Exclusive Interview by Yingke: An Expert from a Former Big Tech Company Developed the World's Only Optical Solution and Has Been a Supplier to Rokid for Years

Author | Zhang Ziyi

Editor | Yuan Silai

Among hardware startup companies, Rokid is already an old - timer. The story of its founder Misa (Zhu Mingming), who was inspired by Google Glass and left Alibaba to found Rokid, took place 11 years ago.

Misa is 49 years old and has witnessed every up and down in the smart glasses industry. In the latest round of financing, Rokid was valued at $1 billion, but its profit situation remains unclear.

But that doesn't matter. In the rumors, Rokid has been on the verge of collapse more than once, but it always seems to survive until the arrival of the next big trend.

This time, Rokid has become one of the "Seven Dragon Balls of Hangzhou", alongside Deepseek and Unitree, and has entered the list of the most - watched technology startup companies in China.

On February 18th, at the High - quality Economic Development Conference in Yuhang District, Hangzhou, Misa stood on the podium wearing a pair of AR + AI glasses and delivered an off - the - cuff speech. "My speech notes are in the glasses, and I turn the pages with the ring on my hand," he told the audience.

Regardless of the display effect, the stock market boiled over it. At the close on February 20th, shares of Boshi Optical, Mingyue Lens, and Jiemeite rose more than 20%. Even indirect shareholders of Rokid hit the daily limit.

Behind this seemingly simple pair of glasses lies a group of low - key hard - tech companies.

36Kr learned that the optical supplier behind Rokid Glasses is Lightship Semiconductor. The company's unique single - optical - engine diffractive waveguide solution is applied to Rokid Glasses. This solution can reduce the power consumption and cost of the product while providing a good display effect.

Lightship Semiconductor mainly provides optical waveguides and optical engines for downstream AR glasses manufacturers. At the same time, Lightship Semiconductor has a complete set of IP and related patents from optical chip design software to product design and manufacturing. Earlier, it received investments from well - known institutions such as Sequoia China Seed Fund, Tencent, and ByteDance. The founder once served as a R & D expert in Huawei's optical devices, and the co - founder is Tapani Levola, the father of diffractive optical waveguides.

As AI glasses become increasingly popular, more and more customers, ranging from listed companies to startups, are approaching Lightship Semiconductor. They have more detailed requirements, as well as more expectations and anxieties.

Currently, the AI glasses supply chain is far from mature. There are significant differences in the solutions provided by different suppliers, and a single mistake can lead to the failure of the product. While companies like Rokid are in the spotlight, the reshuffle in the shadows is extremely brutal.

Therefore, we had a chat with Zhang Xueli, the co - founder and marketing director of Lightship Semiconductor, about the current changes in the industry, the company's R & D achievements, and the new demands of customers.

When Rokid participates in a more significant competition, suppliers like Lightship Semiconductor are the ones providing the ammunition.

"Whether it's AI devices or AI - related optical enterprises, Chinese companies are already far ahead. In the future, the industry will also be in an absolutely leading position globally," Zhang Xueli, the co - founder and marketing director of Lightship Semiconductor, told Yingke.

This generation will witness the rapid rise and growth of new industries and companies, as well as the collapse of some myths. In the long run, companies with strong technological capabilities like Lightship Semiconductor will be the survivors.

The following conversation has been edited by Yingke:

01 AI and AR

Yingke: AI glasses are quite popular these days. What are the technical differences between them and previous AR glasses?

Zhang Xueli: In essence, AI glasses integrate large - scale models. Currently, they mainly interact through voice, along with a camera. In the future, the camera may become a new input method. Most previous AR glasses mainly had two directions. One was to watch video streams and present color images; the other was the single - green direction for information prompts. For example, the cycling AI glasses launched by Liweike in 2022 could provide navigation routes, mobile phone text messages, and call reminders during cycling, and also had a large - scale model for simple chats. Early AR glasses could only be regarded as near - eye displays, and the interaction between virtual and real was not well - reflected.

Now, the popularity of AI glasses is mainly due to the product jointly developed by Ray - Ban and Meta, which has shown the potential of this category. Coupled with the improvement of large - scale model capabilities, there is more room for imagination, so many AI glasses products have emerged in the market. However, I believe that in the next step of AI development, including within Meta, display functions will be added. The combination of AI and AR may be the future product form.

Yingke: Meta's AI glasses have sold well, but it is said that they have suffered large losses and the cost cannot be reduced, making it difficult to form a good profit model in the short term. What do you think?

Zhang Xueli: Currently, AI glasses are mainly used for displaying content, and it is relatively easy to produce such products. However, in the future, a model like Oculus may emerge, where the device itself does not make money, but profits are made through content. This places high requirements on the interactivity and richness of the content; otherwise, users will not buy it.

