Hardly any institutions are not paying attention to optical chips.

"You should stand in the light, not just let the light stand there." This joking remark has become a truth for investors in the secondary market.

Looking at the primary market, optical chips have also become an inescapable topic this year. At the recent China Venture Capital Annual Conference, several investors mentioned that they have been paying attention to optical chip projects recently. Some, although having made early layouts, still chose to invest again this year.

A chip investor also told me that he invested in an angel - round project even before the project was registered. "There's no way. It's so competitive now. By the time it's registered, the valuation will be much higher."

This is not an isolated case. An investor from a CVC - oriented firm told me that they invested in an optical chip project two years ago. Due to the high - level confidentiality of the project, their investment information is still not available in the public market. "Projects like this make money quietly. They don't worry about financing or orders."

On a micro - level, in March and April this year, I noticed that some headhunters started recruiting talents in the field of optical chips, and the salaries were quite flexible.

Upturn

On the surface, in the field of optical chips, there is a lively scene where the primary and secondary markets are in sync. However, in fact, compared with the secondary market, the consensus in the primary market to pursue "light" only started after the Spring Festival this year, which is only about three months so far.

An early investor in Lightelligence said, "When we invested in it in 2020, it wasn't popular. The threshold was high, and the commercial scenarios were not clear. There was no large - scale model at that time, and people didn't know how to use it, where to use it, and how valuable it was. It was quite difficult to raise funds during that period."

J, an early entrepreneur in the optical chip field, recalled that he switched from AI chips to optical chips in 2024. At that time, the primary market's acceptance of this concept was not high. "Many investors only focus on the present, at most on tomorrow or the day after. Optical chips were still too early at that time, so we faced a lot of rejections when raising funds."

Indeed, 2024 was the time when large - scale models were attracting a lot of capital. Even in the chip field, GPUs were much more popular than optical chips. Moreover, there was no so - called "Yizhongtian" theory in the secondary market at that time.

But just two years later, the entrepreneur felt a different situation. "No exaggeration, in the past month, three groups of investors have approached me to start a business. " In addition, his company has been raising funds since the beginning of the year. Its valuation has quadrupled in just two months. Even though they have been very cautious in financing, they have met with at least thirty institutions.

Based on the views of multiple practitioners and investors, a relatively consistent view is that the popularity of the optical concept in the primary market this year is driven by demand, concepts, and IPOs.

Since the beginning of the year, an industry consensus has become stronger. For the computing power system, the era of competing for single - computing power is over. In large - scale model computing power clusters with tens of thousands or hundreds of thousands of cards, more than 90% of the energy consumption is used for data transfer rather than calculation itself.

At the same time, artificial intelligence has elevated data transmission to a higher level. The existing infrastructure cannot meet the requirements for computing, transmitting, and storing data in the order of hundreds of billions. All infrastructure needs to be upgraded. If the interconnection efficiency is not high enough, the cluster's capabilities cannot be fully utilized. So, it has become a new bottleneck to be solved.

Look, the largest IPO in the US stock market that just emerged provides a solution to this bottleneck - concentrating performance on a single wafer to reduce data transfer. However, due to the high difficulty of this approach, few people dare to get involved.

Currently, the mainstream approach in the market is to start from changing the way of computing power transmission. Light is an important solution. The popular CPO (Co - Packaged Optics) can reduce interconnection power consumption by about 70% and increase bandwidth density by 10 times.

Of course, the most direct triggering factor comes from Jensen Huang. He pointed out that to meet the exponentially growing computing demand, optical products must be expanded on an unprecedented scale.

Therefore, at this year's GTC conference, NVIDIA clearly stated that the next - generation AI infrastructure must fully shift to optical connections. The next - generation AI - native computing power architecture, Feynman, will also deeply integrate optical communication technology. This move marks that the inter - chip interconnection will completely shift from traditional electrical signals to optical signals, opening the era of true optical interconnection.

On this basis, NVIDIA also announced an investment of $2 billion each in Lumentum and Coherent, two global giants in optical chips and optical communication components.

