After the excitement of CES, what is the answer for AI to move into the physical world?
After the hustle and bustle of this year's CES, a clear signal has emerged: AI is "pouring" into the physical world from the digital realm with an unprecedented density of physical entities.
From companion robots capable of understanding the environment and initiating interactions to smart wearables that deeply integrate health monitoring and emotional support, the "bodies" of AI have become diverse and tangible. Behind this hardware extravaganza lies an ultimate competition for real - world experience implementation: How can technology move beyond demo videos and transform into daily value that users truly need and are willing to pay for continuously through appropriate hardware forms?
Recently, a discussion on AI consumer hardware by Linear Capital echoed the major trends at CES on a micro - level. Yang Jun, a partner at Linear Capital, along with three AI hardware entrepreneurs in the education, companionship, and health sectors, went beyond the simple discussion of technological capabilities and tackled the most practical issues. The following are their observations and discussions.
Looking back at 2025, if we had to find a keyword for AI, it would probably be: starting to have a "body". And the just - concluded CES 2026 in Las Vegas has condensed this trend into a tangible display — from companion dolls to smart rings, AI is accelerating its exit from dialog boxes and seeking its foothold in the physical world.
This is not just an overflow of technology but more like a real competition for users: Who can use hardware as a container to truly harness the power of AI and turn it into an experience that users are willing to pay for.
This trend coincides with the focus of a recent AI consumer hardware forum by Linear Capital. In this dialogue, Yang Jun, a partner at Linear Capital, and three front - line entrepreneurs — Gu Jiawei, the founder of Lingyu Universe; Xue Lijun, the founder of Ludens AI; and Lou Xiaodu, the founder of Bangtang Technology — had a very practical discussion around three areas: education, companionship, and health. They were not discussing "what AI can do" but "what kind of AI hardware can actually attract customers".
Where exactly does the "Aha Moment" of AI hardware occur? Is it adding AI to old hardware or creating a truly "new species"? And, in an era when pure software is becoming increasingly difficult to develop, why is hardware becoming a card for both offense and defense in the hands of Chinese entrepreneurs? This dialogue attempts to provide some on - the - ground answers.
The Aha Moment in the Process of AI Hardware Entrepreneurship
▌Gu Jiawei @ Lingyu Universe:
We distributed over 20,000 units of our Xiaofangji device in the early stage. From the feedback of this first wave of real users, we observed some very interesting phenomena. The most obvious one is that AI has re - verified the classic "anxiety logic" of educational hardware.
The sales data of these 20,000 units tell us that the core motivation for parents to buy is the same as that for buying BBK products twenty years ago — back then, it was to address the anxiety of "learning computers", and today, it is to give their children an "entry ticket" to the AI era. This indicates that when parents generally hope to equip their children with a "new tool" for the AI era, as long as our product fills this niche, we can naturally meet this huge market demand.
However, parental anxiety alone is not enough. The key lies in whether children can engage with the product. This leads to our second discovery: AI is transforming hardware from a "tool" into a "co - creator". For example, when a child uses the Xiaofangji to view the main KV pattern on the background wall, the multi - modal large - scale generation model "imagines" it as an octopus flying with its tentacles outstretched. This actually means that AI can act as a Copilot, nurturing children's curiosity and bringing their imagination to life. This "co - creation" experience generates stronger user stickiness than simple knowledge dissemination.
Furthermore, behind these interactions, we also discovered the in - depth insight value of data, which is helping parents truly "understand" their children's abilities. We launched a function that generates a Feeds stream based on children's chats and behaviors, and found that parents showed extremely high interest. Through AI's data capture, parents can discover their children's interests and strengths earlier. This means that hardware is no longer just a companion tool but has become a bridge connecting parents and children in understanding. This is also the part where we believe AI hardware has great potential in the long - term.
▌Xue Lijun @ Ludens AI:
When developing our companion robot, our biggest Aha Moment came from the return to "intuitive design". In a child's cognitive world, the logic is very simple: if a robot has eyes and can move, it should naturally be able to "actively" play with me, rather than just being a passive machine waiting for instructions.
With the power of large - scale models, we achieved this — enabling the robot to generate game scenarios in real - time by calling underlying algorithms and truly interact according to the child's ideas. To our surprise, when technology catches up with children's intuition, it brings extremely positive feedback.
This feedback is not just a momentary surprise. More importantly, it brings great joy to children. This pure joy breaks the curse of traditional electronic toys that kids quickly lose interest. It makes children willing to play with the robot continuously and for a long time. This is a very valuable lesson for entrepreneurs: In the AI hardware field, winning "happiness" means winning "retention".