Currently, there are two profit models. One is like domestic mobile phone manufacturers, where the glasses themselves make money, with relatively thin profits but still profitable. The other is the model of content providers, where the device can be given away at a loss. However, currently, it is difficult for the content to make users willing to pay a large amount of money, so it may take time for this model to be profitable.

Yingke: Domestic companies rely on hardware for profits, and the profits are quite meager. All industries are complaining. What do you think of the future prospects of the hardware industry?

Zhang Xueli: Domestic enterprises are best at achieving scale. Once the scale is achieved, even if the profit margin is low, the absolute profit will be large. Just like mobile phone screens and semiconductors, the prices of many domestic mid - end or low - end chips have dropped to unimaginable levels.

The business world is in a dynamic balance. When there is growth in some areas, there will be decline in others. If everyone thinks a certain industry is difficult, it may mean that some people need to switch to another industry. There must be some industries that can make people money at a certain stage. If everyone thinks a certain industry is very profitable, it also means that too many people are entering this industry.

Yingke: Has the AI glasses industry reached a stage of fierce competition and meager profits?

Zhang Xueli: There is such a trend. In fact, when an industry reaches a certain scale and the supply chain becomes more mature, reshuffle is an inevitable process. In the end, it depends on who has better technology and who can burn more money. Money itself is also a part of the ability.

Yingke: There are too many companies making AI glasses now. Some people think that some companies just assemble the display and voice - interaction supply chains to make products, and this method is not reliable. What's your observation?

Zhang Xueli: After the release of many AI glasses products, market feedback has shown that it is much more difficult than expected, especially for companies that have never been involved in the glasses or wearable device industries. Currently, the supply chain is not mature, and there are significant differences between different suppliers. In the process of assembling the supply chain, if there is a mistake in one aspect, the entire product may fail.

However, more teams entering the AI glasses industry is beneficial to the development of the industry because Chinese teams can quickly make the entire industry accessible to global consumers, although some teams may face ups and downs in the process.

Based on my current observation, whether it's AR devices or AR - related optical enterprises, Chinese companies are already far ahead. In the future, the industry will also be in an absolutely leading position globally.

02 Changes in the Interaction Requirements of AI Glasses

Yingke: You've been in touch with Rokid for a long time. Why did you only start collaborating last year? What were the points that impressed them?

Zhang Xueli: In the early days, Rokid invested more energy in the Birdbath solution. From the release of the Vision Pro at the beginning of last year to the improvement of the application scenarios of smart glasses brought by GPT, many domestic companies began to consider making more innovations based on Meta's products and adding display functions. This opportunity was very important. We've always had the technology, but it really takes big companies to enter the market and educate some users.

Now that AI glasses are very popular, the confusion that may be faced in the end is: How does the interaction between me and AI actually happen? When the efficiency of voice interaction cannot meet the requirements, it is found that 80% of human information comes from the eyes, so AR will come into play. It's a gradual process.

Yingke: Does Rokid put forward any requirements for your products? Will you jointly design and develop products?

Zhang Xueli: We've been in touch with Rokid for the past three or four years. In the past, Rokid's AI glasses were mainly used for watching movies, so they adopted the BirdBath solution. Last year, they began to consider the diffractive waveguide solution.

I've always thought that in the early development stage of AI glasses, many ideas come from optical companies. The product solutions of different companies vary greatly, with different advantages and disadvantages. Our solution may have impressed them at that time.

03 The World's Only Technical Solution

Yingke: Could you introduce the products currently being developed by the company?

Zhang Xueli: We are an optical module device manufacturer. In the past few years, we have been researching and developing diffractive waveguides and their supporting miniaturized optical engines. For example, we are making products that are thinner, have a larger display range, higher efficiency, and better color uniformity. In terms of products, consumers will feel that the AR glasses headset is brighter, has a higher color - restoration degree, and a larger display range...

In 2021, we launched a brand - new diffractive waveguide structure globally, which can achieve binocular display with a single optical engine and we also exclusively hold the patent. Many commercial products have applied this diffractive waveguide, but only we can do it. Other waveguide products in AR glasses applications must use two optical engines for binocular display.

Yingke: Is the diffractive waveguide the current mainstream technical route? I see that Apple and Huawei have both made arrangements in this area. Do they need to use two optical engines? Or are there differences in the specific technical routes?

Zhang Xueli: The technical routes are similar, but they all need to use two optical engines. There are many technical routes for AR glasses. Since Google Glass, it has used the prism technical route, which can only provide a small viewing window for display and cannot be placed directly in front of the eyes, but rather diagonally above or below. You have to glance at the glasses to see it.

