The AI glasses battle awaits Xiaomi to issue licenses

"The AI glasses market has been lively for a while, but the real competition has yet to begin."

"I heard Xiaomi paid Goer for 100 million for AI glasses."

"Xiaomi's first phase of AI glasses invested 30 million; you should know that now, you can get an AI glasses solution in Huaqiangbei for 500,000."

"Xiaomi's AI glasses are personally overseen by Lei Jun and the head of Xiaomi's wearables; Xiaomi takes this very seriously."

"Xiaomi gave us 4 million to develop software for AI glasses."

……

In the past few months, discussions about Xiaomi's AI glasses have been incessant.

The reason hardware practitioners are collectively focusing on Xiaomi is that Ray-Ban Meta has rekindled market enthusiasm for AI glasses, yet there are almost no other products available for purchase on the market. Although some manufacturers held press conferences six months ago, their products are still hard to find.

Liu Yunlong, product manager at AI glasses ODM manufacturer Yijing Technology, told Lei Feng Network, "Currently, AI glasses still need some polishing in terms of functionality debugging, performance, audio, and photo effects. The earliest they can be released is Q2 2025, and some good products will come out in Q3." Yijing Technology has undertaken ODM demands from multiple AI glasses companies, and its insights somewhat represent the current state of the entire industry.

This means that mature mass-produced products will only be available on the market by Q2 of this year. The long-rumored Xiaomi AI glasses are also expected to be released in Q2. In other words, as all AI glasses players enter the market, they will face a direct impact from Xiaomi, making the competition incredibly fierce and suffocating.

Some manufacturers have attempted to get a head start. In late December last year, Shanjite Technology launched the AI Snap Mirror co-creation version, priced at only 999 yuan, announcing that it would officially ship around mid-January.

At that time, some industry players expressed concerns, stating, "The AI glasses produced by Shanjite Technology in such a short time may not perform well; if they are not usable, not being able to ship on time might actually be a good thing."

As it turned out, this concern became a reality. Before shipping, Shanjite officially announced that there were issues with the product's Bluetooth calling function, and there were also flaws in the software, which would be continuously optimized through OTA updates, leading to a wave of user dissatisfaction.

The AI glasses market has been lively for a while, but the real competition has yet to begin. Everyone is waiting for the starting gun, and Xiaomi is the most likely to pull the trigger.

01. The Dilemma of the Four AR Dragons: Path Dependence, Scene Misalignment, Valuation Constraints

The Four AR Dragons are pioneers in the smart glasses track, having delved into smart glasses-related technologies for many years and accumulated a certain level of brand recognition. In the eyes of some, they should naturally take the lead in the AI glasses battle, but the reality is quite the opposite The public is always generous with praise for technological innovation, but the real business world is exceptionally harsh, and not all innovations can yield returns. If technological innovation does not grasp the rhythm well, rushing too fast can easily lead to failure, while moving too slowly may result in being abandoned by the times. Hitting the right rhythm requires not only extraordinary business acumen from the founders but sometimes also a bit of luck.

The "Four Little Dragons" of AR are all staunch optimists, targeting the ultimate form of smart glasses from the very beginning—AR glasses. Among them, Yingmu Technology is particularly aggressive, being the only company among the four that insists on not making split BB (BirdBath) glasses, but only producing integrated waveguide glasses.

However, they all seem to lack a bit of luck. Whether it is the split BB solution or the integrated waveguide solution, both are limited by technological maturity, and the experience is hard to describe as outstanding. The BB solution is too bulky to meet daily wear needs, while the waveguide solution is constrained by mass production and how to integrate with prescription lenses.

The market has also provided honest feedback. Yingmu co-founder Lv Yifei admitted to Lei Feng Network: "The users purchasing AR devices are still mainly technology enthusiasts and users in vertical scenarios."

In 2024, global AR device sales are expected to be 500,000 units, remaining basically flat compared to 2023, indicating a ceiling has been reached.

In contrast, AI glasses players represented by Ray-Ban Meta are simplifying based on AR glasses, temporarily abandoning display functions or only providing simple monochrome information prompts. This is technically easier to achieve. Currently, Ray-Ban Meta's cumulative sales in 2024 have already surpassed 2 million units.

