In 2025, AI glasses stepped off the pedestal.
AI glasses are experiencing a roller-coaster fate.
According to IDC data, in the second quarter of 2025, the global shipment volume of smart glasses reached 2.555 million units, a year-on-year increase of 54.9%. The performance of the Chinese market was even more remarkable, with shipments reaching 664,000 units, a year-on-year increase of 145.5%. Major manufacturers have been actively deploying in this field. From Xiaomi to Alibaba, and from Rokid to Thunderbird, almost all relevant enterprises have announced their smart glasses plans this year.
In the first half of this year, the sales volume of smart glasses on the JD platform increased by more than 10 times year-on-year, and the number of settled brands soared by about 3 times. On average, a new product was launched every 9 days.
However, behind the bustling scene, a crisis is looming. According to statistics from industry institutions, the return rate of AI glasses on platforms such as JD and Tmall is about 30%, and on the Douyin platform, it is as high as 40% - 50%. The core issue reflected by consumers behind the high return rate is the "lack of functional practicality".
The once highly-praised AI glasses are now facing unprecedented doubts.
On the Eve of Deflating the Bubble
Xiaomi's AI glasses are almost a typical example of this bubble. In the first week of sales, 70,000 pairs were sold. However, according to public information, the return rate on the Douyin platform is about 40%.
It is understood that from project initiation to product launch, Xiaomi's AI glasses only took about a year. However, this "quick" approach has led to a sacrifice in user experience and a high return rate. The reason is that the product is more like a simple combination of "headphones + camera", lacking the innovation to truly solve users' pain points.
A consumer decided to return the product after using it for 5 days. He frankly said, "Xiaomi's glasses with only a camera and no lens display are more like a simple combination of 'headphones + camera', and the actual experience in real scenarios did not meet my expectations." For example, when using the navigation function, due to the lack of visual display, it was difficult to accurately identify the right path at intersections relying only on voice prompts. Eventually, the user still had to take out the mobile phone to check the map.
A tech blogger found during the test that by simply smearing with a marker pen, the shooting indicator light of Xiaomi's AI glasses could be successfully blocked. It was difficult to tell whether the indicator light was on from the appearance. After turning on the shooting function, people around could hardly notice that the glasses were in the recording state, which raised serious privacy concerns.
Of course, this is not an isolated case. The highly anticipated "Battle of a Hundred Glasses" in 2025 did not happen as expected. The lack of industry standards has become a major drawback.
"Currently, no manufacturer has been able to integrate all aspects of AI glasses, from materials, optics, hardware, software, systems to algorithms," said Cui Haitao, the chairman of Gudong Intelligence, pointing out the current situation of the industry.
The ideal is full, but the reality is skinny. According to industry insiders, in the early stage of the industry, players generally benchmarked against Meta's Ray-Ban. The product design was often influenced by MetaBan from the start, but the focus was more on rapid market commercialization, while the actual products fell far short of users' expectations.
"The core of AI glasses lies in achieving the in-depth service value of 'data precipitation + auxiliary decision-making' based on the technology of 'all-day wearing + visual and auditory perception'." However, from AI algorithm companies, traditional eyewear manufacturers, mobile phone manufacturers, audio terminal providers, Internet companies to telecom operators, there are many players from different camps. But most of them tend to simply stack product hardware, lacking revolutionary innovation in essential functions.
Li Chuangqi, the person in charge of Xiaomi's wearable device department, frankly said that Xiaomi's AI glasses almost used the most expensive components. Although there was a profit margin at the price of 1,999 yuan, it was still a loss considering the overall investment.
Different players with different backgrounds are flocking to this track, each with their own agendas.
Internet giants like Alibaba launched the Quark AI glasses, attempting to "wear the entire Alibaba ecosystem on users' faces" - integrating functions such as Tongyi Qianwen's large model, Gaode Navigation, Alipay payment, and Taobao's image search.
After a series of organizational structure adjustments since the end of last year, the AI application "Tongyi" originally belonging to Alibaba Cloud was merged into the Smart Information Business Group where Quark is located. At the same time, Alibaba promoted the integration of Tmall Genie and Quark products and explored new hardware directions such as AI glasses, which were coordinated by Wu Jia, the CEO of Alibaba's Smart Information Business Group.
Media reports indicate that the standard version of Quark AI glasses is sold at 2,999 yuan, which is on par with the cost. If the company doesn't make money from the product itself, it is speculated that Alibaba's ambition lies in charging for ecological functions and scenarios.