Many AR glasses for watching movies on the market currently adopt the Birdbath solution. Previously, Thunderbird and Rokid both used this type of solution. The advantage is that it has a good color - restoration degree and relatively low cost. The disadvantage is that it doesn't look like ordinary glasses.

Now, people are still exploring other AR optical technical routes to create a product with the appearance of ordinary glasses, so they are moving towards the waveguide route.

The diffractive waveguide has several characteristics. First, it can achieve the appearance design of ordinary glasses. Second, it uses a relatively mature semiconductor manufacturing process, which has the possibility of mass production. Once mass - produced, the cost may decrease exponentially.

Of course, it also has some disadvantages. For example, the display effect and color - restoration degree are not as good as those of the Birdbath technical route. However, for users, its cost and appearance are more acceptable.

In terms of the optical display effect alone, the diffractive waveguide may reach a level of 50 - 60 points, and it is also the most likely to reach a level of 100 points among all current technical routes. Of course, there may be new technical routes.

Yingke: What are the advantages and technical challenges of the company's development of diffractive waveguide products using only one optical engine?

Zhang Xueli: The original intention of designing this architecture was that the cost of optical engines was very high. As a consumer product in the future, consumers would be very sensitive to price. Using a single - optical - engine solution can save the cost of one optical engine and also reduce the power consumption of the optical engine. Reducing power consumption means saving battery capacity. Moreover, the battery technology has not changed much in the past few years. Saving power can also reduce the weight of the glasses.

After actually implementing the single - optical - engine solution, we found that if two optical engines are used for display, absolute binocular fusion must be achieved. The dual - optical - engine solution requires binocular fusion for each complete device, and the precision requirements are high and difficult to maintain alignment. Therefore, the single - optical - engine solution greatly simplifies the assembly process of the complete device and will not cause binocular misalignment during user use.

Yingke: What are the technical challenges of using only one optical engine? Are there any problems?

Zhang Xueli: When using only one optical engine, the optical engine needs to be placed right in the middle of the nose bridge so that the light on both sides can be evenly distributed. This means that the width of the nose bridge of the glasses will be wider than that of ordinary glasses, which poses a certain challenge in design. Second, there is an angle between the two lenses of the glasses, which is not an absolute 180 degrees, but about 174 - 176 degrees. The diffractive waveguide lenses are an absolute 180 degrees. For some designers who pursue a sense of design, the difference in the lens angle is unacceptable. However, from the perspective of cost and engineering implementation, our solution is better.

Previously, Misa, the boss of Rokid, wore AR glasses using our solution at the High - quality Economic Development Conference in Yuhang District, Hangzhou. You can see that the finished product doesn't look bad and is similar to ordinary glasses.

04 The Smell of Price War

Yingke: You just mentioned that there are many optical technical routes. Do you feel that the competition is more intense?

Zhang Xueli: In fact, we very much welcome more competitors in the industry. If there is only one dominant company in the industry, it means that the industry has not yet taken off. More companies entering the market will bring more possibilities.

Speaking of Lightship itself, in addition to the core single - optical - engine technology, for color waveguide display (RGB three - color waveguides), we can use a single layer of glass for the waveguide. Overseas giants such as Microsoft's Hololens and Magic Leap use two layers of glass waveguides for full - color RGB waveguides.

Lightship does this mainly for two reasons. First, using one less layer of waveguide means lower cost. If both layers of waveguides have gratings, there is also the problem of absolute alignment of the gratings during assembly. Lightship's solution can reduce the bare cost by half, and a lot of cost can be saved just in terms of the process. The cost is reduced by more than 50%. Second, two layers of waveguides mean increased weight, and using a single layer of waveguide makes the product lighter, which is more attractive to consumers.

Yingke: Are Lightship's technologies unique? Can other companies not achieve them at all?

Zhang Xueli: If Lightship has made early arrangements in some technical routes, theoretically, other companies cannot use the same technology. If they can bypass it and achieve similar technology, that's also okay. Lightship has always attached great importance to the accumulation of intellectual property rights, and we also respect the IP of our peers. Our initial insistence on using a single - layer waveguide for full - color RGB was to ensure that we would never infringe on the IP of peers such as Microsoft and WaveOptics.

Yingke: Has the number of your customers increased now? What new requirements and expectations do they have for you?

Zhang Xueli: The number of customers has more than doubled every year. In addition to startups, more and more medium - to - large - scale companies, unicorn - level companies, and even listed companies are coming to learn about our products.

Customers' demands for Lightship have always been clear. First, customers need us to provide products with better optical effects, more efficient diffractive waveguide products, clear images, and sufficient brightness of the