After AI hardware left mobile phones, it first encountered human limitations.
Qian Weijing
2026.11.21 12:14
· From Guangdong
Full text 6242 words
In 2026, there will be a clear contrast in the AI hardware industry. Model capabilities are rapidly popularizing, new products are being launched intensively, and the position of mobile phones as personal computing centers has not been significantly shaken.
Manufacturers are still looking for opportunities beyond mobile phones. The ring first enters the body data, and the earphones and recording card save the voice content. Glasses manufacturers are trying to obtain visual information, and brain-computer devices are still verifying action intention recognition. These explorations jointly point to a judgment that future interactive portals will be scattered around the human body.
This judgment has a basis in reality. Mobile phones require users to actively turn on the screen, but a lot of body data and environmental information are continuously generated. If the device can collect this information in a low-disruption manner, AI will have the opportunity to help the user before issuing explicit instructions.
The difficulty also starts here. Once the interactive portal is close to the body, the product evaluation criteria will immediately change. Wearing comfort, data credibility, privacy boundaries and long-term maintenance costs will all affect user choice earlier than model demonstration.
In the past, consumers were willing to pay for a bigger screen, a stronger chip or a novel feature. After AI hardware enters the market, parameters are still important, but a single demonstration can no longer explain the value of the product.
After a device enters daily life, it has to continue to face charging, pairing, authorization, false alarms and maintenance. The operations saved by AI are likely to be consumed by these new links.
Therefore, the inventory of smart hardware in 2026 cannot only look at the number of new products and technical concepts. More reliable clues come from stable use, continuous service and multi-device collaboration.
Public materials and channel information can illustrate part of the facts, but they cannot replace industry judgment. The sales volume of a single company, the judgment of the founder, and the data of stays in the store are just different aspects of observing this change.
Looking back at the end of the year, AI hardware has left the mobile phone screen, but it has not yet completed the leap from product innovation to lifestyle habits.
This wave of change is different from past wearable device crazes. The previous generation of products was more responsible for notifications, step counting and health records, while the AI hardware tried to understand the environment and proactively provide results.
Changes in capability boundaries have led hardware companies to re-evaluate the value of human time. In the past, products competed for user retention, but new devices need to prove that they reduce retention.
This also changes the industry evaluation standard from the quantity of functions to the quality of task completion. Providing reliable results on key tasks supports retention better than stacking dozens of low-frequency functions.
From this perspective, the competitive core of AI hardware has gone beyond model access. Whether equipment, data and services can form a stable closed loop will begin to determine how far the product can go.
The mobile phone was not lost, but the entrances began to disperse.
Mobile phones still have basic capabilities that are difficult to replace with other devices. It masters communication, payment, application distribution, identity authentication and content consumption, and also has a mature large screen and input system.
If AI hardware wants to change this landscape, it cannot just bring a few more convenient functions. It must be something users want to carry around with them and reduce steps in enough daily tasks.
Most of the new products that appeared this year actually took a different path. They avoid the core capabilities of mobile phones and start from the links where processing by mobile phones is not natural enough.
The advantage of the ring is continuous wear, and the advantage of the recording card is the startup speed. While glasses attempt to obtain a first-person perspective, the task of a desktop 3D printer is more direct, turning digital content into physical objects.
This division of labor shows that it is difficult for a new all-round terminal to appear in the industry in the short term. What is more likely to happen is that different devices each take on a small piece of the task, and some system connects them.
The problem is that after the tasks are split, the number of objects that users need to manage increases. Each device has its own power, network, app and data permissions. If collaboration is not done well, the efficiency brought by AI hardware will be offset by the new maintenance burden.
This is why the lead in a single function is not enough to determine the outcome. For a device to enter life, it must pass the hurdle of continuous use.
The user evaluation method is very simple. The equipment should save time, require less operations, and work stably. But the problems on the industrial side are much more complicated. The more data collected by equipment, the higher the security, algorithm and compliance costs borne by manufacturers.