Mobile phone manufacturers like Xiaomi regard AI glasses as part of their "full ecosystem of people, cars, and homes" strategy, hoping that they will form the three core terminals of the Xiaomi ecosystem together with mobile phones and cars.
Start - up companies such as Rokid and Thunderbird choose to focus on specific fields. Zhao Weiqi, the person in charge of Rokid's global developer ecosystem, believes that "the current AI glasses track is still in a state of healthy competition."
However, since the industry has not yet established unified standards and regulations, "issues such as regulating AI usage and avoiding privacy violations are still in the stage of discussion and exploration," said Wang Yiru, the financing director of XREAL. Currently, only some risks can be predicted, and the problems will have to be solved while the industry is developing.
The technology iteration in the AI glasses industry is accelerating, and brands need to seize the opportunities in the supply chain and ecological cooperation. However, the reality is that the industry has fallen into a strange cycle of "high investment, low return, and difficult implementation".
Why Are AI Glasses Progressing So Slowly?
Rokid Glasses made a stunning appearance at this year's World Artificial Intelligence Conference, attracting a large number of audiences to queue up for an experience. However, consumers soon found that there was a huge gap between the ideal and the reality - it tried to replace mobile phones, but the actual experience could not be separated from mobile phones.
Production capacity is also a major problem. Many brands have released amazing demos, but their actual delivery capabilities are far from meeting the demand. Some consumers reported that after pre - ordering Rokid glasses, they had to wait for more than half a year. So far, the production capacity is still limited. Finally, when a part of the products were launched in September, the price soared from the pre - ordered 2,499 yuan to 3,299 yuan.
Industry insiders frankly admit the current situation. Consumers expect AI glasses to have functions such as multi - modal recognition, making phone calls, recording videos, and live - streaming, and at the same time, they should be fashionable and comfortable to wear. The requirements for manufacturers are extremely high, but in reality, the existing technological capabilities are indeed difficult to meet these demands.
This is directly related to the immaturity of the AI glasses supply chain. Since the sales volume of AI glasses is still low and the experience is not satisfactory, the corresponding chips, edge - side AI algorithms, new materials, etc. are not yet mature. Manufacturers also do not have a separate production line for AI glasses in the mechanical production and processing stage.
The industry generally faces the dilemma of the "impossible triangle" - it is difficult to balance performance, lightness, and battery life.
Stronger computing power requires a chip with higher power consumption, which will shorten the battery life. Longer battery life requires a larger battery, but this will increase the weight. Lighter weight is a prerequisite for all - day wearing, but it will limit the battery capacity and heat dissipation space.
Currently, no manufacturer has been able to integrate all aspects of AI glasses, from materials, optics, hardware, software, systems to algorithms.
In terms of algorithms, since there is no dedicated AI algorithm for glasses, AI glasses can only use small models on the mobile phone side, and integration is difficult. In terms of optical modules, they currently support medium - sized spectacle frames, but not large or small ones.
Both Meta and domestic manufacturers mostly adopt "general - purpose" solutions such as Qualcomm's Snapdragon AR series. These chips are essentially variants of mobile phone SoCs and are not designed for the extreme scenarios of AI glasses, which need to be "always on and ultra - low power consumption".
Some industry insiders have bluntly stated that the entire industry is waiting for a dedicated chip that can handle complex AI tasks with milliwatt - level power consumption. Without this "heart", AI glasses will never be able to get rid of the embarrassing situation of being a "power bank accessory".
The deeper problem is that AI glasses lack a real "killer application". Most of the current functions, such as translation, navigation, and taking photos, can be better performed by mobile phones. AI glasses attempt to combine the functions of "headphones + camera + information assistant", but in each individual function, they cannot provide an experience that surpasses that of mature devices.
A tech blogger described his experience after using the image recognition function of AI glasses: He had to take a photo and ask the AI. After waiting for a few seconds to more than ten seconds, the glasses could give feedback. This experience was far from what the manufacturer advertised as "ask the AI whenever you have a problem".
Battery life is also a major problem. A consumer complained, "The shooting time is obviously insufficient. The continuous usage time is only about 2 - 3 hours."
In a real sense, AI glasses are still in their infancy, but the industry has already gone through a reshuffle. In contrast, some products focusing on optimizing existing functions, such as AI voice recorders, have received a good market response.
