AI glasses: Retracing the path of smart speakers

Last year, when Baidu launched its AI glasses product, the industry was extremely excited. The entry of major Internet companies made many industry insiders believe that this was a sign of the eve of the explosion of AI glasses.

This year, almost everyone is waiting for the release of Xiaomi's AI glasses. The industry media interprets it as "expected to push the Chinese AI smart glasses market to a new level."

Those in the industry can clearly feel that, driven by big companies like Baidu and Xiaomi, the "Hundred Glasses War" is on the verge of breaking out. The current atmosphere is very similar to the context when smart speakers were just emerging.

As smart hardware, AI glasses seem to be treading the same path as smart speakers.

"Hundred Glasses War" Replicates "Hundred Speakers War"

Let's rewind the clock 10 years.

In June 2015, Amazon launched a "talking" speaker called Echo, which dethroned Sonos, the former king of the connected speaker industry, and also opened up a brand - new category of smart speakers. By 2016, Echo's market share in the smart speaker market reached a peak of 88%.

With Echo setting an example, tech giants also began to take action. Google launched Google Home integrated with Google Assistant, and Apple focused on integrating the sound quality ecosystem and launched HomePod.

In China, major tech companies also entered the fray and launched smart speaker products. Among them, the most well - known are Baidu's Xiaodu Speaker, Alibaba's Tmall Genie, and Xiaomi's Xiaoai Classmate. At that time, there were a wide variety of solutions, and there were more than a hundred speaker manufacturers on just one street in Huaqiangbei.

This scenario is very similar to the current situation in the AI glasses industry.

According to the Beige Smart Glasses Market Research Report, it is estimated that by 2029, the market size of smart glasses will reach 106.778 billion yuan, with a compound annual growth rate of up to 18.56% during this period.

The emergence and performance of Ray - Ban Meta have injected a shot in the arm into the market. According to statistics from Wellsenn XR, 2.34 million AI glasses were sold globally last year, of which Ray - Ban Meta alone accounted for 2.24 million.

Just as the "Hundred Speakers War" was triggered by Echo in the smart speaker industry, driven by Ray - Ban Meta, a large number of players have flocked to the AI glasses track, showing a trend of a "Hundred Glasses War."

According to public information and channel sources, as of now, at least 50 companies in China are promoting smart glasses projects. These players can be roughly divided into three categories.

Firstly, there are recently established startups focusing on AI glasses. This includes Hive Technology, an enterprise in the Xiaomi ecosystem, which has launched the Jiehuan AI audio glasses equipped with multiple large models. Wang Xiaoyi, the former CPO of JMGO, also has his latest startup project Even Realities related to AI glasses, and it launched its first product, the G1, in July last year.

Secondly, there are emerging manufacturers that rose during the previous wave of AR glasses, such as Thunderbird Innovation, Yingmu Technology, Rokid, Liweike Technology, Xingji Meizu, etc. Liweike Technology launched its first AI glasses, MetaLensChat, in the first half of last year. Xingji Meizu's StarV Air2 smart glasses have launched many AI functions, and Rokid promoted its upcoming new product, Rokid Glasses, at a government meeting.

Thirdly, there are major mobile phone and Internet companies represented by Huawei, Baidu, Xiaomi, and ByteDance. The products launched by these manufacturers include Huawei Smart Glasses 2, Xiaomi MIJIA Smart Audio Glasses, OPPO's Smart Glasses Air Glass 3, etc. Honor has also been reported to have applied for multiple trademarks related to smart glasses.

From the "full - color waveguide solution" promised by Thunderbird X3 Pro, to the latest iterative spatial computing module of Rokid, and then to Xiaomi's binocular display system, supported by various technical solutions, more than 40 AI glasses products were unveiled at the CES 2025 exhibition. According to the plans of various manufacturers, there are at least 50 new models waiting to be released this year.

Beware of Falling into the Trap of a High - Start and Low - Finish

AI glasses currently seem very popular. However, comparing with the development of smart speakers, there are also concerns about a high - start and low - finish situation.

After reaching its peak in 2020, the smart speaker market started to decline. In 2024, the monthly sales of smart speakers showed a double - digit decline compared with the same period last year, and the annual decline exceeded 20%. In the fourth quarter, due to government subsidies and the Double Eleven promotion, the overall decline rate slowed down a bit, but it still exceeded 10%.

The industry generally believes that the high - start and low - finish of smart speakers is because there has been no major breakthrough in the core functions and user experience of the products. The upgrades in voice assistants and sound quality are limited. After the first batch of early adopters have been covered, it is difficult to stimulate users' demand for replacement.

In a typical scenario, the core selling point of smart speakers is voice interaction, but most products perform poorly in long - sentence recognition, semantic understanding, and contextual response. Users often encounter problems such as answering irrelevantly, mishearing, and misoperating.

Digging deeper, although smart speakers are named "smart," they are not as intelligent as expected. Most products only support basic functions such as playing music and controlling household appliances, lacking in - depth active service capabilities. After users' novelty wears off, the frequency of use drops significantly, and they eventually become "electronic white elephants."

Looking back at the development of smart speakers, their fate may have been doomed from the day they were born.