Competition in smart hardware thus extends from single products to system design. The hardware is just the entrance. Data flow, service response and permission management determine whether the entrance can be used for a long time.
In this sense, mobile phones do not lose value because of the emergence of new devices. On the contrary, many AI hardware still require mobile phones to complete setup, view results and handle complex tasks. In the short term, mobile phones are more like dispatch centers. The new device is more responsible for perception and light interaction, and confirmation, payment, editing and sharing are still completed by the mobile phone. But how long this relationship will last depends not only on how many cross-application permissions the AI system can obtain, but also on the user's willingness to authorize personal information.
The fragmentation trend the industry has seen this year does not yet indicate that mobile phones are declining. To be more precise, this is a rearrangement of the extension of mobile phone capabilities. Only when new devices can handle more tasks independently will the position of the mobile phone become significantly looser.
For startups, this rearrangement brings opportunities as well as misjudgments. Opportunities come from small needs that mobile phone manufacturers have ignored in the past. Special equipment can make these experiences better.
Misjudgment occurs here as well. After small demands are turned into products, they may not be able to support a company's long-term growth.
The health reminder function can be placed in the ring or become part of the mobile phone application. The same is true for recording summaries, which can be done with cards, office software and mobile phones. Hardware companies must prove that the convenience of standalone devices outweighs the cost of additional hardware.
This cost is not just the selling price. Users also have to allocate attention to the device, deal with charging and malfunctions, and understand new data permissions. Many products emphasize AI capabilities when they are released, but fail to answer a more basic question: why do users not use mobile phones to complete the same thing?
If the answer is just a few less screen taps, the market will quickly hit a ceiling.
The more convincing reason for independent hardware is that it can continuously collect information that is difficult to obtain with mobile phones, or complete tasks when hands are inconvenient, vision is limited, and time is tight.
This explains the focus on rings, earphones and glasses. They occupy different positions on the body and can cover situations where the phone is inconvenient to handle.
But location advantages don't automatically translate into business value. The frequency of collection, data accuracy and quality of feedback determine whether this information can be used in the long term. The industry used to measure hardware success by shipments. AI hardware also needs to add a set of indicators, such as the proportion of active devices, data continuity and service utilization.
After a batch of equipment is sold, if it remains idle for a long time, no matter how high the shipment volume is, it will not be able to form a stable entrance. Judging from this standard, many companies are still in the verification stage this year. They prove that the product can be built, but they don’t fully prove that users will stick around.
This is also the background for capital markets to become cautious. The model concept can no longer support valuation alone, and hardware sales cannot cover up insufficient retention. When investors start asking for usage data, industry competition will shift from the release rhythm back to the product itself.
Mobile phones continue to occupy the center, which just shows that the new entrance has not yet formed a strong enough substitute capability. A more reasonable judgment at this stage is that AI hardware is supplementing mobile phones, and there is still a long way to go before rewriting mobile phones.
Complementary relationships may also generate business value, but they require manufacturers to reduce their reliance on alternative narratives. Only when a startup company first establishes stable use in tasks that mobile phones are not good at can they be able to talk about the next step of expansion.
Low-frequency rigid needs retain users better than novel features
To judge the value of AI hardware, you must first look at the frequency of use. No matter how advanced a feature is, if the user only thinks of it once every few weeks, it will be difficult for the device to support its independent existence.
The opportunity for smart rings comes from continuous wear. Only when sleep, heart rate and exercise data are accumulated for a certain period can valuable trend judgments be made.
RingConn entered this direction earlier. This year’s new products include vascular health trend reminders and vibration reminders. This product change fully exposed the industry’s difficulties.
Recording data is relatively easy, but interpreting data is more difficult. If health reminders are too frequent, users will gradually become numb. Prompts are insufficiently based and may also cause anxiety. Moreover, the ring category also faces medical boundaries. Consumer-grade equipment can help users observe trends, but it is difficult to directly undertake diagnostic functions. Once this line is crossed in publicity, subsequent trust is difficult to repair.
