The Hundred-Mirror War in AI Glasses: Return Rates of 30%-70%, and the Industry Stuck Between Early Adoption and Daily Use

As 2026 draws to a close, there is no sign of the AI glasses race cooling down.

Rokid released three new products in one go at the end of October; Thunderbird Innovation (RayNeo) officially announced its fourth-generation full-color AI glasses the same day; the Qwen N1 series entered mass production and delivery in October; and with Huawei, Xiaomi and other companies having previously entered the field from adjacent sectors, domestic AI glasses have ushered in a veritable "battle of a hundred glasses."

Looking at the data, growth is proceeding in tandem. Counterpoint Research data shows that in the first half of 2026, global AI glasses shipments grew 263% year on year.

The excitement is real, but if you look away from the frequency of new product launches and shipment volumes and toward the pair of glasses sitting on users' noses, you see a different set of numbers.

Data released by RUNTO Technology shows that in the first half of 2026, the industry's return rate was as high as 30%-70%. Many users still experience dizziness and discomfort when wearing AI glasses.

At the summer launch event in May this year, RayNeo CEO Li Hongwei acknowledged an awkward reality for the industry: over the past two years, the industry has launched more than 200 new AI features, but users' long-term usage rate is less than 6%; the remaining 94% of users still stick to the most basic functions such as asking about the weather, stocks, and news.

The "iPhone moment" for AI glasses seems farther away than the capital-market narrative suggests.

1. The battle of a hundred glasses has no true "number one"

English text is being prepared.

On October 27, Rokid launched three new AI glasses products covering different price points in one go, from a screenless lightweight entry model to a full-color binocular display flagship, while simultaneously launching a companion interaction ring.

Rokid Glasses 2, image source: Rokid WeChat official account

On the same day, RayNeo officially announced its fourth-generation full-color AI glasses, "RayNeo iO+", featuring an on-device large model and extreme weight reduction, with an appearance that emphasizes everyday wearability.

On October 16, INMO Tech released the INMO GO3; Huawei also completed deep HarmonyOS ecosystem adaptation and AI capability OTA upgrades in October...

Earlier, on October 13, Alibaba's Qwen released the N1 and N1 Pro AI glasses and AI clip-on earbuds, all equipped with a heterogeneous three-chip platform of Snapdragon AR1, BES, and Himax.

Looking back at the mainstream new products launched by players over the past 10 months, it is not hard to see that the commonalities at the product level are more striking than the differences.

On-device large models or on-device AI Agent capabilities have become standard at launch events, and full-color MicroLED optical waveguides have become the collective direction of this round of high-end new products.

In this round of launches, Rokid Color and RayNeo iO+ coincidentally made "full color" a core selling point. In 2025, monocular information prompts were the mainstream; by 2026, binocular displays have become standard on mid-to-high-end models, upgrading from monochrome text prompts to full-color video playback.

In addition, lightweight design continues to be pushed to the extreme, and the weight race is intensifying. The screenless Rokid Style weighs 38.5 grams, the Li Auto Livis about 36 grams, and the RayNeo V4 is kept at 38 grams; mainstream display-equipped models are also generally pressed to within 70 grams.

Technologically, multi-chip heterogeneous solutions are becoming widespread, with players beginning to use chips of different process nodes and architectures to handle imaging, everyday tasks, and low-power sensing separately, trying to find a balance among computing power, power consumption, and battery life.

Undeniably, supply chain maturity and technology trickle-down will naturally lead to configuration convergence. The problem is that when all players, in a similar time window, using a similar supply chain, tell a similar story, brand competition will only become more intense.

Capital is the accelerant of this competition. Incomplete statistics from RUNTO Technology show that in the first half of 2026, there were at least 64 financing, M&A, and strategic investment deals in the global smart glasses industry chain, with a disclosed total amount of about 9 billion yuan.

Behind the seemingly contradictory rankings is the intensity of industry competition.

At present, AI glasses are still in an early stage where no absolute leader has emerged. "Industry number one" is both an endorsement of brand strength and a core bargaining chip for gaining supply chain pricing power, capital favor, and user mindshare. Therefore, brands coincidentally choose the statistical dimension most favorable to themselves to define the battlefield, turning differences in methodology into a competitive tool.

XREAL, citing iResearch's "2025 Global AR Glasses Industry Research Report," uses global consumer-grade display-equipped AR glasses sales revenue as its statistical basis and claims to have ranked first in the global market for three consecutive years;

RayNeo, based on IDC and Counterpoint data, says that in the first half of 2026 RayNeo ranked first in global AR glasses shipments, and in the Chinese market, RayNeo has maintained the number one AR glasses sales position for five consecutive years;

Rokid cites NielsenIQ retail monitoring data for the first three quarters of 2026, saying that in the AI display smart glasses sub-market, Rokid has ranked first in the Chinese market for three consecutive quarters;

Xiaomi, citing Counterpoint's Q1 2026 market tracker report, takes first place with a 28% share under the shipment basis of all categories of smart glasses (including screenless audio glasses).

