Introduction: The AI optical communication supply chain, which was the most sought after by capital over the past year, has collectively suffered a severe plunge.
On October 8, the optical chip sector suddenly faced a stress test.
Yuanjie Technology and Everbright Photonics both hit the 20% daily limit down, Dongshan Precision hit the daily limit down, and a batch of optical communication supply chain companies including Henan Shijia Photons Technology, Tengjing Technology, and Dekeli saw significant plunges. The Wind Optical Chip Concept Index fell by 8.18% on that day.
Among them, the one receiving the most attention remains Yuanjie Technology.
This optical chip leader, whose share price once approached CNY 2,000 and ranked among the most expensive stocks on the A-share market, closed at CNY 1,290.40 on October 8, dropping 20% in a single day.
What is even more intriguing is that the limit-down occurred when the company's fundamentals appeared to be at their best.
In the first half of 2026, Yuanjie Technology achieved an operating revenue of CNY 925 million, a year-on-year increase of 351.40%; net profit attributable to shareholders of the parent company reached CNY 607 million, a year-on-year increase of 1,212.20%; and net profit attributable to shareholders of the parent company after deducting non-recurring gains and losses was CNY 514 million, a year-on-year increase of 1,033.70%. Revenue from data center business reached CNY 774 million, a year-on-year increase of 640.29%, while the company's overall gross profit margin rose to 80.42%.
Why would a company with profits growing more than 12 times in half a year and a gross profit margin exceeding 80% suddenly hit a 20% limit down?
On the surface, the answer points to the FCC.
During the National Day holiday, overseas institutions discussed the regulatory path by which the U.S. Federal Communications Commission (FCC) might further restrict Chinese-made optical modules, triggering market concerns about the overseas business of China's optical communication supply chain.
However, if we separate the policy text from market rumors, we will find that things are not as simple as the market performance suggests.
The FCC does indeed have a formal rule about to take effect.
Its Third Report and Order passed in July attempts to plug the so-called "component loophole" in the equipment authorization system: if a device contains hardware components with logical functions produced by entities on the FCC Covered List, the FCC authorization for new devices may be restricted under specific conditions. The relevant rules will officially take effect on October 13.
But this is not a "3.2T ban" targeting Chinese optical modules.
The widely circulated claims in the market that "restrictions will start from 3.2T products" and "existing 800G and 1.6T products will not be affected for now" mainly come from deductions by overseas institutions, not from the officially released rule text by the FCC.
Overseas institutions judge that if further restrictions on Chinese-made high-speed optical modules are implemented in the future, they are more likely to be executed in phases starting with the next-generation 3.2T optical transceivers, while mainstream 800G and 1.6T products at this stage will not be directly affected for the time being.
In other words, what the market was trading on October 8 was not an already occurred business loss, but a policy risk that might occur in a few years.
This is also the most debatable aspect of this limit-down.
What AI optical chips are truly beginning to be graded on may not be their performance, but their valuation.
Risks Shift from Complete Systems to Upstream
An interesting aspect of the market is that if the so-called FCC risk targets China's optical communication supply chain, then theoretically, those most directly exposed to the U.S. data center market should first be the optical module system manufacturers.
However, on October 8, the most intense venting of sentiment occurred in upstream optical chips.
Yuanjie Technology and Everbright Photonics hit the limit down, while the overall reaction of optical module leaders was relatively mild.
One of the reasons is that this is not the first time optical module companies have faced this policy expectation.
In early August this year, news already emerged that "the U.S. might restrict imports of new Chinese optical modules."
At that time, leaders such as Zhongji Xuchuang, Xinyisheng, and Tianfu Communication saw significant intraday declines, which were quickly narrowed.
On August 5, the A-share optical module sector ultimately even rose by 3.98%. Listed companies such as Zhongji Xuchuang publicly responded at the time that the U.S. FCC had not yet issued restrictive documents in this field, and the company's operations were normal.
That time, it was equivalent to the optical module leaders having already undergone a round of policy risk pricing.
By October, the risk narrative had changed.
Overseas institutions began to further deduce: even if Chinese companies are not directly banned from producing optical modules in the future, the proportion of U.S. domestic chips and components in next-generation high-speed optical modules might be increased through methods such as the localization ratio of the Bill of Materials (BOM).
What is truly directly impacted by this assumption shifts from "who produces the optical modules" to "whose chips are actually installed inside the modules."
Risks begin to migrate from downstream complete systems to upstream value.
This is precisely the reason why optical chips were collectively sold off by the market.
Some institutions have assessed that the potential regulatory framework may drive the use of more products from U.S. suppliers for high-value components such as DSP, PD, EML, and CW Laser. Although this still belongs to policy deduction at present, for the market, this exactly hits the most important valuation logic for domestic optical chips previously—moving from domestic substitution to global substitution.
Over the past two years, the capital market's imagination for domestic optical chips has gone beyond "replacing overseas products in the domestic market."
A more enticing story is that, along with the upgrade of 800G, 1.6T, and future 3.2T optical modules, Chinese optical chip manufacturers have the opportunity to follow domestic optical module leaders into the supply chains of North American cloud computing giants.
The former determines the domestic market space, while the latter determines the ceiling of the global market. A significant part of the high valuation given by the market to companies like Yuanjie Technology is precisely built on the second possibility.
Therefore, even if the potential rules may not truly have an impact until the 3.2T era, and even if existing orders for 800G and 1.6T are not constrained in the short term, stock prices will still adjust first.
The stock market has never traded on this year's profits; it trades on the discounted value of future profits, especially for a stock that has already risen to over CNY 1,000.
Shanghai Securities News, citing investor communication information from Zhongji Xuchuang, reported that the true scale volume of 3.2T optical modules is expected to be achieved around 2028 to 2029, and from 2026 to 2028, the mainstream products in the industry will still be 800G and 1.6T.
In other words, even according to the policy path circulating in the market, what is affected is a rather distant market.
But for high-valuation growth stocks, three years is never considered distant. Because today's market capitalization is inherently supported by profits three years from now.
A Single News Item Pierces Sector Sentiment
If it were only the FCC rumors, Yuanjie Technology might not necessarily have to hit the limit down.
On October 8, another piece of news fermented simultaneously: optical chips might see price cuts.
It was once rumored in the market that prices for optical chips supporting 1.6T were already under pressure. With the ramp-up of silicon photonics solutions, supply improvements, and changes in customer procurement strategies, the high premiums previously seen for some high-end optical chips might begin to loosen.
But just like the FCC 3.2T rumors, this news currently also lacks confirmation from listed companies.
Everbright Photonics explicitly stated that it had not received any notice of price cuts for optical chips; Henan Shijia Photons Technology and Etern Company Limited also stated that they were not aware of any such situation. Henan Shijia Photons Technology simultaneously judged that this stock price adjustment might be more related to the market's discussion of FCC policy during the National Day holiday.
Therefore, at least as of October 8, the evidence is insufficient to explain this plunge as "1.6T optical chips have seen comprehensive price cuts."
But the capital market remains sensitive.
Because the price cut rumors touch upon another extremely important variable in the valuation of Yuanjie Technology: how long can the 80% gross profit margin be maintained?
In the first half of 2026, the overall gross profit margin of Yuanjie Technology reached 80.42%. Among them, the gross profit margin of the company's data center business even reached about 84%.
For a manufacturing company, this is an almost luxurious level of profitability.
It means that at least at this stage, high-end optical chips are still in a very obvious seller's market.
Demand is growing rapidly, while the number of suppliers capable of stably providing high-power, high-speed laser chips is limited, giving the upstream strong bargaining power.
This is also the core reason why optical chips have become one of the most fiercely rising segments in the entire AI hardware supply chain over the past year.
GPUs are very expensive, optical modules are in short supply, and optical chips are even more scarce.
The more scarce a segment in the supply chain, the higher the profit margin.