There has always been a joking saying in the primary market: "Elon Musk and Jensen Huang point out the direction for Chinese investment."

So, NVIDIA's full convergence on the CPO technology route, the high - profile display of the CPO switch system at GTC/OFC, and the huge investment in the supply chain have directly ignited the primary market's crazy pursuit of "light".

This enthusiasm reached a new high after the listing of Lightelligence.

Many investors said that Lightelligence's listing was a very inspiring event. As the world's first AI optical computing power stock, its performance after listing far exceeded expectations, playing a positive role in the market. Correspondingly, a number of optical computing companies have emerged.

A detail is that in the US optical chip field, there is a project called Light Matter that is comparable to Lightelligence. Previously, its valuation was much higher, but after Lightelligence's listing, the valuation gap between the two companies has been narrowed.

The deeper reason is that, in addition to cutting - edge fields such as large - scale models, commercial spaceflight, and quantum, another track with a certain chance of listing has emerged. At the same time, the stock prices of light - related companies in the domestic secondary market have also risen significantly. For example, the market value of Zhongji Innolight in Suzhou has exceeded one trillion. Perhaps even its founder didn't expect such a large increase this year.

This shows that optical chips not only have an exit strategy but also can achieve a good exit.

Two Paths

So, in this previously quiet track, money, people, and speculators who smell the opportunity have all come.

Investor A said bluntly that many people in China didn't know what CPO was last year. But just one year later, almost all institutions are starting to invest in "light" projects.

Investor B also said, "Now investment institutions don't look at orders or performance. They only look at a company's position in the industry and invest based on that because everyone knows this is a definite opportunity."

What many people didn't expect is that there aren't many projects in the market. According to China Venture Capital, as of April this year, there are less than a hundred core manufacturing and design enterprises in the narrow sense of optical chips. The first one is Accelink Technologies from Wuhan, which originated from the Solid - State Device Research Institute of the Ministry of Posts and Telecommunications in 1976 and was listed on the A - share market in 2009. Its revenue exceeded 10 billion last year, and its market value exceeded 100 billion in April this year. It is one of the "Seven Stars in Optics Valley".

The core reason for this phenomenon is that the threshold in this field is too high. Light is a relatively niche and professional field, and the domestic research foundation is not deep. "Different from large - scale models, where you can get involved with a little mathematical foundation and some relevant theories, and there are many schools and research institutions related to large - scale models, covering a wide range."

Under the imbalance between supply and demand, the choices of institutions have also diverged. Some prefer to invest in existing projects in the market, but this means accepting high valuations and high premiums.

In this regard, Fei Jianjiang, the general manager of Yuanhe Origin, gave two "unexpected" examples. The first is an optical chip company based on next - generation new materials. The team used to do research at the University of Hong Kong and was far from commercialization. In the early stage, it mainly had R & D needs. "We were already prepared for a long - term battle, but unexpectedly, it became popular this year, and the valuation soared to more than one billion. Many institutions came to inquire, which was a pleasant surprise."

The other case is more dramatic. A company that manufactures optical chip devices had a valuation of just over one hundred million at the end of last year, but this year, more than a dozen institutions flocked to it, and the valuation directly soared to nine hundred million. In fact, it has not yet commercialized, and its business is uncertain. Many investors are simply attracted by the concept and the track.

Some people are not willing to pay for high valuations and choose to "organize a project" themselves. After all, the valuation of new projects in the angel round can reach four or five hundred million. This approach may seem speculative, but it requires higher requirements for investors because it needs investors with real industrial resources and industry insights to take the lead.

For example, the optical interconnection company, Optolink, which we wrote about before, was incubated by Zhenzhi Venture Capital. Its founder, Chen Chao, also from Zhenzhi Venture Capital, holds dual master's degrees from MIT and Duke University and has the same teacher as the founder of Ayar Labs in the United States.

An investor who has designed a similar project told me that even "organized" projects do not mean that other investment institutions cannot participate. Different from other tracks, the optical chip field cannot be directly accelerated by capital. The engineering and industrialization of the entire optical chip field require a long - term process, and funds at different stages should work together.