Ray-Ban Meta physical image

Faced with the curve-breaking success of Ray-Ban Meta, the AR "Four Little Dragons" are confronted with a difficult choice: should they adhere to long-termism and continue to stubbornly pursue AR technology, or bow to reality and first align with Ray-Ban Meta?

"The speed at which Ray-Ban Meta surpassed one million shipments is faster than that of the iPhone, indicating that AI glasses will be a trillion-dollar consumer battlefield. This will be a war of opportunism, and long-termists who have always been driven by technology cannot grasp the 100-fold growth in the next three years," said Li Nan, founder of Angry Meow Technology, during a conversation with Lei Feng Network. He shifted from the 3C accessories sector to the AI glasses track in 2024.

However, following Ray-Ban Meta means giving up their own strengths to engage in close combat with competitors, which not only offers no advantages but may also risk losing the AR mindset that has been painstakingly built up in the past.

Therefore, the AR "Four Little Dragons" appear to be very restrained in this wave of AI glasses frenzy: XREAL has yet to publicly release its AI glasses product, with founder Xu Chi emphasizing that "direction is more important than speed," indicating a reluctance to blindly chase traffic amid the AI craze. However, reports suggest that XREAL began research and project initiation for AI glasses as early as June 2024, even considering the addition of brain-machine interfaces as the ultimate solution for AI glasses.

Rokid continues to adhere to its display route, adding waveguide display technology to its new product, Rokid Glasses. Its founder Misa believes that Meta's AI glasses' lack of a display module is a relatively conservative approach.

Yingmu Technology and Thunderbird have compromised to varying degrees. At the end of December last year, Yingmu released the INMO GO2, which features dual-eye single green display, focusing on business scenarios. Shortly after, Thunderbird launched the display-less AI glasses Thunderbird V3, aligning its functional positioning and technological route with Ray-Ban Meta. However, it is understood that over 60% of Thunderbird's R&D resources are still devoted to AR display technology, remaining trapped in the ideal realm of display technology.

The hesitation and entanglement of the "Four Little Dragons" in the face of the AI glasses craze stem partly from the sunk costs formed by long-term investments in display technology, and partly from the "high valuation" that hangs over them like the sword of Damocles.

Currently, the valuations of the AR Four Little Dragons range from 4.5 billion to 10.5 billion yuan, with Rokid having been trapped by numerous betting agreements three years ago. XR industry expert Wang Zhen stated that high valuations mean there is no turning back, and the decision-making chain will be longer than that of typical startups, leading to the AR Four Little Dragons' passive decision-making, which may not necessarily allow them to benefit from the AI glasses boom.

Following in the footsteps of Ray-Ban Meta means forming a separate team and reinvesting in product development and R&D. Currently, the cash flow of the Four Little Dragons is not abundant, and high valuations narrow the exit paths, affecting new capital's confidence in entering. If "valuation inversion" occurs, they will be caught in a game with old shareholders. In contrast, the newly emerging AI glasses startups can enter the market with a lighter load.

"Next, everyone needs to find ways to make subtractions to survive the transition period and quickly enter the AI glasses track, which will likely take 3-5 years," Wang Zhen said.

02. The Dilemma of Startups: Short Time, Limited Funds, Difficult to Get on the Table

Compared to the AR Four Little Dragons, new-generation AI glasses startups like Yiwen Technology and Shanjiji are fortunate enough to have secured significant funding during the market's peak, while also being able to push products to market along the technology routes already validated by Ray-Ban Meta for commercial monetization, supporting the company's exploration of next-generation products.

However, their advantages end there.

To some extent, products like AI glasses are not suitable for startups to pursue. The latter lacks sufficient voice in both the "body" and "soul" of AI glasses A practitioner candidly told Lei Feng Network that whether the hardware of AI glasses can be well-made does not depend on startups, but rather on the upstream supply chain. A typical example is chips.

Currently, there are no mature chip solutions available in the market, which is a collective dilemma faced by AI glasses companies.

