From large models to AI companionship: What are the cyclical patterns behind the rotation of AI hotspots?
How hot is the AI hardware market? According to data released by 36Kr, as of the first half of 2025, there were 114 investment and financing deals in the fields of embodied intelligence and AI hardware in China, with a total financing amount exceeding 14.5 billion yuan, far surpassing the 92 deals and 9.8 billion yuan in the whole year of 2024. In May 2025 alone, the funds flowing into AI hardware accounted for more than half of all investment and financing.
Ropet, a smart product by Mengyou Intelligence, is one of the players in this "hardcore entrepreneurship" wave. On October 17th, Ropet was officially pre - sold on JD.com in China. Ropet was previously unveiled at the 2025 Consumer Electronics Show (CES) and was reported by media such as Forbes, BBC, and CNBC, achieving over 6 billion global exposures and, to some extent, boosting the AI companionship product market.
Image source: Ropet
He Jiabin, the co - founder and CEO of Mengyou Intelligence (Ropet), positions Ropet as a "weak robot". "Ordinary robots serve humans, while Ropet simulates a living being that needs to be cared for. Through being cared for, it comforts the users."
He Jiabin is not a typical AI entrepreneur. He graduated from the Beijing Institute of Fashion Technology, majoring in product design. His career has spanned across Microsoft Research Asia, Baidu, Lingoo Technology, and ByteDance. He led the development of the LUKA picture - book robot and the PICO 4 VR all - in - one device. His experiences made him more concerned about a question: Why do people want to form an emotional connection with a machine?
Not long ago, at the AI Creators Carnival (ACC) jointly initiated by Silicon Valley Insider, Zhongguancun Science City, and Zhongguancun Startup Street in Beijing, Li Feng, the founding partner of Frees Fund, had an in - depth conversation with He Jiabin. They discussed the current stage of the AI boom, analyzed how intelligent companion products are defined, and recalled the ups and downs of Mengyou Intelligence's entrepreneurial story.
We edited some parts of the conversation, hoping to provide new perspectives.
First, let's share some conclusions:
- The future of AI lies in "embodied intelligence". "Embodied intelligence" doesn't necessarily mean bipedal or humanoid forms; it can also be various hardware carriers with the ability to interact with the world and collect data. Such terminals will create new opportunities for consumer products.
- "Useful" products have the advantage of low market education costs but quickly fall into price - performance competition. "Fun" products are difficult to educate the market about in the early stage, but once successful, they can command a premium and form unique competitive barriers.
- Revolutionary technologies in the hardware field usually go through three stages: The first stage is the explosion of underlying technologies. The second stage is to explore the imaginative space of the technology. The third stage is to achieve commercial value. In the first wave, the technology is still in its "core". In the second wave, the technology has some application space. In the third wave, the technology interacts with the real world and is implemented.
/ 01 / Ropet: 100% "Fun", 0% "Useful"
Li Feng: As AI hardware becomes more and more popular, the AI companion toy market has attracted particular attention. We are curious about what form AI hardware should take to achieve the function of emotional companionship. Please introduce the Ropet product first.
He Jiabin: Ropet is a brand - new pet - shaped product designed specifically for women. It provides emotional value through AI and helps users relieve loneliness.
Our first step was to create an AI pet with an attractive appearance that appeals to female users, so that users would be willing to take it home and place it in a prominent location, such as a desk. People spend a long time at their desks from graduation to retirement. Moreover, the desk is a fixed location with a stable power supply. If there is a small creature like a "desk buddy" that can provide emotional value and be present for a long time, its companionship duration and power - on rate will be very high.
Image source: Ropet
With the user's careful care, Ropet will gradually accumulate data on the owner's emotional changes and "understand you better as you raise it". It will learn the user's catchphrases and may even start calling the user "mom" slowly.
We hope to awaken the feeling of "love" in the lonely world of female users. In the high - pressure urban life today, such emotional experiences are particularly scarce. Our original intention of starting a business was to bring new product experiences with technological innovation to more women through AI and robot technology. This is our initial product logic.
It's worth mentioning that Ropet is also the intersection of the AI and trendy toy product fields and also integrates pet elements. We call ourselves a "future pet company" and hope that Ropet will become a new "silicon - based pet" species. In the future, it may exist in different forms - it can be placed on a desk, hung on a bag, put in a car, or move freely at home.
Li Feng: From a functional perspective, smart hardware in the market can be roughly divided into two categories: "useful" and "fun", and some products have both characteristics. In your choice, what are the proportions of "fun" and "useful" in your product?
He Jiabin: We chose 100% "fun" and 0% "useful".
What I understand as "useful" is that a robot can help you solve specific problems and improve efficiency, such as cleaning, reminding, and tutoring. This is the tool attribute. "Fun", on the other hand, is purely about providing emotional value.
The children's picture - book robot "Luka" that I created in my previous entrepreneurship was an attempt between the two, about a 50 - 50 split. For parents, it was "useful" as it solved the need to cultivate children's reading habits. But for children, it was "fun" - through IP images and interactive games, it made the "anti - human" activity of reading interesting and increased user stickiness.
However, when a product needs to please two different groups of people (such as parents and children) with inconsistent needs, the product definition becomes blurred.
As Wang Ning, the founder and CEO of Pop Mart, said: "If you can pull out the head of MOLLY and it becomes a USB flash drive, would you still buy so many?" A USB flash drive is a "useful" tool, while MOLLY is an emotional carrier that is "good - looking" and "fun". Once you have tool - attribute expectations for a product, its "fun" factor will definitely be compromised.
So this time, I chose to focus on "female - oriented" products instead of children's education. Because female emotional companionship is a direction that can be explored deeply and purely.