Timing

Although many institutions choose to enter the market now, some institutions have already started to make layouts in this field, such as Sequoia Capital, Hillhouse Capital, Yaotu Capital, Zhongke Chuangxing, and Yuanhe Origin. And they have enjoyed the dividends of this wave in advance.

As for whether it is a good time to invest in optical chips this year, different institutions have different views.

Liu Haofei, the founding partner of Shengjing Jiacheng, believes that some investment institutions started to make layouts in optical chips a few years ago. However, so far, not all the expectations at that time have been realized. Now, more money is flowing in, and after a certain period of precipitation and selection of technical routes, the probability of success may be higher than before. At the same time, from an objective development perspective, the technical route of optical computing has not yet converged, so there is still a time window for continuous verification or falsification.

Fei Jianjiang said that from an industrial perspective, it is not too late to invest in light - related fields now. However, from an investment timing perspective, the valuations have risen significantly, which is not a very comfortable time for investors. Investing in optical chips now is a bit like investing in GPUs in 2020 and 2021. There is already a certain consensus in the market, but it will still take time for large - scale commercialization. The commercialization of the optical industry is far from the explosive stage, and there is still a lot of room for improvement in the future.

He believes that the listing of Lightelligence this year is similar to the listing of Cambricon back then. Both are industry benchmarks. In terms of commercialization, the products have not really been mass - produced, and the companies are still in the red. In fact, Cambricon didn't have large - scale shipments and achieve its first profit until the first quarter of 2025, and its stock price reached a historical high in 2025. Before the real realization of commercial value, the market also had doubts.

Of course, some people think it's too late. This is because if you start to develop optical chips now, it will take time to verify, tape - out, and complete the whole process, and you may miss this wave of opportunities. This kind of cyclical opportunity is similar to that of memory chips. Once you miss it, it's gone. Currently, there is already a target in the Hong Kong stock market, and there may be one or two more in the A - share market in the future. So, there aren't many opportunities left for new targets.

The investor further revealed that in the optical field, a company that has been established for less than three years already has a valuation of tens of billions. How can newly - started companies catch up?

New projects that have lost the first - mover advantage in terms of timing are all trying to break through by tying up with the industry.

Here are two cases to share.

Liu Haofei mentioned PhotonCore, a project that he is quite optimistic about. It has just completed its angel - round financing. With its unique technology reserve, it has a good foundation in engineering and ecosystem construction and is well - received in the market.

He said that although being a genius teenager is one of the labels of Yang Qisheng, the founder of PhotonCore, what really impressed him was Yang Qisheng's charisma. "During the entire research and development process, he can effectively get the support of different universities - he has collaborated with several universities, not just limited to his own laboratory. People are willing to work on this, which increases the certainty."

Fei Jianjiang also said that it is not difficult for them to get shares in popular projects in the market because entrepreneurs value their resources and layouts in the entire optical industry chain. This ecosystem will be of great help to early - stage projects in terms of industrial collaboration. "In the future, light will replace existing silicon - electrical related things, but ultimately, it will be applied to data computing, transmission, interconnection, etc. Coincidentally, we have made early layouts in these fields. Some of the companies we invested in have been listed on the STAR Market, and some have occupied important positions in the industry."

It can be seen that the ecological card and the industrial chain are becoming the entry tickets for the second - half competition in the optical chip field.

With the entry of capital and the increase in projects, the golden age of optical chips has officially begun. Hundreds of millions in angel - round financing and the scramble for non - commercialized projects, these classic scenarios that have been staged in fields such as large - scale models, embodied AI, and GPUs are also being repeated. For new entrants, the only proposition is how to seize the era's dividends of "light in, copper out" and find the correct path to cross the "valley of death". After all, those who can stand in the light are never speculators chasing the trend, but those who truly master light.

This article is from the WeChat public account "China Venture Capital", author: Zhang Xue, published by 36Kr with authorization.