Liu Yunlong, product manager at Yijing Technology, told Lei Feng Network that Qualcomm's AR1 is recognized in the industry as the chip solution for AI photography glasses, featuring mobile-level ISP that can achieve smartphone-like shooting quality, and it has already undergone market validation, with Ray-Ban Meta using this chip.

Having reached a cooperation with Qualcomm in 2017, Yijing naturally chose to develop a full-chain solution for AI glasses based on AR1.

However, the drawbacks of AR1 are also quite obvious for startups. First, it is expensive, with a single unit priced at 70-80 USD, accounting for nearly half of the entire solution cost; second, even if small and medium-sized manufacturers are willing to invest heavily in AR1, they cannot obtain the goods.

"We want to launch AI glasses as soon as possible and have considered Qualcomm AR1. At the time of project initiation, AR1 was still in the exclusive supply period for Meta. We chose the Spreadtrum W517 because they were willing to support us, and on the other hand, the W517 had already been validated on the Xiaotiancai watch. We assessed that it was more suitable for AI glasses, and currently, there are no other solutions that have truly shipped, so the uncertainties everyone faces may be greater than those faced by Shanjiji," said Zhang Bo, CEO of Shanjiji.

Lei Feng Network learned that besides Qualcomm AR1 and Unisoc W517, a commonly adopted solution is represented by Hengxuan's MCU combined with MTK or ISP. Qualcomm 5100 with audio chips is also seen as an option, with Niumiao Technology and Zhijing Unknown adopting this solution. Among them, Niumiao Technology's AI glasses, originally planned for release at the end of November last year, have been postponed until today. Jianan Technology adopted Junzheng's chip solution, and after the launch event at the end of December, it is still in the process of upgrading and debugging, unable to ship.

If small and medium players are willing to wait, they can expect more mature chip solutions by the end of this year.

An investor familiar with the upstream and downstream of the industry chain, Li Wei, revealed that Hengxuan is expected to launch a main control chip integrated with ISP SoC by the end of this year. This solution will be cheaper and have better power consumption control, making it more suitable for AI glasses than AR1, and Huawei is waiting for this solution.

As for the software aspect, the large models that have sparked this wave of AI glasses are more of a game for the giants. The AI glasses startups can only choose which large model to connect to. Ultimately, the points where everyone can compete in differentiation are limited to product definition and interaction design, but this does not constitute a core barrier.

Compared to larger companies with more leverage, the later small and medium players come to the table, the fewer opportunities they have. The key is to stay at the table as long as possible. This has led to a strange phenomenon where various manufacturers compete to hold launch events, creating a lively atmosphere, but there are no mass-produced products available in the market Many people compare the war of AI glasses to the smart speaker battle from eight years ago, where Alibaba and Baidu entered the market using price to gain market share, nearly driving small and medium-sized companies that had just started making smart speakers to the brink of extinction.

As a participant in the smart speaker war, YueRan Innovation CEO Li Yong has believed from the beginning that AI glasses will inevitably be a field where major companies will collectively enter. To avoid a head-on collision with these giants, Li Yong decided not to consider making AI glasses when starting his business, but instead turned to creating AI toys that major companies do not pay much attention to.

However, unlike smart speakers, AI glasses represent a more fragmented market. Due to the fashionable and trendy attributes of glasses, along with the fact that everyone's face shape is different, there are countless styles of frames available in the market. AI glasses will ultimately present a similar pattern, which means that no single manufacturer can dominate the market. Small and medium-sized startups may not lead the market, but as long as they define their products well and design good interactions, it should not be difficult to carve out a share.

An investor told Lei Feng Network that the eyewear category has a strong logic for mergers and acquisitions. "Essilor and Zeiss have acquired over 50% of the companies in China's eyewear distribution channels. From the perspective of investment exit, investing in a team, waiting for it to build a brand, sell volume, and be acquired by a channel is also a good choice."

03. Why is Xiaomi the "chosen one" for AI glasses?

Considering various limitations, a senior industry player told Lei Feng Network that AI glasses are more of an opportunity for smartphone manufacturers, large model manufacturers, and national-level app manufacturers.