Recording hardware faces work efficiency issues. Users do not lack recording tools. What consumes time is the organization, search and distribution after recording.
Mobvoi expanded its TicNote product line this year. TicNote Pods and TicNote Watch respectively cover independent communication and wrist-based recording, and cooperate with enterprise WeChat, indicating that such devices are entering the enterprise collaboration process from personal tools.
After entering the enterprise, recording hardware must assume higher data responsibilities. The recording storage location, access rights, retention period and model training purposes all need to be written into the contract.
The consumer side first requires ease of operation. After entering the enterprise, risk controllability will rise to the same or even higher position. After the two sets of requirements are superimposed, the threshold for recording hardware will be significantly higher than that of ordinary electronic products.
3D printers have different paths. Tuozhu released its new flagship product H3C this year and has achieved cumulative sales of more than one million consumer-grade devices. Its user goal is very specific - to turn digital designs into physical objects.
Generative AI makes modeling easier, but the cost of manufacturing does not disappear. After the model is generated, structural strength, material selection, slicing parameters and failure rate must also be dealt with.
This type of product has a chance to stay because the results of use are easy to verify. The printing results can be seen immediately, and users can clearly judge what the device has done for them.
The situation faced by desktop companion robots is more complex. The need for emotion and companionship is real, but product value is hard to judge from a single demo. A few dozen seconds of interaction in a shopping mall cannot cover the noise, dialects, network fluctuations and context jumps in a home environment. Consumers will encounter these restrictions only after purchasing.
Companion products also need to answer price questions. A device costing several thousand dollars can only bring a short-term novelty, and users will soon recalculate its place in the home.
Many products this year are stuck at this step. The release stage can attract attention, but after delivery, it is difficult to enter the fixed use stage.
Offline channels can detect this gap earlier. INNO100 co-founder Zhang Wenhan has long been concerned about offline sales. Store data shows that there is often a clear gap between onlookers, stay, repeated trials and final payment.
INNO100 is jointly created by North American creative crowdfunding platform Kickstarter and Shenzhen overseas platform GoGlobal Global Service Center. Relevant data is affected by store location, customer flow and product selection, and cannot directly represent the entire market, but it can remind the industry that online popularity and real purchases are not always consistent.
From a retention perspective, low-frequency rigid demand is more reliable than high-frequency entertainment. Users may not use it every day, but as long as the device can solve problems when they arise, the product has a reason to continue to exist.
Low-frequency rigid demand will also affect product pricing. The device does not need to pursue daily activity, but it must provide sufficiently certain results at critical moments. Such requirements are more likely to support specialized equipment because users are willing to accept additional hardware for certain results.
On the contrary, high-frequency but low-value functions are easily absorbed by mobile applications. If product positioning cannot distinguish between these two types of needs, subsequent pricing and subscriptions will be biased.
Smart glasses and brain-computer interfaces push privacy to the forefront
Vision is regarded by many manufacturers as the entrance to the next generation, because sight can reflect attention, and the first perspective can also help AI understand the environment.
XREAL launched AURA with Google this year. The product is based on Android XR and integrates Gemini. XREAL founder Xu Chi has long been optimistic about glasses as the entrance to the next generation of personal computing. This direction has technical logic and a high practical threshold.
Glasses need to be worn on the face for a long time, and weight, heat, battery life and display effects will all affect wearing them. If any experience is not good enough, it will be difficult for users to stick with it.
A more sensitive issue is public borders. Taking pictures with a mobile phone requires actively raising the device, and the glasses may continue to record without others noticing.
Users' willingness to wear them only depends on personal choice - which will affect the popularity of smart glasses. Display technology and model capabilities can continue to iterate, and social acceptance takes longer to build.
Brain-computer interfaces face stricter constraints. BrainCo Technology this year demonstrated a brain-controlled robot training platform and an intelligent bionic hand, and also disclosed a device under development that assists appetite suppression through brain electrical stimulation.
