Among entrepreneurs, OBSBot CEO Liu Bo is already seen as something of a veteran.
In 2016, Liu graduated from Zhejiang University and went to Shenzhen to start a business. Then, as now, his background did not stand out much among the high-profile “genius youth” founders often associated with China’s technology sector.
He does not share many sentimental stories. The experience he still remembers most vividly happened in 2019, when a mistake in product definition left the company with unsold inventory. It had only enough cash left to last a few months. It laid off half its staff and halted related businesses. In his memory, that was the quietest winter.
The turning point came in 2020. As remote work became normal, demand surged for webcams used in meetings. Liu realized that most webcams on the market were fixed, while people working from home did not sit upright all day. They moved, stood up, and changed posture. He decided to make a webcam that could track intelligently and frame automatically. In 2020, OBSBot released its Tiny series cameras.
That product saved the company. In the years that followed, OBSBot took more than 50% of the high-end webcam market, ranking first. It also raised a nine-figure USD sum from investors including Didi, HongShan, Forebright Capital, CMB International, HKX, Brizan Ventures, and other institutions.
In Shenzhen, a crucible for hardware startups, it is rare for a company to make imaging products, survive for a decade, and still maintain 50% annual growth.
Occasionally, when Liu looks back, he still finds it hard to believe. “Today, the companies that can stand firm in imaging are all established giants and top players. Overseas, there are longstanding Japanese giants such as Sony, Canon, Panasonic, and Nikon. In China, there are DJI and Insta360. We are just a very, very small company in this industry. Even under those conditions, we have been doing this for ten years. For the first five years, the company had almost no revenue. We survived in that environment by sheer force,” Liu told 36Kr.
A large part of the reason may come down to restraint. Liu understands clearly that webcams are a niche category and are unlikely to become a massive market. A segment needs to grow to at least USD 10 billion before giants see enough reason to enter and divide the market. Webcams do not reach that scale, though they can still support a company with several hundred employees.
So OBSBot has remained focused on this scenario. Along the way, it has built its own moat in supply chains and technology. On that foundation, the company has begun investing in R&D around a broader imaging ecosystem and experimenting with more innovative products.
“Always work around your own positioning. That is the foundation of everything,” Liu told 36Kr.
The following transcript has been edited and consolidated for brevity and clarity.
36Kr: Did the growth of the company’s webcam business last year come mainly from new users or replacement demand?
Liu Bo (LB): We have to look at the domestic and overseas markets separately. In China, the live streaming segment has entered a stage of steady development. The overall user base is still rising steadily. Some product shipments come from experienced streamers upgrading their equipment, while others come from newcomers buying gear as they enter the industry. It is hard to precisely separate new demand from replacement demand, and the overall growth curve will not be especially steep.
Overseas, the market leans more toward replacement demand. The main webcam user base is concentrated in office settings, and the user base is stable. What we are doing is gradually converting traditional webcam users into our product lineup, which is more intelligent and better designed.
36Kr: How do you assess the webcam market?
LB: We know very clearly that, even if user education and market penetration were pushed to the extreme, webcams would still be a small but meaningful category. The main scenarios are office meetings and personal live streaming. It is very hard for the category to become a mass market opportunity. A segment needs to grow to at least USD 10 billion before giants see enough reason to enter and take a share of the market. Webcams clearly do not reach that scale.
But that does not mean we have no room to grow. Beyond our core market in PTZ (pan-tilt-zoom) webcams, OBSBot is also entering other niche areas.
Webcams can support cash flow and generate profits, while also supporting investment in innovation and R&D. That is what we want to do over the long term.
36Kr: OBSBot has released several new products and made major software upgrades. What are the most prominent pain points you see among users at this stage?
LB: We began focusing more on the domestic market in 2023, with promotion and sales centered on the Tiny series. After roughly two years, this category has become a top seller in the industry. To address current pain points around traditional webcams in remote work and video meetings, we will soon release a new conference webcam called Meet Flip.
The core capabilities of a webcam can be summarized in three dimensions: image quality, audio performance, and a complete set of intelligent features. We have never stopped iterating on the software side, because domestic live streaming scenarios are relatively deep. You could say we helped create the category of live streaming webcams, and it contains many small but real scenario pain points.
For example, many streamers want the camera to automatically follow hand movements. When they are showing a product, they want the lens to intelligently zoom or pan. Some streamers also need the camera to automatically switch tracking targets when multiple people take turns entering the frame, rather than staying fixed on one person.
In fact, many features people take for granted today were first developed by OBSBot and brought to market by us. We were the first brand in the industry to pioneer artificial intelligence auto-framing, gesture-based subject selection, and tracking shots. The gesture-based subject selection paradigm, gesture-based zoom paradigm, and the physical tilt-down auto-sleep design that protects user privacy and is now replicated across the industry were also first developed by our R&D team and introduced to the global market.