The reason behind this contrast is that the latter solves specific pain points and provides a truly practical user experience, rather than just talking about concepts.
The Imagination and Reality of AI Glasses
The dilemma faced by AI glasses is essentially a mismatch between the supply chain capabilities and market demand.
On the one hand, the functions promoted by manufacturers are not what consumers really need, and the experiences consumers expect cannot be realized due to technological limitations. On the other hand, manufacturers are painting a picture of the "next - generation computing platform", while consumers are becoming more and more clear that they need tools to solve practical problems.
According to market feedback, AR glasses with displays have dominated the market. Statistics from CINNO Research show that in the first quarter of 2025, the sales volume of consumer - grade AI/AR glasses in the Chinese market increased by 45% year - on - year.
Among them, AR glasses with displays account for 80% of the market share, while AI glasses without displays only account for 20%. This indicates that users' demand for glasses with display is far more important than AI.
The optical module accounts for up to 40% of the total cost of the device. The mainstream MicroOLED and MicroLED + diffractive waveguide solutions are expensive, with complex processes and low yields, which have kept the price of consumer - grade products above 3,000 yuan for a long time.
This has also made most manufacturers stop at small - scale trials.
The high cost not only restricts the willingness for mass production but also keeps the user group limited to geeks, niche creators, etc., making it difficult to achieve large - scale popularization.
At the same time, the sales channels are also in pain. When traditional eyewear stores sell smart glasses, they need to overcome difficulties such as consumers' product awareness, inventory management, and after - sales service. Most traditional consumers focus on vision correction. If eyewear stores sell smart glasses, they need to spend a lot of time introducing the functions and usage methods.
However, challenges and opportunities coexist.
Meta's Ray - Ban Display was almost sold out within 48 hours after its launch in the US market. The on - site experience appointments at most stores have been booked until November. The original price of the product is $799, but the resale price on Xianyu has soared to 11,000 yuan, with a premium of nearly 100%. This shows that as long as the product is excellent enough, the market demand is real.
In the B - side market, AI glasses are showing greater potential. On the streets of Hangzhou, traffic police wearing Rokid AI glasses can lock in illegal vehicles in 3 seconds. The law - enforcement videos have spread from local media to national media such as CCTV. This has opened up the application space of AR glasses in B - side/G - side scenarios, such as law enforcement and operations.
Market segmentation may be the way to break the deadlock. Research institutions point out that smart glasses are challenging traditional devices such as tablets, portable displays, smartphones, and action cameras in specific application scenarios through market segmentation strategies.
Data from the first half of 2025 shows that the market shares of four types of segmented products, namely audio glasses, shooting glasses, split - type AR glasses, and integrated AR glasses, in the Chinese online retail market are 42.3%, 12.4%, 41.1%, and 4.2% respectively. Different product routes determine their respective core application scenarios and competition targets.
In the long run, the AI glasses industry needs to return to rationality and find its own development rhythm. According to industry insiders' speculation, in the next three years, three types of dominant enterprises may emerge: large - model holders, hardware ecosystem integrators, and scenario solution providers.
A more practical development path is that when the industry has a large enough order volume and clear technological requirements and trends, suppliers will be motivated to make technological breakthroughs and invest in production capacity. In the future, a separate glasses supply chain may be formed.
For the industry, as cross - border players continue to pour in, the industry will accelerate the shift from "functional competition" to "ecological war". The most important thing now is to re - think the positioning of AI glasses.
It may not necessarily be a replacement for mobile phones but could be a new type of device - an extension of mobile phones, providing specific value in specific scenarios.
Data from market research institutions show that the long - term potential of AI glasses is still highly regarded. The shipment volume of AI/AR glasses in China is expected to reach 900,000 units in 2025, and the industry scale will exceed 100 billion yuan in three years. Technology is only real when it is applied to real life.
The emerging technology industry always goes through a cycle from overheating to rationality. An industry insider frankly said, "The bubble is not a bad thing. It exists in any emerging industry. People of different qualities will jump into this pool to seize opportunities. We should view the bubble with patience and confidence."
In the next 3 to 5 years, the AI glasses market will explode, and the cooperation between more brands and manufacturers will promote the diversified iteration of product forms and functions. After the "Battle of a Hundred Glasses", each company will find its own ecological niche.
The short - term cooling of the market may be the beginning of long - term rationality.