AI Glasses: The First Battle Among Major Manufacturers

On the surface, the competition among big manufacturers in the field of AI glasses seems to be a flourishing scene, but actually, there are hidden concerns.

Recently, a category once considered niche and luxurious is quietly gaining unprecedented attention: AI glasses. Baidu has just released the Xiaodu AI Glasses Pro, which integrates AI object recognition, real - time translation, and first - person shooting; Rokid has introduced its AR glasses originally designed for industrial scenarios into daily life, focusing on lightweight recording and intelligent interaction; Xiaomi also officially unveiled its Xiaomi AI Glasses at its "Full Ecosystem of Vehicles, Homes, and Wearables" strategy press conference, attempting to break the boundary between wearables and smart homes. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_2b7e3714c954437fb1c74917bdde089e@000000_oswg665968oswg1080oswg720_img_000?x - oss - process=image/format,jpg/interlace,1" > According to incomplete statistics, nearly 70 companies have flocked to this field. From internet giants to native XR companies, the intensity of their actions and the extensiveness of their layouts almost replicate the situation on the eve of the explosion of smartwatches or TWS earphones back then. The International Data Corporation (IDC) said that it is estimated that the global shipment of smart glasses will exceed 23.687 million units in 2026, among which the shipment in the Chinese smart glasses market will exceed 4.915 million units. So, the question arises: in today's era when smartphones are highly popular, why are technology giants still willing to invest resources in a field where clear user habits have not yet formed? Is it a new carrier for the implementation of AI, or an ideal product that requires a long education cycle? < h2 >The Rise of AI Glasses AI glasses are not a new species. As early as 2012, Google Glass tried to initiate a wearable revolution, but it quickly faded away due to privacy disputes, useless functions, and high prices. In the following more than a decade, the industry fell into a long technological hibernation period: the cost of optical modules remained high, battery life was difficult to improve, AI computing power could not be localized, and the intelligence of glasses had not yet formed an irreplaceable use value. However, after 2023, several key variables emerged simultaneously: breakthroughs in edge - side large models, the maturity of the supply chain, and the approaching ceiling of the smartphone market. The giants urgently need a physical carrier that can carry AI capabilities and be used frequently. AI glasses happen to meet these requirements. So, we saw Baidu publicly release the Xiaodu AI Glasses Pro; Rokid abandoned the pure B2B route and launched glasses supporting real - time translation and AR navigation; Xiaomi promoted the glasses with the selling point of ecological synergy, enabling seamless linkage between the glasses, mobile phones, and cars. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_f970caa65c044207981d8b984a18a1f2@000000_oswg232930oswg739oswg445_img_000?x - oss - process=image/format,jpg/interlace,1" > However, different from previous years, this year's R & D strategy for AI glasses is more focused and practical. Manufacturers are no longer pursuing showy features but are returning to a simple question: what should a good pair of AI glasses really look like? The premise of intelligence is usability and ease of use. A good pair of AI glasses must first be a pair of wearable glasses - with the weight controlled within 50 grams, the nose pads fitting the Asian face shape, and the battery life lasting at least one day's commute. Secondly, it doesn't have to be all - powerful, but it must be reliable enough in several key scenarios, such as being able to understand the translation just by looking at the menu when traveling abroad, automatically transcribing conversations and generating meeting minutes during meetings, or identifying obstacles in front for the visually impaired. The more focused the scenarios are, the greater the value. Finally, while providing basic functions such as voice and recognition, the interaction should be almost imperceptible. Users don't need gestures, gazes, or complex instructions. They just need to naturally say "What's this?" or look at an object, and the service will follow. In other words, today's AI glasses are undergoing a profound process of demystification. They can no longer simply pursue showy features but should rely on the large model capabilities of cloud - edge collaboration to provide definite value with extremely low interaction costs. < h2 >The Battle of a Hundred AI Glasses There was the battle of a hundred large models before, and now there is the battle of a hundred AI glasses. In the AI glasses battlefield in 2025, no two companies have exactly the same strategy. This competition, seemingly centered around a pair of