Competition over smart glasses among Chinese tech heavyweights has intensified toward the end of the year. After Baidu and Alibaba’s Quark unit launched their respective products, Li Auto unveiled its own device, Livis, on the evening of December 3.
To seize early momentum, Li Auto skipped the usual preorder phase and shipped units directly, pricing its Zeiss-equipped artificial intelligence glasses as low as RMB 1,699 (USD 237.9). An employee at Li Auto told 36Kr that within two hours of the launch event, “inventory sold out, and nearly all prepared stock was gone.”
Beyond these visible competitors, more companies are quietly advancing their plans. 36Kr recently obtained exclusive updates on the roadmaps of several major players:
- Globally, Google now has two AI glasses in the proof-of-concept stage and has partnered with Qualcomm (chips), Samsung (reference design), and Foxconn (manufacturing).
- Among Chinese smartphone makers, Vivo has spent the past year refining its approach to AI glasses.
- Xiaomi, meanwhile, has long been preparing its second-generation AI glasses. After the departure of product lead Li Chuangqi, the project was transferred to Dou Zifei, formerly the company’s director of extended reality products.
- Several consumer electronics makers have also entered the field, including action camera brand Insta360 and home security camera maker Ezviz.
- Among major internet companies, Tencent has taken a more cautious approach, conducting internal research into AI glasses applications.
- ByteDance, by contrast, has been more aggressive. Beyond its Doubao phone, the company initiated two AI glasses projects last year. According to 36Kr, ByteDance’s glasses equipped with an optical display are slated for release in the fourth quarter of next year, supported by a proprietary spatial computing chip. Its non-display AI glasses are expected to launch in the first quarter.
AI glasses have become the latest device format carrying the weight of the tech giants’ broader ambitions. For consumer electronics companies, they are a natural extension of existing hardware expertise. For internet companies, they represent not only a critical entry point after competition in foundational models and applications, but also an entry point with significant implications for data collection.
Wrestling for data
“Li Auto’s priority should be building a Level 4 car first,” said Wang Xing, Meituan’s founder and a Li Auto investor, during a meeting in late 2024. His comment stalled the glasses project internally for several months. But in April this year, Li Auto founder Li Xiang decided to proceed anyway. Within the company, the cost of building AI glasses was compared to “the price of a bumper on a car.”
AI glasses, however, were only part of Li’s larger ambition. A person familiar with Li told 36Kr that he has “never believed Li Auto is merely a car company.” Li expressed interest in AI glasses three years ago and once held a private discussion about the category with Luo Yonghao, founder of the startup Thin Red Line. At the time, Li believed AI glasses were the best vehicle for collecting real-world data.
That belief aligns closely with today’s data bottleneck in AI. As models advance toward multimodal and world model architectures, online text and video data have become increasingly redundant, while real-world data remains scarce. In the future, such data may be the key differentiator, providing the access and volume needed to build robust physical-world models and unlock the next stage of scaling.
In the near term, user behavior captured on hardware devices is equally valuable. Historically, that data has come from phones, tablets, and computers—platforms protected by entrenched permission walls. AI glasses, however, serve not only as distribution terminals for large models but also as continuous data collectors that help refine those models. Every app invocation, interaction, and preference captured at the device level can, theoretically, feed back into model improvement.
Hardware makers play defense
For all-rounders like Alibaba, Google, and Baidu, expanding their hardware entry points is almost inevitable. Their ecosystems already include nearly every competitive element needed for the glasses race: computing power, foundational models, strong app ecosystems (search, payments, navigation), and massive user bases. Google even controls its own operating system.
AI glasses represent the only major hardware gateway missing from their maps, and this may explain why Alibaba and Baidu have adopted particularly aggressive approaches. A person at Alibaba said the company is pursuing a “saturation attack” in the category, covering both hardware and software. Song Gang, who leads smart terminal development at Alibaba’s intelligent information business, told reporters after Quark’s AI glasses launch that Alibaba plans to invest in the category continuously and for the long term.
A Baidu employee told 36Kr that AI glasses have been designated a key strategic product internally. After launching its Xiaodu glasses at last year’s Baidu World conference, the company spent the following year reevaluating the product definition.
Consumer electronics giants, by contrast, are taking a more measured approach. AI glasses combine elements of prescription glasses (or sunglasses), wireless earbuds, voice recorders, and action cameras. For these companies, entering the category is as much a defensive move as a search for a second growth curve. Industry insiders who have spoken with several smartphone makers told 36Kr that, aside from Xiaomi, companies such as Vivo and Oppo have been cautious over the past year.
Full-blown competition
Competition in AI glasses is now possible because hardware barriers have fallen sharply. As tech giants jumped in, the smartphone supply chain followed suit, including suppliers such as GoerTek, Luxshare Precision, and Foxconn. Several standard reference designs have already emerged. Among the many AI glasses launching late this year and early next year, a default configuration has taken hold, built around Qualcomm’s AR1 and BES2800 chips.
An industry veteran told 36Kr:
“China’s AI glasses supply chain now works at this speed: sample in three months, mass production in six, shipment in nine.”
This collective push has shortened development cycles and kept hardware costs in check.
“A walk through Huaqiangbei shows you everything: AI glasses with Bluetooth audio for RMB 35 (USD 4.9), AI glasses with cameras for RMB 499 (USD 69.9), and glasses with an all-green AR display for RMB 999 (USD 139.9),” the person said.
But when the hardware threshold drops so low that companies from nearly any background can cheaply and efficiently produce AI glasses, what becomes the industry’s true inflection point?
The harsh reality of the category is its low tolerance for flaws. Any weak link can sink an entire product, and several cautionary tales have already emerged this year. Xiaomi’s first-generation AI glasses saw return rates as high as 40%. The issues were not catastrophic: some stemmed from unoptimized Bluetooth connections, while others were related to poor chip selection or adaptation.
Players with deep resources are undoubtedly the strongest contenders in the race to build the most popular AI glasses. But this phase will ultimately test something less tangible: how much patience each company is willing to invest in securing the next major hardware gateway.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Qiu Xiaofen for 36Kr.