Similar needs are also prominent in office meeting scenarios. Users often complain that in multiperson meetings, cameras cannot automatically identify the current speaker, causing the image to point at an empty seat. Remote collaboration may require showing a whiteboard or physical object, and users want the camera to switch to an overhead or close-up mode with one tap.
These needs are scattered across different scenarios, but the underlying logic is the same: users want cameras to understand shooting intent more intelligently and automatically complete target recognition, tracking, and angle adjustment, rather than requiring repeated manual adjustments.
36Kr: Traditionally, using a larger image sensor requires more internal space in the device. Tiny 3 upgraded to a larger sensor while continuing to shrink the overall device size. What were the engineering difficulties?
LB: The first-generation Tiny used a 1/2.8-inch sensor. Later, it was upgraded to 1/1.5 inches. Tiny 3 ultimately reached 1/1.28 inches, with a much larger photosensitive area overall.
It was very challenging to compress the device to its current size. The real bottleneck was not the CMOS (complementary metal oxide semiconductor) sensor, but a full set of optimizations involving the mainboard, motor miniaturization, and refined adjustment of whole-device power consumption. For Tiny 2, we had to fine-tune heat dissipation in great detail just to barely balance device size and heat. With Tiny 3, the overall optimization of the hardware system means that even though the device is smaller, heat dissipation is better and more stable.
On the audio side, which users can perceive most clearly, Tiny 3 is equipped with a new spatial microphone array. In practice, it solves two problems. First, precise sound source localization. The machine can clearly distinguish where sound is coming from in the environment. Whoever is speaking, it turns toward that person. This is especially useful in multiperson conversation scenarios.
Second, it enables native stereo capture from the device itself. In most scenarios, external microphones are omnidirectional and lack spatial positioning, which makes recorded audio sound flat and lacking in layers. Tiny 3’s stereo capture can perceive the entire sound field and record real spatial audio. In a multiperson conversation, for example, it can clearly distinguish the positions and distances of different speakers: who is on the left, who is on the right, who is farther away, and who is closer. This improvement may seem subtle, but it has significant value for live streaming, multiperson meetings, and real-world recording.
36Kr: At the user experience level, AI imaging products show an obvious trend toward functional homogenization. What is the root cause?
LB: The homogenization users perceive is essentially the result of overcrowding in the segment.
At present, the largest and most mainstream part of the imaging sector remains handheld imaging devices. Their product form has inherited the position once occupied by traditional digital video cameras, and the core market is large enough that every company wants to compete for it. They compete on image quality, audio pickup, and basic specifications, chasing one another with spec-stacking upgrades. In the end, functions and experiences naturally become more and more similar. The same applies to action cameras. That segment has long been defined and locked in by leading brands. Industry players follow and copy one another, eventually falling into homogenized internal competition.
But there is a key difference here. Handheld devices, action cameras, PTZ webcams, and other categories correspond to different user needs and solve different core problems. Each has its own dedicated use scenarios, so different categories still retain their uniqueness and will not fully converge.
Returning to OBSBot itself, the scenario we focus on is clear and well defined. Of course, in each product generation, peer brands will inevitably follow and copy. That is normal in the development of a niche segment, and we accept it calmly. But the core principle will not change. We must follow real user needs, adhere to long-termism, and focus on core technology breakthroughs and product refinement with patience. Only then can we avoid being pulled off course by the rhythm outside.
36Kr: OBSBot has been around for ten years. Looking back, as a student team with no experience making hardware from scratch, what was the biggest obstacle to completing mass production, iteration, and delivery smoothly?
LB: Fundamentally, it was a logic problem. Not knowing anything was actually fine. Students like us, who came from robotics competitions, did not initially have these process concepts in our heads. We slowly figured out the rules of the industry and walked the path that way. The people who truly struggle to adapt are some entrepreneurs who have gone through scaled training and come out with internet or pure software thinking. The logic of the two models is completely opposed, so they first need to change their way of thinking.
36Kr: What are the parts of the business that you still believe require humility and respect after all these years?
LB: First, you have to recognize that the hardware sector requires time, patience, and detailed iterative refinement. I think that is the foundation. You cannot rush it. Second, you have to keep creating differentiation and build sufficiently high technical and patent barriers. We attach great importance to R&D. R&D personnel account for more than 60% of the company, and every year we put 20% of company revenue into R&D.
From a purely technical perspective, many companies can achieve many functions. But hardware cannot bypass R&D cycles, production costs, and user switching barriers. Every investment must be weighed carefully.
36Kr: Some believe hardware itself iterates quickly and does not easily form barriers, so startups should put more energy into developing software and algorithms. How do you see this?