lightweight glasses, actually reflects the different corporate genes, resource endowments, and strategic focuses of different companies. Some bet on the semantic understanding ability of large models, some rely on the synergy effect of the hardware ecosystem, and some choose to deeply cultivate vertical scenarios to leverage the mass market with professional capabilities. The competition is no longer about whose screen is brighter or whose chip is faster, but about who can more naturally embed AI into people's real life flow. Roughly speaking, the participants can be divided into three forces: internet companies with large models, terminal manufacturers with hardware ecosystems, and native XR companies that have been deeply involved in the XR field for many years. They each enter the field with their own advantages and also carry their own unavoidable shortcomings. Internet giants such as Baidu, ByteDance, and Alibaba are obviously aiming for the intelligent ecosystem. They are not obsessed with creating the lightest and thinnest glasses but hope to turn their self - developed large models into personal assistants that users can't do without. Baidu's Xiaodu AI Glasses rely on the Wenxin Yiyan large model and can understand dialects and vague references; Alibaba's Quark AI Glasses are deeply integrated with Alibaba's applications such as Gaode Navigation, Alipay's "Scan to Pay", and Taobao's real - time price comparison, and build a closed - loop ecosystem of "life services + commerce" through the Tongyi Qianwen large model and Quark AI technology. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_801305e008424bb4a869de0a425ccf27@000000_oswg197888oswg704oswg435_img_000?x - oss - process=image/format,jpg/interlace,1" > These companies have a large user base, a mature digital service ecosystem, and a deep understanding of high - frequency scenarios. However, due to the lack of hardware engineering experience, details such as quality control, mass production, and wearing comfort are often not as good as those of hardware manufacturers. For example, the Xiaodu AI Glasses Pro do not follow the "AI + AR" route and do not have the mainstream screen display function in the industry; the Alibaba Quark Glasses weigh 42 grams, which is slightly inferior to manufacturers like Xiaomi that are good at hardware in terms of lightweight design. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_1658f2855ade4d50b3be646febf534b1@000000_oswg114224oswg1080oswg754_img_000?x - oss - process=image/format,jpg/interlace,1" > The second type of players are hardware ecosystem manufacturers represented by Xiaomi. They may not be leading in large models, but they have a large device matrix, a mature supply chain, and a user base of hundreds of millions. For them, AI glasses are not an isolated new product but a sensing node in the "full ecosystem of vehicles, homes, and wearables". However, these manufacturers are weak in core AI capabilities and cannot meet the professional needs of geeks during the product trial period. For example, although the Xiaomi AI Glasses sold 70,000 pairs in the first week of their release, seemingly popular, the return rate on the Douyin platform was as high as 40%. Users complained that the shooting was like blind shooting and the voice assistant was easily mis - activated. The third type are consumer - grade AR manufacturers such as Rokid and XREAL. In the past decade, they have continuously invested in near - eye display, spatial positioning, and human - factor interaction and have accumulated profound underlying technologies. Rokid was deeply involved in industrial AR in the early years and is now feeding these capabilities back to the consumer market; XREAL adheres to the split design and focuses on creating an immersive movie - watching experience after connecting to a mobile phone. However, these native XR companies are also facing the dilemma of a weak ecological foundation and difficult commercialization. As independent technology manufacturers, Rokid and INMO cannot build a powerful usage ecosystem like Huawei and Xiaomi. Their functions are single, and the compatibility in different systems is insufficient. Due to the lack of an ecosystem, the overall sales of these native XR companies are currently difficult to compare directly with those of technology giants. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_09029dd764ed4852a6fd8a91dbf36b88@000000_oswg71964oswg1080oswg182_img_000?x - oss - process=image/format,jpg/interlace,1" > It can be seen that different companies have different strategic characteristics, but the focus of competition is similar: whether they can identify high - value scenarios while having natural interaction. For example, Alibaba's Quark AI Glasses were able to attract users during the internal testing phase and topped the Tmall XR device Double 11 hot sales list during the pre - sale period. The reason is that it accurately tapped into the scenario of searching for answers to exam questions by taking photos. Voice activation is more convenient than uploading photos through an app, and the cool AR prompt projection has also stimulated children's interest in learning. Learning with just a blink of an eye, this core selling point has directly unlocked the huge consumer market of exam - oriented students. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_c2ecfb8d56d34a6eafe208f188558670@000000_oswg230558oswg677oswg751_img_000?x - oss - process=image/format,jpg/interlace,1" > The Ray - Ban Meta, which is very popular overseas, has accurately unlocked the fashionable social scenario. Among the 2.34 million units of global AI glasses sold in 2024, the Ray - Ban Meta accounted for 95.7%. The success of the Ray - Ban Meta proves that when the technology is not yet disruptively leading, the product must first be a pair of good - looking glasses. Meta cooperated with the famous eyewear brand Ray - Ban, and the fashionable appearance has a high acceptance rate. At the same time, it can share the captured content to social platforms such as Facebook/Instagram with one click, which solves the convenience and sharing desire of first - person recording and forms a core rigid demand. Whether it is the social sharing of the Ray - Ban Meta or the payment and navigation closed - loop that the Quark Glasses are trying to build, successful products all try to provide unique and irreplaceable value in one or a few specific scenarios; products that are temporarily lagging behind all exaggerate the toy attributes and have useless practical value. It can be seen that the competition of AI glasses is essentially a competition of how to encapsulate AI capabilities into real life. Whoever can identify high - value scenarios earlier and transform them into natural interactions without learning may win the first batch of loyal users. < h2 >The Long and Difficult Road to the Implementation of AI Glasses The road to the implementation of AI glasses is much longer and more rugged than expected. In the past few years, it has indeed stepped out of the closed loop of the laboratory and is no longer just a novelty for a small group of geeks. With technological upgrades, it has begun to frequently appear in the core positions of international exhibitions such as CES and MWC, been written into the annual strategic reports of many technology giants, and become a new narrative thread in the capital market. According to the latest data from IDC, the global shipment of AI glasses in the first quarter of 2025 was 1.487 million units, a year - on - year increase of 82.3%, and the growth rate in the Chinese market was as high as 116.1%. < img data - img - size - val = "" src = "https://img.36krcdn.com/hsossms/20251121/v2_f4e182d8ca264b23a60e0fd55eb0b33b@000000_oswg318720oswg824oswg500_img_000?x - oss - process=image/format,jpg/interlace,1" > However, although the competition of AI glasses among large manufacturers seems to be in full bloom on the surface, there are hidden worries. Most of the AI glasses on the market are still at the trial - product stage. The monthly active user retention rate of mainstream products is generally low; a mainstream AI glasses product fell into a return rate dilemma of as high as 40% - 50% one month after its release. Consumers often complain that the price is high but the experience is fragmented, feeling that they have been deceived by another IQ tax. Without a stable interaction paradigm, it is difficult to have a prosperous application ecosystem. Specifically, the entire industry has not yet formed a unified interaction language. Some control the menu by blinking, some rely on specific wake - up words, some must transfer instructions through a mobile phone app, and even the operation logic of different models of the same brand varies greatly. This fragmented experience not only increases the user's learning cost but also discourages developers. All signs indicate that AI glasses are still in the critical climbing period of product verification and are still a long way from real mass popularization. What it needs is not more press conferences and PPT demonstrations but respect for human - factor engineering, in - depth cultivation of real scenarios, and continuous exploration of "why users are willing to wear them every day". This also reflects that technological feasibility does not equal product usability, and capital enthusiasm does not equal user acceptance. Looking to the future, AI glasses may experience a cruel reshuffle. This article is from the WeChat public account < a target = "_blank" rel = "noopener noreferrer" href = "https://mp.weixin.qq.com/s?__biz=MzUxNTUyMjE4Mw==&mid=2247532257&idx=1&sn=78d9ff93a3eeebd7222bf09d708b94cb&chksm=f85f4356bafa46980ba9ac6c4e96511271ad2659597869fee83729f8176fdbe3a8c1e19cd76a&scene=0&xtrack=1#rd" >"Brain Intelligence" (ID: unity007), author: Shanhu. It is published by 36Kr with authorization.

该文观点仅代表作者本人，36氪平台仅提供信息存储空间服务。

AI glasses, the first battle among major manufacturers