Amazon developed Echo to find a carrier for its voice recognition technology, Google Assistant. Although the emergence of the technological wave created conditions for the development of smart hardware, even from today's perspective, there are still too many problems to be solved in voice recognition technology. The gap between "usable" and "easy to use" has not been bridged.

Smart speakers have demonstrated the development trend of a product with immature technology but a rapid entry into the market. In contrast, in the technical aspect, AI glasses seem to have more problems to solve.

From the user's perspective, AI glasses are first and foremost a pair of glasses, and then we can talk about other AI functions. This requires AI glasses manufacturers to find an optimal balance among weight, battery life, and functions. However, current AI glasses products have not yet emerged from this "impossible triangle."

A typical example is the Vision Pro. Although it is impeccable in terms of functions, its weight of nearly 650 grams makes it difficult for ordinary people to like, except for die - hard enthusiasts.

Some products subsequently launched by AI glasses manufacturers are committed to "weight reduction" and strive to make AI glasses closer to the weight of ordinary glasses. One of the main reasons why Meta Ray - Ban has been selling well is that it has reduced the weight to 49 grams. Thunderbird V3 has gone a step further and reduced the product weight to 39 grams.

Although the newly launched AI glasses on the market have reduced the weight to a certain extent, they are still heavier than ordinary glasses, which usually weigh 20 - 30 grams. Especially after adding lenses, depending on the lens power, the weight of AI glasses will increase by more than 10 grams, and wearing them for a long time will still cause discomfort.

As for battery life, almost all AI glasses still do not have a perfect solution. Meta Ray - Ban can maintain continuous shooting for a maximum of 4 hours. Thunderbird V3 claims to have a 7 - hour battery life and even reserves a function to attach a "charging case" to the glasses, but there is still a huge gap compared with the usage time of ordinary glasses.

Based on the above analysis, it is not difficult to see that AI glasses products are not fully finalized yet. The faster the industry develops, the greater the harm of the hidden dangers it leaves behind.

Xiang Wenjie, the co - founder of Lingban Technology, once said, "To achieve full - color, a wider range of display, better performance, longer battery life, lighter weight, lower price, and for large models to reach the expected capabilities, and to increase the annual shipment volume to 10 million or even 100 million units... This is what the entire industry needs to do in the next three to five years."

The Turnaround Brought by Large Models

Although smart speakers are currently in a development bottleneck, with the penetration of large models in various fields, they may enter a new stage.

Currently, Alibaba's Tmall Genie and Baidu's Xiaodu Speaker have both integrated their respective large models. The person in charge of Alibaba's Tmall Genie business center once said in an interview with the media, "The implementation of AI large models is a long - term process, a process of continuous growth and continuous satisfaction of people's needs, which is parallel to the development of hardware."

In this process, smart speakers will continuously unlock new skills for users and continue to bring surprises.

Compared with the previous generation of products, smart speakers integrated with AI large models are indeed more intelligent, with significant improvements in voice recognition, natural language understanding, and dialogue capabilities.

The aforementioned person in charge of the Tmall Genie business also predicted that future smart speakers will not be dull machines that only answer questions one by one. With the help of large models, smart speakers will optimize the voice interaction link, actively judge the current scenario, have multi - modal interaction capabilities, and be able to intelligently judge the current state, rather than just passively accepting.

Judging from some functions of the Tmall Genie X6, which has integrated Tongyi Qianwen, some features have been initially realized, and more diverse applications will be launched in the future.

AI glasses, being a late - comer, have more advantages in the implementation and application of large models.

With the rise of DeepSeek, it has created conditions for AI glasses to reduce training costs by using low - cost and high - performance models, and the ultimate result is to lower the entry threshold for AI glasses.

According to a research report by Northeast Securities, the open - source nature of DeepSeek allows developers to make in - depth customizations, and the API price is low, which is conducive to the popularization and explosion of edge - side AI.

In addition, Huatai Securities believes that as a brand - new consumer electronics category, the AI glasses platform does not have the situation where brand manufacturers (such as Xiaomi and Apple) need to coordinate with existing Internet companies (such as WeChat and Meta) to open AI traffic entrances, as in the smartphone ecosystem. Instead, it is more likely to be the first scenario for AI applications to be implemented.

With more powerful functions, lower prices, and interactive innovation after integrating cutting - edge technologies such as AI, AR, and eye - tracking, the industry generally believes that AI glasses have the potential to replace smartphones.

Mark Zuckerberg once said in a conversation with Jensen Huang that smart glasses will be similar to mobile phones and will be an always - on version of the next computing platform.

Li Hongwei, the founder and CEO of Thunderbird Innovation, also believes that just like smartphones, although there are still professional cameras, music players, smart watches, and other devices on the market, as a general - purpose computing platform, smartphones fully integrate these scenarios and create more diverse application possibilities.

By reverse - reasoning, AI glasses also have similar capabilities and even unique advantages in interaction, so they naturally have the potential to become the next - generation general - purpose computing platform.

* The pictures are from the Internet. Please contact us for deletion if there is any infringement.

This article is from the WeChat official account "Insight New Research Society" (ID: DJXYS - 0309), author: Chen Wen. It is published by 36Kr with authorization.