Viewed from another angle, the true number one seems not to have appeared. At present, no brand has established an absolute advantage simultaneously across all categories, all channels, and all regions, nor has anyone been able to pull decisively ahead through overwhelming product experience.

2. Privacy controversies, a battle over roadmaps—how far are AI glasses from daily life?

What is more worth pondering than who is truly number one is that, after this hundred-mirror battle has raged until now, no brand has ever been able to pull decisively ahead through an overwhelming product experience; the industry as a whole is stuck at the critical leap from early adoption to daily use.

In June this year, Rokid's "secret filming" incident brought this risk starkly before the public for the first time.

At the time, some netizens discovered in Rokid's official user community that a user wearing the brand's smart glasses secretly filmed flight attendants on a flight and publicly posted the video. In addition, the community contained a large amount of content filmed of passersby in public places such as subways, parks, and shopping malls, most of it without the consent of the people involved, with faces clearly recognizable.

The incident quickly escalated, not only plunging Rokid into a public opinion storm but also branding the entire AI glasses industry with the controversial label of "secret filming tool."

At a time when products are homogenized, perception capability is the core lever for brands to build differentiation. RayNeo Innovation's V4 upgraded the camera module and stabilization algorithm, strengthening first-person shooting and on-device visual recognition capabilities; Rokid focuses on POV shooting and object recognition; Xiaomi, meanwhile, integrated the camera with payments and smart home control, enabling the quick interaction of "look and scan to pay."

RayNeo Innovation V4, image source: RayNeo Innovation WeChat official account

The higher the camera resolution, the more sensitive the audio pickup, and the richer the AI-recognized scenarios, the more prominent the product's selling points. But it is precisely these core capabilities that constitute the source of privacy risk. Without a camera, real-time translation, image recognition, and scene perception would be impossible; with a camera, the hidden dangers of secret filming, indiscriminate filming, and data leaks are also hard to avoid.

Manufacturers have by no means not tried to strike a balance. After the controversy, Rokid quickly issued a statement, cleaned up violating content, banned the account involved, and emphasized that its products come standard with a hardware-level recording indicator light that is forced on during recording.

As Rokid founder Zhu Mingming recently mentioned in a media interview: "Privacy and protection in the AI glasses field are a contest of spear and shield; this is a proposition the industry must confront head-on. What companies can do is raise the cost of making mistakes to the maximum extent."

RayNeo Innovation is deeply engaged in diffractive optical waveguides + MicroLED self-developed optical engines, controlling costs through high yield rates and focusing on cost-effectiveness and mass production capability for the consumer market; XREAL takes viewing glasses based on the BirdBath + Micro-OLED solution as its base, rapidly scaling up with a mature solution while gradually upgrading toward the optical waveguide route; Rokid adopts a dual-line layout across all price segments, using BirdBath for entry-level models to lower the threshold for purchase and equipping flagship models with diffractive optical waveguides to push into the high end; INMO adheres to an integrated lightweight route, continuously reducing volume on the basis of monochrome diffractive optical waveguides and focusing on the form-factor advantage of daily wear.

But no single route can be perfect. Although diffractive optical waveguides are thin, light, and easy to mass-produce, their light efficiency is generally only 10%–20%, resulting in poor visibility under strong outdoor light; the BirdBath solution offers stable image quality and controllable cost, but is rather bulky and suffers from serious front light leakage, making its appearance hard to integrate into daily life; high-end geometric optical waveguides offer balanced experience, yet due to complex processes and high cost, they remain difficult to bring into the mass consumer market.

The inherent limitations of the technical routes translate directly into users' daily pain points. Dizziness and visual fatigue are the most common feedback: because of the visual discrepancy between the displayed image and the real environment, plus insufficient interpupillary distance adaptation accuracy, many users experience dizziness and eye strain after wearing the glasses for half an hour.

Poor outdoor visibility, 3–5 hours of battery life, and a nose-pinching feeling caused by a forward center of gravity all make all-day wear a luxury. These fragmented yet high-frequency experience problems are precisely the key factors preventing AI glasses from breaking out of their niche.

Data from the "2026–2027 China AI Glasses Industry Trend White Paper" released by iiMedia Research shows that the main consumers of AI glasses are AI enthusiasts and tech early adopters, accounting for 32.5% and 32.2% respectively, together making up more than 60%.

On one side is the barrier of public trust erected by privacy controversies, which leaves AI glasses constantly facing acceptance disputes in public scenarios such as commuting, venues, and offices; on the other side is the experience compromise under the roadmap battle, which prevents products from supporting all-day, multi-scenario daily wear. With internal and external dilemmas intertwined, AI glasses still find it difficult to truly enter ordinary households.

