From new toys to new gateways: The AI "battle of a hundred mirrors" is approaching.
One day at the beginning of 2024, Li Chuangqi, the person in charge of Xiaomi's AI glasses, initiated an unusual conversation.
He dragged a document into the chat window between himself and Lei Jun on Xiaomi's internal office software and typed the following text: "... I've made additional considerations regarding the intelligentization direction. I believe glasses are an important strategic product and supplement for the wearable business."
Then, he hit the Enter key on the keyboard.
This report is called "Feasibility Considerations for Xiaomi's Wearable Business in the Glasses Category", and it is also the background and premise for the project establishment of Xiaomi's first AI glasses this year. At that time, there were still doubts about this project within the company. If Li Chuangqi wanted to continue promoting it, he needed higher - level approval and greater support.
Lei Jun's reply came quickly. He first "pinned" the document - this is a prominent reminder, indicating the person who made the "pin" action attaches great importance at the moment and can easily find it when reviewing later. Then Lei Jun immediately said, "You should have a detailed discussion with Jiutang (Chief of Staff of Xiaomi Group) and Zhang Lei (Vice President of Xiaomi's Mobile Phone Department and Head of the Wearable Department)."
For the entire Xiaomi AI glasses project team, the meaning was quite clear. Although they had previously experimented with two product forms, "glasses cameras" and "audio glasses", this was the first time the concept of "AI glasses" was established at Xiaomi, and the project was soon officially launched internally.
Once a certain entrepreneurial track becomes popular, the outside world is used to using the term "Hundred X Battle" to describe its popularity. After experiencing intense battles such as the Hundred Group Battle and the Hundred Model Battle, the new concept being discussed in the Internet world is - the "Hundred Glasses Battle".
The fuse was lit from Zuckerberg's company. In September 2023, with the second - generation Rayban - Meta glasses, Meta washed away the failure of the first - generation Ray - Ban Stories.
This smart glasses, which integrates the Llama model, has AI applications such as real - time translation and voice Q&A, and combines audio and shooting functions, impressed the market with its price of $299, weight of 48g, and simple sunglasses appearance. It sold one million pairs globally in less than a year, becoming the first and so far the only big hit in this field.
This is a huge stimulus to domestic technology companies with sharp senses. From startup teams to AR/VR companies, from mobile phone manufacturers to Internet giants, all are showing a posture of scrambling to enter the market.
They are all vying for an opportunity: Will AI glasses be another consumer - level entry point after smartphones?
Seize the Beachhead and Wait for the Inflection Point
At the end of 2022, when ChatGPT emerged, Xiaomi's Wearable Department conducted a study on the application of large - model products. One of the conclusions was that future intelligent hardware products must be able to see, hear, and sense.
Deducing from the latter two requirements, manufacturers have many sensors and body parts to design (such as earphones, bracelets, etc.). But when it comes to the first requirement, only glasses remain, because the entrance for vision is unique.
Alibaba also has a similar judgment. Quark AI glasses announced their technological progress at WAIC2025. Song Gang, the person in charge of the terminal business of Alibaba's Smart Information Business Group, believes that AI glasses will be the most important product form in the intelligent wearable direction. They will become another pair of "eyes" and "ears" for humans, with strong scene penetration ability.
"Glasses will definitely become the'sensory center' for the next - generation human - machine interaction, which will lead to a sharp explosion of personal data volume," Song Gang said. "It is expected to become the most important personal mobile entry point after smartphones."
But AI glasses are still waiting for their inflection point.
This inflection point will be reflected by sales volume after the supply chain is truly mature. Xiaomi once made a judgment of "5 million units in 3 years". After crossing this inflection point, the strategic position of AI glasses will be more clear. If it doesn't come for a long time, it may return to the category of "electronic toys".
Although the Hundred Glasses Battle has not really started, the ecosystem of participants is already extremely large and diverse. Among Internet giants are Alibaba and Baidu; among mobile phone manufacturers are Huawei and Xiaomi; there are also AR/VR companies such as TCL XESS and Rokid, as well as numerous startup companies such as AnkerMake, In Air, Hive Technology, Li Weike, and Yingmu Technology.
Charting: He Miao; Data compilation: Zhou Mo
After identifying the direction, "seizing the opportunity" has become the common mindset of many competitors.
After the project establishment of Xiaomi's AI glasses, it was codenamed "O95". In addition to core members like Zhang Lei and Li Chuangqi, many members were transferred from other departments as needed while promoting the project.
Zhang Lei and Li Chuangqi would say to the department heads of other departments, "I need these people. Tell me who you think is suitable and transfer them to me quickly."
