestic demand in China.
The growth of the three companies cannot be explained solely by telecommunications investment within China. Investments in North American cloud, AI servers, and GPU supply chains are having a major impact. Although the extent to which companies directly disclose customer names differs from company to company, the market is quite aware that the final demand is strongly tied to the AI CAPEX of US big tech.
The third is high profit margin.
Advanced optical modules are currently not simple assemblies. Because optical design, thermal design, yield, parts procurement, and customer certification are involved, there are a limited number of companies that can mass produce it. Therefore, at least for now, supply capacity is being evaluated more than price competition.
This is stimulating investor sentiment for A shares.
"Even though it's AI, it's a Chinese stock, and it's really showing results." This combination is strong. Because it is so strong, it also reverses quickly.
2027 Forecast: Bullish scenarios factor in far into the future
Market discussions heading into 2027 are quite bullish.
However, I want to be careful here. Forecasts for 2027 vary widely depending on the securities company, and are more like a ``tug-of-war of bullish scenarios'' than a consensus.
| Stocks | Issues the market is looking at in 2027 | Views |
|---|---|---|
| Middle Ages | Can net profit be maintained at over 20 billion to 25 billion yuan? | The annualized rate for Q1 2026 is already at a high level. Focus on 1.6T quantity and unit price |
| Xinyisheng | Opinions are divided on net profit ranging from 20 billion to 40 billion yuan | The old range of 6 billion to 8 billion yuan is too conservative compared to the 2025 result of 9.532 billion yuan |
| Tianfu News | Net profit expected to increase to 4 billion to 5 billion yuan | Winner-distributed position in upstream parts from the module body |
In particular, we need to update our perspective on Xinyisheng.
The figure of ``net profit of 6 billion to 8 billion yuan in 2027'' might have worked based on past assumptions. However, the company made a net profit of 9.532 billion yuan for the full year of 2025, and 2.780 billion yuan in Q1 of 2026 alone. What the current market is pricing in is a much larger profit stage.
However, it is dangerous to draw a straight line based on bullish forecasts.
If mass production of 1.6T increases from 2026 to 2027, sales and profits will increase. On the other hand, as mass production advances, competition will also intensify. Price declines, yields, material shortages, delays in customer certification, currency exchange, and customs duties. Any one of these will destroy the profit margin assumption.
Optical module stocks have big dreams. Therefore, the expected numbers will be higher. It's better not to just swallow the heat of the market here.
Three risks that Japanese investors should be aware of
US-China decoupling
The biggest risk remains the US-China relationship.
Yizhongtian's appeal lies in its ability to benefit from North American AI investment. However, it also means that they are subject to US regulatory risks.
There is a non-zero possibility that regulations will expand not only to high-performance GPUs but also to communications components for AI data centers. To what extent can Chinese-owned companies be involved in U.S. AI infrastructure? This line can change quickly depending on politics.
Each company is proceeding with the development of Southeast Asian production and global supply networks. Tianfu News has set up a dual headquarters system in Suzhou and Singapore, and production bases in Jiangxi and Thailand. However, relocating the factory does not completely eliminate the risk of origin of capital and technology.
Technology shift and commoditization
The second factor is the speed of technological change.
Currently, 800G and 1.6T are the main focus, but next will be new themes such as 3.2T, CPO, silicon photonics, and optical I/O. It is likely that the current pluggable optical modules will continue to play a leading role for some time, but not forever.
Furthermore, as mass production progresses, price competition will emerge. In industries where Chinese manufacturers are strong, it is not uncommon for profit margins to drop the moment supply capacity increases.
In this theme, maintaining not only the growth rate but also the gross profit margin is quite important.
Valuation and price movement of A shares
The third factor is price movements in the A-share growth market.
China A-share startup boards are bought all at once when expectations are high. On the other hand, if expectations are seen to be even a little too high, even a good financial result will be sold.
Yi Zhongtian is close to exactly that situation.
Performance is strong. The theme is also strong. But the stock price is already looking ahead to profits. Stocks that have been bought in anticipation of 2027 profits will generally fall sharply if the gross profit margin or order intake comments are even slightly weak in the quarterly results.
If you approach it with the mindset of Japanese or US stocks, you'll be surprised at the price range.
Investment strategy
Yizhongtian is quite attractive as a "pickaxe stock" in the AI gold rush.
However, it is quite difficult to buy.
In the short term, it is easy to react to NVIDIA financial results, North American cloud CAPEX plans, 800G/1.6T shipping comments, and US-China regulatory headlines. Even if the financial results are good, if the stock price moves ahead of the rest, you will run out of material.
In the medium term, there are four points to look at:
| Check items | How to view |
|---|---|
| 1.6T mass production progress | Center of profit growth from 2026 to 2027 |
| Gross profit margin | See if price competition has started |
| Overseas production | How much can we reduce the US-China risk |
| North American AI CAPEX | Main focus on whether final demand will continue |
Personally, I think it would be better to separate the roles rather than viewing the three companies in the same way.
Chuchu Kyokudan is a favorite large-cap stock. Shinyisheng is a growth stock with high profit elasticity. Tianfu Tsushin is a brand that rides on the entire theme with upstream parts. Understanding this difference will change the way you position yourself.
The worst thing to do is to jump after a sudden rise, thinking that it's still going up because it's AI. Yizhongtian is a strong theme, and many investors are already thinking the same thing.
Summary
China's three optical module companies, Yizhongtian, are in a truly important position in the AI data center market.
The stronger the demand for GPUs, the faster the network will need to be. The spread of 800G and 1.6T is strongly tied to AI investment by NVIDIA and North American cloud.
The achievements of Zhongchu Xuan, Xin Yisheng, and Tianfu Communications are already strong. So it's only natural that the market values it highly.
However, the forecast for 2027 takes into account quite far into the future. US-China decoupling, technology shift, price competition, and overheating of A-shares. These four things should always be kept in mind.
Yizhongtian is a very attractive group of stocks in the AI infrastructure market. At the same time, it is also a group of stocks for which expectations are too high.
If you're going to invest, I think it's more practical to check the 1.6T's mass production progress, gross profit margin, overseas production, and North American CAPEX, and carefully pick up on the market when it's on the decline, rather than going crazy right after the financial results.
Reference information
- TrendForce “Global AI Optical Transceiver Market to Reach US$26 Billion in 2026” https://www.trendforce.com/presscenter/news/20260420-13017.html
- TrendForce "800G and above optical transceiver modules share to exceed 60% by 2026" https://www.trendforce.com/presscenter/news/20260210-12919.html
- Securities Times ``Central Ascendant's profit increases by 262% in one season'' https://www.stcn.com/article/detail/3756603.html
- Ge Longhui "New Yi Sheng 2025 Year Report and 2026 Seasonal Report" https://www.gelonghui.com/news/5229259
- Sina Buying Co., Ltd. “Tianfu News 2026 Seasonal Report” https://finance.sina.com.cn/jjxw/2026-04-21/doc-inhvfivp0263832.shtml