3. Hard to commercialize, low repurchase rates—AI glasses have yet to see their "iPhone" moment

New product iterations are accelerating, financing continues to rise, and major companies are entering across sectors; the excitement on the supply side is almost at its peak. But peeling back the shell of hype and looking at the real market size, profit model, and user retention, the slowness and weakness of the commercialization process are equally clear.

Online monitoring by RUNTO Technology shows that in July 2026, sales of smart glasses on the three major platforms of JD.com, Tmall, and Douyin totaled 74,000 units, down 15.3% year on year and down 20.9% month on month.

The global market is the same. According to RUNTO Technology's estimates, Meta's total smart glasses shipments for the full year are expected to be about 9 million units, firmly ranking first globally, but its main volume products remain the audio and camera models without displays.

Even more alarming is the slowdown in growth momentum. IDC data shows that in the second quarter of 2026, domestic smart glasses shipments were about 611,000 units, down 8.0% year on year, showing negative growth for the first time.

According to estimates in an AI glasses industry chain report released by Guotai Haitong, the gross margin on complete hardware for domestic consumer-brand companies is about 15%–20%; after deducting channel commissions, marketing spending, R&D, and after-sales expenses, most leading brands are near the break-even line.

As the industry leader, XREAL is even more illustrative of the problem. According to its Hong Kong stock prospectus, the company's revenue in 2025 was 516 million yuan, up 30.8% year on year, and its gross margin had risen to 35.2%, significantly higher than the industry average, but its adjusted net loss for the full year was 250 million yuan, with the core of the loss stemming from R&D investment and global channel-building expenses.

Image source: XREAL WeChat official account

Leading companies, relying on product mix optimization, can only narrow their losses; the profit pressure on mid-tier brands under the price war can be imagined.

The industrial and cultural tourism projects deployed by Rokid, INMO and others are mostly customized development, with fragmented demand and long delivery cycles; they are more of a cash flow supplement and cannot support the commercial imagination of the consumer segment. Even a global giant like Meta has not yet found a path to monetize value-added services beyond hardware, and the entire industry has still not found a clear second profit curve.

A survey in the iiMedia Research "2026–2027 China AI Glasses Industry Trend White Paper" shows that domestic AI glasses consumers' usage frequency is concentrated in the medium-frequency range of 1–2 times per week, accounting for 37.8%; next is 3–5 times per week, accounting for 28.4%; the share of users who wear them at high frequency every day is extremely low.

Many investors compare AI glasses to the iPhone and look forward to the industry ushering in an "iPhone moment," but the business logic of the two is fundamentally different.

When the iPhone was born, mobile phones were already a rigid-demand product that everyone had to own; what it completed was an experience-upgrade substitution. AI glasses, by contrast, must create from zero a new demand for daily wear. Creation is far more difficult than substitution, and this is also the core reason why commercialization has been slow to break through.

One obvious change this year is that, in the face of the commercialization dilemma, players have begun to break out of the old path of piling on features and explore new directions for breakthroughs, with the most central trend being extension toward personal AI Agent scenarios.

Rokid positions its glasses as an entry point for a portable AI assistant and has launched native Agent functions, supporting first-person-perspective schedule management, information reminders, and scenario-based proactive services, attempting to turn the glasses from a passive query tool into a portable Agent that provides proactive services;

RayNeo focuses on office scenarios and has built conference Agent capabilities, enabling real-time multilingual transcription, automatic outline generation, and to-do synchronization, strengthening its irreplaceability in business scenarios;

INMO, meanwhile, has partnered with large-model vendors to launch an AR-native Agent, supporting integrated triggering of real-scene navigation, object recognition, and real-time translation, exploring a new form of AI services under the first-person perspective.

Image source: INMO Tech WeChat official account

In addition, deepening ecosystem linkage is also a common choice. Huawei and Xiaomi rely on their phone and smart home ecosystems to connect interaction between glasses and all-scenario devices; automakers such as Li Auto and NIO use glasses as an extension of the in-car ecosystem to create an immersive in-car experience.

However, these attempts are still in an early exploratory stage, have not yet formed a phenomenal rigid-demand scenario, and have not fundamentally driven wearing time or user retention. To break through in the commercialization of AI glasses, the industry ultimately still has to answer the simplest question: "Why would users wear them every day?"

The hundred-glasses battle will eventually come to an end, and what remains will not be the model with the most features, but the one users are truly willing to wear every day. From competing on specs to returning to experience, from fighting over shipments to increasing wearing time, from telling stories to making the business work, players need more patience.

This article is from the WeChat official account "Lianxian Insight" (ID: lxinsight), author: Wang Huiying, published by 36Kr with authorization.