These directions have clear technical value, especially in the field of medical rehabilitation. However, there is still a gap between the booth recognition effect and daily stable use.
Products involving health intervention must also undergo clinical verification, safety assessment and regulatory approval. The promotion rhythm of consumer wearable devices cannot be directly applied to brain-computer devices. Data ownership, purpose of use, retention period, third-party access and exit mechanisms will all affect user trust.
In the past, hardware companies could rely on parameters and appearance to establish differentiation. After entering the field of body data, trust itself has become part of product capabilities. Once there is a problem with privacy processing, it will not only affect a certain product, but may also drag down the entire category.
In addition, the industry's discussion on privacy has increased significantly this year, indicating that manufacturers have realized that after being close to the body, data collection can no longer continue to use the extensive method of the mobile Internet era. This will raise the threshold for starting a business and force companies to put compliance and security first.
Privacy concerns can also change product pricing. Compliance, security audits, and data storage all require ongoing investment, and low-priced hardware cannot cover these costs in the long term.
From a business perspective, equipment for enterprise customers is often easier to establish a charging model than for consumer-grade products. Enterprises are willing to pay for controllable processes and are more aware of the cost of data breaches.
The difficulty with consumer products is that it is difficult for users to judge the cost of privacy in advance. Many people only re-evaluate their devices after a data breach or permissions dispute arises. Vendors therefore cannot put privacy statements in lengthy agreements. The data scope, purpose of collection, retention period and deletion method need to be presented in a way that ordinary users can understand.
For glasses and brain-computer devices, this transparency has a direct impact on social acceptance.
Business model and synergy capabilities determine the next stage
The business model of AI hardware has not yet been completely closed. One-time hardware sales can hardly cover ongoing model calls, cloud storage, and algorithm maintenance costs. Many companies are turning to subscriptions as a result. Recording transcription, health analysis, cloud storage and intelligent body services are all packaged as ongoing charging items.
Users are willing to pay for subscriptions, which is the premise of the business model. Services must consistently save time, reduce errors, or deliver results that are difficult to obtain through other channels. If a subscription only charges for basic functions, users will be lost quickly. Hardware companies will also suffer from thinning equipment profits and insufficient service revenue.
The capital market has begun to adjust its judgment this year. Financing and initial sales are still important, while repurchase, retention, return rate and renewal rate are placed higher.
Zhang Jun, vice president of investment at Zhixing Institute, continues to pay attention to smart hardware projects this year, and the investment pace remains cautious. His concerns are very representative. After the category is verified, the supply chain will quickly flatten the appearance and function, and the early advantages can easily be diluted.
The ultimate test is the engineering efficiency of the Chinese team - after new overseas categories receive market feedback, domestic manufacturers often quickly launch products with similar functions and lower prices. However, this ability shortens the innovation implementation cycle and compresses the window period for startups. The first movers have not yet formed a stable user base, and the followers have begun to lower prices.
A feature race ensues. The number of sensors and model capabilities continues to increase, and the core purpose of some products becomes blurred. The more complex the equipment, the higher the learning and maintenance costs for users.
A mature supply chain can solve manufacturing problems, but demand research must still be completed separately. Target users, usage frequency and retention after three months cannot be solved by accelerating proofing.
Some overseas teams are more accustomed to starting from a narrow group of people, verifying specific needs first, and then expanding functions. The speed is slower and product boundaries are relatively clear.
This difference is not absolute, but reflects the habits of the two products. Chinese companies are good at turning ideas into products quickly. What needs to be completed next is demand definition, data governance, after-sales service and content ecology.
As the number of devices continues to increase, collaboration will become a new focus of competition. If rings, headphones, glasses and recording equipment are separated from each other, all users get is a more complex list of equipment.
Vendors are looking for a unified scheduling layer. The dispatch layer needs to process visual, sound, and body data simultaneously and return the results to the appropriate device. Mobile phones continue to undertake identity management and complex operations in this system.
Platform companies have natural advantages in this aspect. The account system, application ecology and cloud capabilities are all in their hands. If a startup only makes a single device, it can easily be integrated into the platform.