LB: We can go back to 2015 and 2016. At that time, many innovative internet hardware companies emerged, and huge numbers of smart bands, smart plugs, smart speakers, and similar projects appeared together. The popular model was to offer free hardware and monetize associated software subscriptions. The threshold for that kind of standardized hardware was not high. In Shenzhen’s Huaqiangbei, three people could build an identical product within six months.
History has already played out once. The conclusion was that the vast majority who jumped onto the bandwagon faced attrition.
Today, many algorithm teams want to use hardware as a traffic entry point and then build an advantage with self-developed software. But in reality, that is very hard to execute. In the later stages, pure software competition basically moves toward industry consolidation. The pattern is the same: only a few companies or brands can build a software ecosystem with exclusive barriers. Players that cannot do so eventually fall into long-term internal competition. The large model field is similar. Huge numbers of players pile in and compete on computing power, while profit margins continue to be compressed.
There are no shortcuts in imaging hardware. There must be enough difficulty. Only companies with high-threshold hardware, mature supply chains, self-developed underlying algorithms, application development, and a complete set of comprehensive capabilities can create some distance from others and possibly survive. Frankly, even with all the technical accumulation OBSBot has built over the years, when we first enter a completely new product category, we still inevitably need a two-year R&D and refinement cycle.
Once a hardware product lacks differentiation, it will inevitably enter a price war, and the whole industry will gradually become unprofitable.
36Kr: How do software teams and hardware teams differ in working styles?
LB: Hardware costs money, and the switching threshold is high, so every piece of hardware must be refined to the highest standard. Once you open a mold, it costs money. Production is also a real investment. Once you ship a batch of goods, you must make sure there are no problems. If you rush to release an immature product and problems arise, brand reputation and users are directly damaged. An entire warehouse of goods could be stuck in your hands.
By contrast, software development emphasizes rapid trial and error and high-frequency iteration. If there is a new idea, you put it online immediately and adjust it at any time. One side allows trial and error and pursues speed. The other must be rigorous and seek stability. The two ways of working are completely different.
36Kr: In terms of imaging hardware, do you think new entrants still have opportunities?
LB: The industry is not something a team can enter casually. The entire industrial chain does not have mature, general-purpose supporting solutions. Every key link can only be polished and fully understood from beginning to end by yourself.
That explains why the companies that can stand firm in the segment today are all established giants and top players. Overseas, there are longstanding Japanese giants such as Sony, Canon, Panasonic, and Nikon. In China, there are DJI and Insta360. Compared with them, we are just a very, very small company.
Even under those conditions, we have been doing this for ten years. For the first five years, the company had almost no revenue. We survived in that environment by sheer force.
Behind this are high barriers in every direction. First, the product has to be compact enough, which creates strict requirements for power consumption control. To adapt to that, you have to thoroughly reconstruct the underlying R&D system. Computing power optimization for small hardware platforms alone is a major difficulty and cannot be made up for in a short time. Solving power consumption control alone is enough to block most players.
Second, product refinement and superior user experience depend on the system integration of multiple capabilities: product definition, structural R&D, full supply chain control, and refined operations around R&D and production costs. As these links stack up one by one, the barriers naturally rise.
36Kr: Why is it difficult for overseas competitors to copy this?
LB: For Chinese hardware entrepreneurs, supply chain is only one part of the capability. Domestic teams have formed a clear comprehensive advantage in the R&D and integration of complex hardware and systematic products. Overseas competitors find it hard to copy because, on top of a mature supply chain, we have very strong advantages in technology R&D.
The imaging segment is interdisciplinary. The algorithms, motion control, and optical coordination we have accumulated over the years have allowed us to build a full-chain AI imaging automation system. That is our core technical asset. Combined with the domestic supply chain’s deep customization and fast engineering implementation, we can turn technical ideas into products.
As global e-commerce infrastructure matures, sales channel barriers have largely been dissolved. In reaching users, we now stand on the same starting line as international brands.
Then there is talent. The overall advantage of China’s engineering education system and logical thinking training means domestic practitioners’ comprehensive capabilities have already reached parity with overseas peers, and this gap will continue to widen in the future.
36Kr: What is the decisive factor that creates a long-term gap between companies?
LB: For a company to survive over the long term and develop healthily, strategy is absolutely key. You must think clearly, what exactly is the company doing? What thinking lies behind it? Strategy determines the subsequent layout.
Of course, in the early stages, we may inevitably make some intuitive decisions. But in the end, what is reflected is also a high degree of coordination between strategic thinking and execution.
So strategy comes first. Around a clear strategy, you then build the organizational structure and plan the pace of development. The industry itself needs different types of companies to coexist, and each company follows different logic. Only by finding your own position, clarifying development goals, adhering to long-termism, valuing R&D investment, and moving forward along that direction can a company walk steadily.
Always work around your own positioning. That is the foundation of everything. We have always done it this way.
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Huang Nan for 36Kr.