Capital thus gave the highest valuation to the most scarce segment.
But all high-gross-margin manufacturing industries cannot avoid a problem: high profits themselves create competition.
When a product has an 80% gross profit margin, industrial capital cannot remain indifferent for a long time.
Existing companies will expand production.
Competitors will develop new products.
Downstream customers will proactively cultivate second and third suppliers.
Manufacturing yield rates will improve.
Unit costs will decline.
Ultimately, the supply-demand relationship will gradually shift from "severe shortage" to "tight balance," and then to mature supply.
The most scarce product last year is not necessarily scarce forever.
This is also why, even without a real price cut notice, a price rumor can still quickly pierce sector sentiment.
What investors are truly worried about is not "how much cheaper it is today," but rather they begin to doubt:
Is the 80% gross profit margin truly a long-term moat, or just a phased dividend in the early stage of AI infrastructure expansion?
The two answers have completely different implications for valuation.
If Yuanjie Technology's profitability mainly comes from technical barriers, then with the upgrades to 1.6T and 3.2T, the company may still be able to maintain high profit margins by continuously enhancing product performance.
But if the current ultra-high gross profit margin contains a large price dividend brought by supply falling short of demand, then along with production expansion and intensified competition, the profit margin will ultimately return to normal levels.
Yuanjie Technology is expanding production, and so are other companies in the industry.
This is precisely the most difficult question for the capital market: the continued growth in AI optical communication demand does not mean that optical chip prices will rise forever.
Market space expansion and product price declines can completely happen simultaneously.
The history of the semiconductor industry over the past few decades has been exactly like this.
With continuously growing demand, ongoing technological upgrades, and continuously declining unit prices, ultimately only those companies that can continuously reduce costs through process, yield, product iteration, and customer stickiness will truly retain profits.
Therefore, after October 8, the market may no longer be satisfied with asking: how much 1.6T is shipped?
But will further ask: after the massive shipment of 1.6T, how much can actually be earned?
A Divergence Initiated by the Limit-Down
It is equally too simplistic to interpret Yuanjie Technology's limit-down as the end of the AI optical communication industry logic.
At least for now, no data supports this judgment.
In the first half of 2026, Yuanjie Technology's data center business revenue reached CNY 774 million, a year-on-year increase of 640.29%; 100G PAM4 EML products have achieved batch delivery, 200G PAM4 EML is advancing customer verification, and high-power CW products continue to scale up.
Everbright Photonics achieved an operating revenue of CNY 391 million in the first half of the year, a year-on-year increase of 82.67%, and net profit attributable to shareholders of the parent company of CNY 30.3378 million, a year-on-year increase of 238.04%. The company attributes an important reason for the performance growth to the rapid growth of the optical communication business previously laid out.
From the perspective of industry prosperity, at the China International Optoelectronic Exposition held in September this year, multiple supply chain companies still told Securities Times that orders for high-speed optical modules are strong, 800G has become the main product, 1.6T is in the stage of sample delivery and batch introduction, and an important factor currently restricting industry expansion is even still the insufficient capacity of some core components.
Therefore, October 8 did not prove that AI demand has disappeared, nor did it prove that 800G or 1.6T suddenly stopped selling; even the 3.2T policy risk that the market worries about most is currently still at the stage of deduction by overseas institutions.
This is also the most interesting part of this plunge: the fundamentals did not change drastically within a day, yet the valuation changed.
From this perspective, Yuanjie Technology's limit-down is more like a valuation switch.
Previously, when the market priced optical chip companies, the three most important words were: shortage, upgrade, and domestic substitution.
As long as AI capital expenditure continues to expand, 800G iterates to 1.6T, and the localization rate continues to increase, investors will be willing to continuously push back earnings forecasts.
The market space in the next three years is huge, so today's market capitalization can also be huge.
The core of this trading method is calculating the space.
But when a company's profits truly begin to materialize, and the materialization speed is as fast as a 12-fold increase in half a year, investors will instead enter the next stage: calculating profits.
How long can profits grow?