Of course, it is challenging to create a "purely fun and useless" product - how to prove its value to users? Fortunately, we are in a great era: today, almost all products with emotional value have more market demand and can command a premium. This is our opportunity and our strength. Therefore, we firmly define our product as "100% fun". (Welcome to read the report "From the Century - long Changes in the US and Japanese IP Industries, Looking at the Development Trend of China's IP Economy" by Frees Fund)
I hope the product can have the core of a "cultivation game", using AI to make the "pet" more lifelike and the interaction more natural.
Li Feng: From an investor's perspective, "useful" products have the advantage of low market education costs but quickly fall into price - performance competition. For example, household appliances such as vacuum - cleaning robots and washing machines are "useful", but no one "plays" with them. Eventually, products compete on price and parameters.
For "fun" products, it is difficult to educate the market in the early stage, but once successful, they can command a premium and form unique competitive barriers. You don't compare "which chip is used in my toy" but rather whether it is unique and can touch your heart.
/ 02 / From Graduating from the Beijing Institute of Fashion Technology to Starting an AI Companion Robot Business: How to Build Cross - disciplinary Advantages?
Li Feng: In 2025, we invested in Ropet. Before that, we had seen many projects in the "AI companion toy" field, but many AI toys seemed to be just for the sake of being associated with AI rather than being products truly based on consumer needs. I even decided not to pay attention to such projects for a while.
But later, more than one colleague strongly recommended that I meet He Jiabin. After meeting him, the deepest impression he gave me was that he first thought about "how to create something people like" and then considered "what role AI could play in it".
At that time, my biggest question was not "whether this project could succeed" but "as a man, how could he think so carefully?"
He Jiabin: Many people are curious about why I seem to understand women well.
The origin was that I studied product design at the Beijing Institute of Fashion Technology for four years. Many majors at the institute, such as fashion, jewelry, and luggage, focus on consumer insights related to women. In such an environment, I was always thinking about how to form unique cross - disciplinary advantages as a student with an industrial design background. The answer after thinking was: embrace technology. Technology is the primary productive force, and great product revolutions often stem from breakthroughs in underlying technologies.
Ropet was unveiled at China International Fashion Week. Image source: Ropet
After graduating in 2014, I caught up with the wave of "mass entrepreneurship and innovation". At that time, we were in the "weak AI" era, and people were exploring the possibilities of various scenarios. I joined Microsoft Research Asia and engaged in interdisciplinary research on human - computer interaction, exploring with scientists how to make technology more "likable" and easier for users to accept, and giving technology a "beautiful" appearance.
For example, we were developing a first - person - perspective AI assistant, that is, AI glasses. One of the key problems we tried to solve was: how to make women willing to wear a device with a camera on their heads? Because wearing AI glasses at that time might make them seem like an "outsider" and attract strange looks on the street.
Li Feng: AI glasses are very popular in the United States. For example, Meta collaborated with Ray - Ban on sunglasses - style AI glasses. But AI glasses may face some challenges in China.
First, more than 80% of office workers in the United States commute by car, with an average one - way commute of 40 - 50 minutes. They are used to wearing sunglasses to and from work, taking them off after wearing for two hours each day, which is very natural. And AI glasses can be an additional function of sunglasses, fitting into users' daily lives.
In China, the situation is different. First, the myopia rate is high, and many people wear framed glasses. Second, some female users, in particular, don't like wearing framed glasses and prefer contact lenses or colored contact lenses. According to internal observations from Tongxue (the parent company of Colorkey), in 2020, less than one - tenth of the users were non - myopic, and they chose to wear colored contact lenses for beauty. By 2024, this proportion had approached one - third. Especially during the pandemic, when masks covered the face, the eyes became the limited window for expressing "beauty".
In this context, asking them to wear a bulky pair of AI glasses not only covers their carefully - made - up eyes but also may cause discomfort and the need for frequent charging, reducing acceptance.
He Jiabin: Indeed, consumers' choices are deeply influenced by local living habits. We are making a global product, and the usage habits of users in different regions vary greatly.
Initially, we defined our target users as urban single women aged 24 - 30. This group is heavy users of platforms such as Xiaohongshu and Instagram and is willing to share new lifestyles. But after delivering nearly ten thousand devices, we found more user profiles: in Japan, more than 50% of users are women aged 50 - 70; in North America, many products are bought for children. Of course, half of the users are still young women.
Image source: Ropet
Li Feng: Besides Microsoft, you have also served as a product leader in several well - known companies. Before starting your business, you were at ByteDance, responsible for the product definition and design of PICO VR glasses.
You left ByteDance just as the AI boom was rising. What is the relationship between the two? Did you plan to start a business and make hardware first and then were activated by AI, or did you decide to start a business because of the AI boom?
He Jiabin: I joined ByteDance in 2021 and left in 2023, staying for exactly two years. My original intention of joining a large company was to accumulate experience for my next business venture.
Before that, I explored innovative hardware at Baidu's Deep Learning Lab and then co - founded Lingoo Technology with my former boss, focusing on children's companion robots. We achieved millions of unit sales in five years but also encountered a growth bottleneck.
Therefore, I chose to join a large company to "lie low" and improve my capabilities. One was to experience making hardware with sufficient budget, such as the PICO 4, which had an investment of billions, something a startup could hardly imagine; the other was to accumulate supply - chain resources. The Apple supply - chain resources we use now were obtained through "lessons learned the hard way" at that time.
Back to Uncle Feng's question: Did I decide to start a business first and then encounter the AI boom, or vice versa?
My answer is: I always wanted to start a business and create innovative hardware from scratch. Joining a large company was to