"O95" successively introduced project managers, product managers, software engineers, and hardware engineers from within the company, and also borrowed employees from departments such as the Camera Department and Xiaoai, as well as core partners from outside. After the project was officially launched, the number of people increased exponentially, and at the peak, thousands of people were involved in the operation.
The office was also temporarily opened up. On the 5th floor of a building in the Xiaomi Science and Technology Park, the team found a spare parts warehouse that was previously used for mobile phone testing. After cleaning it up, they named it the "Black Room", and dozens of people stationed there. Others came here continuously every day to discuss requirements.
Startup companies are probably even more eager. AnkerMake is a company that has entered the AI glasses field from the power bank business. In the past, it targeted the top 10% of users in the power bank market. "But our ceiling is very low. If we only focus on power banks, our annual sales will only be a little over one billion yuan," Chen Pan, COO of AnkerMake Technology, told Jiemian News reporters.
In 2021, the company replenished a batch of "ammunition" from investment institutions. CEO Zhang Bo, COO Chen Pan and others began to look for the next core product line. That's when Rayban - Meta came into AnkerMake's view.
Almost in the same rhythm as Xiaomi, near the end of the first quarter of 2024, the AnkerMake team finally sorted out basic issues such as consumption scenarios, user profiles, costs, and pricing, and decided to enter the market. But different from Xiaomi, which was working silently at this moment, AnkerMake held a press conference in late May of that year.
Looking back, this was a press conference where marketing came first without a product yet. As one of the earliest companies expected to replicate Rayban - Meta, it indeed attracted a lot of market attention for AnkerMake. But the company's fundamental purpose was to send two strong signals to the industry: looking for money and looking for people.
"AI glasses are not something that can be accomplished with 30 - 50 million yuan. We still want to raise another round of funds," AnkerMake said. That year, AnkerMake received Series A/A+ financing from institutions such as Guangyuan Investment, Future Light Cone, Yuntian Lifly, and Oasis Capital, and completed the transaction within the year.
"The transaction must be completed within four months" became a hard requirement for investment institutions by AnkerMake at that time. The team wanted to spend more energy on the product and consume as little energy as possible on financing.
In addition, the most crucial position the team lacked at that time was a person who understood both AI and applications. Zhang Bo and Chen Pan flew from Shenzhen to Beijing several times to persuade this person. The news was finally officially announced in November 2024 - this was Pan Xin, who later joined as CTO, had experience in core AI positions at Google and ByteDance, and was serving as a co - founder at Lingyiwanwu at that time.
However, due to multiple factors, AnkerMake's first - generation "AI Selfie Glasses" A1 was not successful, and was even a failure in a certain sense. Chen Pan admitted that besides accepting the reality, the team still had a chance to make up for it.
What couldn't be made up for was the missed opportunity. Chen Pan pointed out the root cause: For a small factory only well - known in the power bank field, if it wants to enter a category with more advanced technology and greater market potential like AI glasses, it can only occupy a position in advance.
"It's very difficult to squeeze in later," Chen Pan said.
The Battle between Big and Small Companies
In 2023, Li Chuangqi advised many friends in the investment field not to invest in AI glasses.
Because in his view, this is an opportunity reserved for mobile phone manufacturers with a wearable ecosystem and leading Internet companies that master large models and national - level apps, and there are only these two types. "Because what users need is not a single piece of hardware, but a complete set of services."
According to this logic, the ecological advantages of manufacturers such as Apple, Samsung, Huawei, and Xiaomi will only become more obvious. Unless they can achieve full compatibility in terms of permissions, the irreplaceable product value of independent manufacturers will become a bit vague.
A product manager of AI glasses at Goertek also told Jiemian News reporters that after serving multiple customers, he personally is most optimistic about mobile phone manufacturers making AI glasses. Because they have a strong hardware foundation, with a complete and mature system from R & D, supply chain to quality management, and the interaction with mobile phones can penetrate to the lowest level.
Startup companies obviously don't buy into this logic. "Among ten friends, maybe ten people use the same mobile phone, but it's impossible for ten people to wear the same pair of glasses," Chen Pan said. It is a product with such a personal attribute and can show individuality, unlike mobile phones that can be hidden in the trouser pocket at any time. Small factories firmly believe that the market is so large that there must be room for them to survive.
In AnkerMake's view, within the next three years, the gap in the hardware supply chain of AI glasses will definitely be narrowed. "If you can make this hardware, I can too. There won't be any engineering problems," Chen Pan said. This is consistent with the view of Huang Hai, the founder of In Air. The Chinese hardware supply chain is "open - source".