How much can the gross profit margin be maintained?
Can they still enter the North American supply chain in the 3.2T era?
Will customers demand price cuts?
Will new capacity bring price competition?
After the penetration of silicon photonics, how exactly will the value of different chips change?
Every question will directly change the final valuation.
This means that the optical chip sector is undergoing a transition from "industry Beta" to "company Alpha."
In the past, as long as they belonged to the AI optical communication supply chain, valuations could be raised together.
It may become increasingly difficult in the future.
Both Yuanjie Technology and Everbright Photonics are called optical chip companies, but the profitability stages of the two enterprises are already completely different.
In the first half of the year, Yuanjie Technology's net profit attributable to shareholders of the parent company was CNY 607 million, net profit after deducting non-recurring gains and losses was CNY 514 million, and net cash flow from operating activities was CNY 488 million; profitability and cash flow have been materialized on a large scale.
Everbright Photonics' net profit attributable to shareholders of the parent company in the same period was CNY 30.3378 million, but net profit after deducting non-recurring gains and losses still recorded a loss of CNY 10.3844 million.
In the past, the market could cover both companies simultaneously with the four words "domestic optical chips."
This may not be possible in the future.
Entering 1.6T equally, some companies sell EML, while others sell CW Laser;
Moving towards silicon photonics equally, the value of some segments increases, while the value of some components decreases; facing 3.2T equally, some enterprises have already entered the overseas customer system, while others are still in the verification stage; even with high revenue growth equally, some come from true technical barriers, while others may just come from short-term industry shortages.
The more mature the industry, the greater this difference.
Thus, the truly important question in optical chip investment is changing: from "how big is the market exactly" to "who exactly do the profits belong to."
This is also the key to understanding the market performance on October 8.
Capital No Longer Just Buys "Light"
The FCC risk circulating in the market is currently more of a long-term policy game.
The so-called restrictions starting from 3.2T do not appear in the FCC's existing formal rule text; nor have 800G and 1.6T suddenly lost the market due to the aforementioned policy deduction.
Industrial demand has not suddenly reversed either.
What has truly changed is the risk appetite of capital.
Optical module companies have already experienced a round of policy rumor shocks in August; this time, the market reserved the most intense sentiment for optical chips, which had larger previous gains, higher valuations, and were simultaneously at the center of the domestic substitution narrative.
This seems like policy trading.
Essentially, however, it is more like a valuation stress test.
During the previous rally phase, the market was willing to believe three things simultaneously: AI investment will continue to grow, the share of domestic optical chips will continue to increase, and the scarcity of high-end optical chips will persist.
After October 8, these three things have not been falsified.
But for the first time, investors are beginning to question their boundaries so intensively.
If supply chain localization requirements really emerge for 3.2T in the future, how many overseas orders can domestic chips still get?
If capacity is released, how long can the 80% gross profit margin be maintained?
If technologies such as silicon photonics and CPO iterate rapidly, will the most scarce products now still be scarce three years later?
These questions could be left to the future in the past.
They cannot be now.
Because Yuanjie Technology is already a company with a market capitalization of CNY 100 billion.
When a company's half-year profit grows from tens of millions to CNY 600 million, the capital market will no longer merely reward it for "entering the AI track."
In the next stage, what it needs to prove is: this CNY 600 million is not an accident, the 80% gross profit margin is not due to a temporary shortage, there is 3.2T after 1.6T, there is an overseas market beyond the domestic market, and there is next year's growth after this year's growth.
This is the truly difficult part of a CNY 1,000 stock.
When an industry goes from 0 to 1, the capital market buys imagination.
When it goes from 1 to 10, it buys penetration rate.
After reaching 10, it ultimately still has to return to profits, cash flow, and the competitive landscape.
On October 8, the AI optical communication industry did not stall.
It is just that the capital market took out its calculator so seriously for the first time.
Over the past year, capital was buying "light."
From now on, it may have to start picking companies within the "light."
The valuation of AI optical chips has begun to be graded.