▌Lou Xiaodu @ Bangtang Technology:
Firstly, there is a qualitative change in commercial value. We found that simply adding AI to hardware often doesn't result in a favorable ROI. However, if AI can provide new and interesting experiences in specific scenarios, users' willingness to pay will increase significantly — they may be willing to pay $20 per month instead of the original $9.9 subscription, and the renewal rate is extremely high.
Secondly, there is a three - dimensional upgrade of services. If you ask ChatGPT "How did I sleep last night?", it can't answer. But if you combine the private data of IoT devices with large - scale models, it can immediately provide personalized care and suggestions. This "understanding you" experience makes hardware services instantly more three - dimensional.
Thirdly, there is explosive power in the marketing aspect of AI. Once the product is successful, AI can help us generate a vast amount of content on global social media at extremely low cost, requiring only 1% of the previous number of employees. This explosive power, whether for pure AI agents or AI hardware, was unimaginable before.
The Difference between Old and New Hardware Lies in the People Doing It
▌Gu Jiawei @ Lingyu Universe:
Regarding the idea of "old hardware + AI", I first go back to the demand itself to do a cost - benefit analysis. New technologies only make it more efficient to meet existing needs, but it is very difficult to create a new demand out of thin air.
AI toys are the most obvious direction but also the most likely to fail, because there are two major mismatches. Firstly, there is a demand mismatch. Current AI toys are just learning to "talk", but how many children really have so many questions to ask continuously? The demand for high - frequency conversations may not exist in real - world scenarios. Secondly, there is an economic mismatch. Talking essentially consumes a lot of energy — the cost of tokens, IC circuits, and IP all need to be considered. If these fundamental issues are not resolved, the idea of creating an "AI version of Pop Mart" is difficult to justify economically.
Based on this judgment, I avoided obvious tracks like AI glasses and instead chose what Duan Yongping called the "niche market" — similar to the next - generation "Xiaotiancai". This is a huge market where a single product can sell 12 million units and generate an annual profit of 2 billion. However, today's "Xiaotiancai" is like Nokia in the past. It is strong in channels and supply chains but weak in AI and content capabilities. This also leads to our core opportunity: defining new AI hardware with "AIGC content and AgentOS services".
The times have changed. Parents are worried about their children's eye health from staring at small screens, which is a pain point. This unique pain point, combined with the new capabilities brought by AI, gives rise to real opportunities for new AI - native product categories. What we need to do is to seize this opportunity with a new definition in this era.
▌Xue Lijun @ Ludens AI:
Actually, the so - called difference between old and new product categories essentially lies in "the people doing it". In the past, people from the hardware background dominated, with the logic of "hardware + AI" — first creating a clear hardware shell and then adding AI functions. But today, more people from the AI background are entering the field, and the thinking has completely reversed: We first plan "what kind of AI interactions we want to create" and then deduce what kind of chips and peripheral devices are needed accordingly.
This shift in thinking of "defining hardware with AI" brings an unexpected data - driven advantage. In the past, robots were pre - programmed to perform set actions, requiring extremely high precision for motors and batteries. But today, with stronger AI perception and decision - making capabilities, robots can adapt to the environment more intelligently and "take shortcuts", without having to move mechanically all the time. This actually extends the lifespan of key components by at least 50% - 60%, fundamentally optimizing the cost structure of hardware design.
▌Lou Xiaodu @ Bangtang Technology:
Our strategy for hardware development is different. We do not reject the combination of existing hardware and AI. Instead, we focus on the AI needs of the 40 million European and American female users we have accumulated. Most of these users are between 25 and 45 years old. When exploring their needs, we discovered a huge and overlooked "new scenario" — mental health and religion.
Relevant data shows that 90% of women in the real world have mental health problems, and 40% of them go to churches to confess to priests. For this scenario, we developed a software called "AI Priest", and the results were astonishing: the customer acquisition cost (CPI) dropped from the original $1.5 - $2 to $0.3 - $0.5, and the user renewal rate increased from 70% to 80% - 90%. This proves the strong demand for this new scenario among users.
Based on this, when it comes to the question of whether to do "old hardware + AI" or create a "new species", our answer is: We can reshape the "old" hardware foundation with new scenarios based on user insights. We decided to develop smart rings. On the one hand, traditional thermometers are "anti - human" as they are only used once in the morning, while rings are more comfortable to wear while sleeping. On the other hand, we found that in religious scenarios, European and American women already like to wear rings and even have the habit of "twirling rings".
Therefore, if a product already has a hardware foundation and has been performing well so far, when we introduce it to a brand - new scenario and reshape it using users' strong willingness to pay for AI, the unit price can reach hundreds of dollars, and the market will become huge. This is the new path we are exploring.
What Kind of Founders Can Win in This Wave of AI Hardware Enthusiasm
▌Gu Jiawei @ Lingyu Universe:
There is a core consensus in AI hardware development: "Stay away from mobile phones". This can be divided into two distinct directions: one is to engage in "atom - level interactions", such as humanoid robots and smart home appliances, which are beyond the physical capabilities of mobile phones. In such projects, the proportion of hardware is still significant. The other is to pursue "ultimate efficiency logic", targeting users who do not use mobile phones or simplifying multi - step tasks on mobile apps into one - click operations through AI hardware.
I firmly choose the latter. Based on this, I believe that the next - generation entrepreneurs are more likely to come from a software background. In our product definition, 80% of the value lies in software, and hardware is just a carrier. For this type of AI hardware, the key to success does not lie in stacking hardware parameters but in the ability of software and AI to define interactions.
▌Xue Lijun @ Ludens AI:
Since we started our business, the most profound feeling is that the domestic supply chain has become extremely mature, and any short - comings in the R & D process can be quickly addressed. Therefore, the most core ability for the next - generation AI hardware founders is no longer to manage the supply chain but to have a global perspective and cross - cultural understanding. This is because consumer demands have changed. Today's consumers have seen a wide range of products in every category, and they no longer need to be taught how to use smart devices. Their demand has shifted from seeking better cost - performance to better experience and more emotional value.
This change has directly led to the reconstruction of the business model. In the past, when I was at DJI, the logic was to continuously sell hardware, and once the product was sold, there was little interaction if there were no major bugs. But in the AI hardware era, the logic is completely reversed: We actually hope that users replace hardware less frequently and instead continuously purchase content and services based on the hardware.
Take our companion robot as an example. We hope that users will continuously buy clothes, games, and warranties for it. This is the biggest opportunity for entrepreneurs today: finding a high - frequency, sticky scenario that can provide data feedback in the global market. By using data to refine AI, when AI becomes powerful enough, it can flexibly change its appearance and emotional value to reach different user groups.
▌Lou Xiaodu @ Bangtang Technology:
If we limit the scope to the consumer market, I believe that the next - generation AI hardware founders are likely to be Chinese.
Although AI is very popular in the United States, most entrepreneurs in the Bay Area are working on enterprise services or large - scale models and lack a complete industrial chain support. In contrast, after the globalization wave of TEMU and TikTok, China has accumulated a group of all - around talents with global experience in various fields such as engineering, product management, customer acquisition, and brand building.
We have recently recruited many people with such backgrounds and found that their strategies for entering new vertical markets are completely different from traditional smart hardware approaches. The essence of this difference lies in: They no longer operate with traditional "hardware thinking" but use the thinking of Internet product managers to develop hardware. This keen insight into product trends is a necessary condition for the success of the next - generation AI hardware companies.
What Advantages Does the Chinese Environment Offer to AI Hardware Entrepreneurship
▌ Gu Jiawei @ Lingyu Universe:
Recently, from a capital perspective, I noticed a different aspect: Why has AI hardware suddenly become so popular in China this year?
The core reason is that it is extremely difficult to invest in pure software in China. Basic model companies control the underlying data, can easily see your strategies, and may enter the market to out - compete you. Software entrepreneurs find it hard to defend their positions. However, hardware is different. It is like a "physical container" — It is easy for large companies to copy pure software, but it is difficult for them to quickly follow up on hardware projects in the early stage.
More importantly, it is easy to put together a demo, but it is difficult to sell one million units. Many companies can create a demonstration sample, but to actually distribute products, establish channels, build brands, and operate IP across the entire value chain like we do, the industry know - how accumulated in this process is the most difficult - to - replicate barrier in the Chinese market.
▌Xue Lijun @ Ludens AI:
I feel very strongly about the advantages of the Chinese entrepreneurial environment, especially being in Shenzhen. These advantages are mainly reflected in three aspects:
Firstly, the "stake - out" strategy enabled by physical proximity. We have a simple and practical method during R & D called "staking out". Whenever there is a problem with the supplier's product, we drive directly to the factory and stay there until the problem is solved. In this process, you can not only solve the problem