But there are still opportunities for startups. Vertical groups, professional scenes and higher data security standards may still form independent spaces. In addition, multi-device collaboration will also change the distribution of benefits. The scheduling layer controls the task entry, which also affects data attribution and service revenue.
Open standards can alleviate equipment fragmentation, but it is difficult to completely solve the problem. As data sensitivity increases, companies are more inclined to keep information in their own systems.
Privacy isn’t just about compliance terms, either. Recording equipment can affect other participants being recorded. Glasses and health devices will also touch the boundaries of public space and the preservation of body information respectively. User trust takes a long time to build, and it can disappear quickly because of an accident.
In 2026, the industry has proven that AI can leave the mobile phone screen and enter rings, headphones, glasses, printers and brain-computer devices. But it has yet to be proven that these devices can remain in people's lives for long periods of time.
The next stage of competition will no longer be determined by the number of new products. What's more critical is to reduce usage hassles, protect user data, and allow different devices to form effective collaboration.
After smart hardware is brought closer to the body, manufacturers gain new access and accept stricter constraints. Products that ultimately stick around need to solve real problems while reducing interruptions, and gain people's trust over the long term. (This article was first published on Titanium Media APP, written by Qian Fang)
AI is rewriting the industry, who will hand in the paper first? TMTpost Media's "The Great Migration of Intelligence" will start from ten questions and observe how AI has been implemented in various industries from the battle of models. Next, we will organize more topics and look at the intelligent migration of different industries through technological trends and industry observations in China and the United States.
AI hardware has left mobile phones, but the real test is to stay in people’s daily lives. What equipment can become a necessity? How to collaborate with multiple terminals?
From December 7th to 8th, come to the 2026 T-EDGE Innovation Conference "Smart Migration" at the Zhongguancun Exhibition Center in Beijing to find answers. Listen to front-line practitioners dismantle product opportunities at keynote speeches, touch hard-core new products with your own hands at the AI Demo Show, and see the true value of AI implementation in two days.
Click to purchase tickets to join the 2026 T-EDGE Innovation Conference Beijing.
Respect for the original work, if you are qualified, you should be appreciated.
According to the real-name requirement of the Cybersecurity Law, please bind your mobile phone number before posting a comment.
Express
12:15
Bitcoin fell below $56,000, down more than 6% on the day
12:05
The National Development and Reform Commission held a video conference to supervise the payment and use of government investment funds and project reserve work
11:44
Bitcoin falls below $57,000, hitting lowest level since September 2024
11:36
The Mongolian section of Hanoi has entered the cold flood season from 2026 to 2027
11:11
SoftBank’s Masayoshi Son discussed acquiring SK Hynix for US$1 trillion
11:05
U.S. cargo Dragon spacecraft breaks away from the space station and sets off for return to Earth
10:31
U.S. presidential envoy discusses U.S.-Iran mediation work with senior Qatari officials
10:30
Quantum computer "high-speed memory" is realized on a real machine for the first time in the world
10:27
机构:需求降温,11月电视、显示器及笔电面板价格或全面持平
10:25
郭明錤:谷歌预计将调整TPU 8t与8i系列2027年出货比重
10:22
阶段性调整未改底色,多家机构看好A股估值与韧性
10:21
网传零跑A10后备厢被几箱啤酒压穿,零跑回应
10:19
国际金价延续跌势
10:16
美联储官员保尔森:支持审慎加息,必要时可连续加息
10:15
若无法完成对软银的出售,ABB据悉考虑将机器人业务分拆上市
10:12
Anthropic IPO获散户投资者约1150亿至1200亿美元认购需求
10:10
金饰克价继续下跌,老凤祥单日跌34元
外卖公司Wonder正考虑出售Grubhub
10:09
Databricks据悉洽谈新一轮融资,估值有望达到2250亿美元
10:08
CrowdStrike据悉正洽谈收购网络安全公司Zscaler
扫描下载App