It's not excluded that there may be some leading details that can be based on the resources of big companies. The photo - taking speed of Xiaomi's AI glasses is 0.87 seconds in some user tests, while the corresponding value of AnkerMake A1 is nearly 2 seconds. This more than one - second gap is the result of the joint adjustment between Xiaomi's Camera Department and the chip supplier, which is a kind of ability configuration that small factories don't have.
Chen Pan said that AnkerMake can achieve a time of more than one second with efforts, but this is not an effect that can be achieved by improving a single link. It requires the optimization of a whole complex chain including the camera, software algorithms, and hardware computing power. It is expected that this optimization ability may eventually be mastered by OEM factories, so there is no absolute barrier.
In terms of software positioning, many interviewees pointed out that at present, AI glasses are an accessory attached to mobile phones, and all interactions ultimately need to return to the mobile phone and the application entry.
This is correct, but Huang Hai believes that if startup companies want to get out of the situation of homogeneous competition with big companies, they must get rid of the "accessory" thinking, treat AI glasses as an independent entity, build an intelligent hardware with a complete application ecosystem operating system, and provide rich enough content and interactions - so that AI glasses can become the end - point of high - frequency behavior paths, rather than a transition.
This matter is more complex, but the possibility of building a barrier is also greater. If the threshold is too low, big companies will easily dominate the market when they enter.
Therefore, AnkerMake will focus on the software application experience of AI Agents in the future. The team has built Loomo OS and has invested nearly forty people in it. In Air has also established its own spatial operating system, and the ratio of the software and hardware teams is approximately 2:1. Huang Hai believes that even if big companies want to develop a completely identical spatial operating system, it will take at least more than a year.
It is a consensus among everyone that AI applications will become the core of products in the future, but it is still difficult to establish differentiation at present. Its core capabilities are concentrated in several major directions: recognition, organization, memory, retrieval, and translation.
Xiaomi also racked its brains to innovate an AI application, but it kept cycling through the process of looking and negating. Because at present, AI applications mainly improve efficiency and cannot create an opportunity similar to mobile payment or short - video platforms, that is, the long - awaited Killer App in the industry.
It is also difficult to draw a clear gap at the model - calling level. Almost all manufacturers have the same idea: build a "transfer box" that integrates different model interfaces, and according to user needs and model performance evaluations, throw different questions to different models.
Li Chuangqi believes that the top priority for AI assistants is to understand people better, rather than just staying at the level of tool - calling and function - control. If the chips and underlying models evolve at an accelerated pace, more sensors can be introduced to deepen the perception of people, so that System 1 (fast thinking) of the human brain can be located on the glasses and mobile phones, and System 2 (slow thinking) can be located in the cloud.
Obviously, AI glasses are still far from their ultimate imagination. "If we aim for a full score, we can only get 60 points at best now," Li Chuangqi said.
Mature the Supply Chain: It Doesn't Matter to Lose Money, First Gain Popularity
The overall judgment of the supply chain by market entrants is similar - "immature". In March this year, at a closed - door meeting attended by industry management departments, the participating enterprises mentioned that the biggest contradiction in the AI glasses industry at present is that consumers' expectations far exceed the capabilities of the entire industry.
This makes the core logic of Xiaomi's selection of suppliers ultimately "choosing the most expensive ones". "This is not a preset. We've almost looked at all the solutions on the market and evaluated them all. We found that the most expensive ones are indeed the best," Li Chuangqi said.
Startup companies can hardly copy this logic, or sometimes they are not allowed to adopt it. AnkerMake also wanted to use Qualcomm's chips. The license fee for one item was one million US dollars. AnkerMake was willing to invest this money, but Qualcomm, which has a large number of orders from giant customers, didn't have time to serve them.
This made AnkerMake finally decide to use domestic chips, based on the Unisoc T760. To some extent, this is a concession. Because the difference in user experience between the two chips in high - frequency functions such as taking photos and videos is already very obvious.
"There's no other way. You have to make a concession," Chen Pan said. "You can either wait for Qualcomm for a year, but will the market wait for you?"
Each company's first - generation AI glasses products are more or less launched with the mentality of "losing money to gain popularity". Behind this, the most important goal before the market explosion is the running - in of production capacity and yield rate between brand owners and OEM factories.
Charting: He Miao; Data compilation: Zhou Mo
Since the end of last year, Xiaomi, which is good at supply chain management, has successively placed material orders. The inventory preparation time for some parts of the supply chain is as long as five months.
But there is no particularly reliable reference in China for how many units to stock. Xiaomi can only establish a sales model by deducing from the target market