At this stage, AI glasses are naturally accessories for smartphones, relying on smartphones to provide sufficient computing power, so it is logical for smartphone manufacturers to make AI glasses. At the same time, AI glasses also require support from large models, and large model manufacturers inherently possess core competitiveness. Additionally, national-level app manufacturers also have certain opportunities, as they have a large user base, low promotion costs, and can customize suitable products based on user needs.

Among them, smartphone manufacturers undoubtedly have the greatest chance of success:

First, smartphone manufacturers have more advantages in supply chain management and more bargaining power with suppliers;

Second, the initial market education for AI glasses relies on offline experiences, and smartphone manufacturers already have complete offline channels that can balance experience and after-sales service;

Third, AI glasses will compete in image quality in the future, which is also a strong point for smartphone manufacturers, with imaging algorithm teams often consisting of hundreds or even thousands of people, possessing rich experience.

Looking at the layout of domestic smartphone manufacturers in AI glasses: OPPO and vivo are still reassembling their previously dissolved XR teams, Huawei is unable to use Qualcomm's AR1 due to U.S. sanctions and is still waiting for a mature chip solution, while only Xiaomi shows the greatest determination and the fastest progress.

From the leaked product configuration information, Xiaomi's first AI glasses have impressive competitiveness. According to the BOM disassembly forecast report from Wellsenn XR, Xiaomi uses the Snapdragon AR1 dual ISP, Sony IMX681, 1200 pixels, with a battery capacity of about 160 mAh, equipped with 2+32GB of storage, and the estimated comprehensive cost for each pair of glasses is about 1281 yuan In line with Xiaomi's consistent style, the launch of this AI glasses is expected to create a significant shock in the industry in terms of pricing.

In addition to the hardware itself, Xiaomi has also made arrangements in the area of large models. In June last year, Xiaomi's large language model MiLM officially passed the large model filing, and the related models will gradually be applied to Xiaomi's automotive, mobile phone, smart home, and other products. According to insiders close to Xiaomi, thanks to the large shipment volumes of products like Xiaomi speakers and Xiaomi TVs, the number of users for Xiaomi's large model has already exceeded ten million.

Beyond Xiaomi, ByteDance is another active super player in the AI glasses market. It is reported that ByteDance is having three internal teams—PICO, Oladance, and educational hardware—compete to develop AI glasses.

The decision-making logic behind ByteDance's development of AI glasses is still to find traffic entry points for its own large model application. In addition to internal competition, ByteDance is also widely seeking external cooperation, and it is said that projects on the crowdfunding platform KickStarter have basically all been swept through by ByteDance.

Many people are not optimistic about ByteDance's hardware endeavors. "Internet companies lack sufficient respect for hardware products, wanting to turn hardware into traffic entry points, but compared to the speed of expanding internet users and traffic growth, they underestimate the growth rate of hardware, ultimately resulting in nothing substantial." said an AI hardware entrepreneur.

From mobile phones to study lamps to VR headsets, ByteDance's repeated failures and attempts in hardware products somewhat confirm this point. This includes the recently popular AI headphones OlaFriend, which, after making a splash, seems to have returned to calm.

The ongoing craze for AI glasses continues to stimulate the nerves of more major companies. With the decline in sales of Baidu's Xiaodu speakers and Alibaba's Tmall Genie, Baidu and Alibaba have already reassigned members of their original smart speaker teams to work on AI glasses. The home appliance manufacturer Skyworth's subsidiary, Coocaa, established a separate AI hardware division at the end of last year and considered whether to develop AI glasses. After conducting research on the upstream and downstream of the industry chain, they decided to pause due to the lack of cost-effective mature solutions, and it remains uncertain whether they will restart in the future.

Among the many players that have emerged, Xiaomi is undoubtedly the strongest in terms of overall strength. However, business competition has never been a simple comparison of balance sheet strength; countless cases in business history have repeatedly proven that organizational agility has more survival value than asset thickness.

Although most AI glasses startups may ultimately struggle to escape the fate of being "accompanying runners" for major companies, there will always be those who are unwilling to accept this fate, and there will always be those who create miracles. This is the greatest charm of the business world.

Author of this article: Wu You, Source: Lei Feng Network, Original Title: "AI Glasses War, Waiting for Xiaomi to Deal Cards"

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk