Original · Unique Research · 2026-09-01
Editor's note: The first-person report and its judgments belong to the original Chinese author. This
English rendition retains the opening essay, four insight sections, speaker list, and the full roundtable transcript. The title's framing about "users forgetting it exists" is the original author's thesis attributed to the panel discussion. Product, user, revenue, and technical figures are speaker self-reports attributed to the named individuals, not independently audited findings.
AI Industry Observation
"AI hardware's inflection point isn't when everyone's talking about it; it's when no one is."
The person who said this is Li Feng (李峰), founder of Langhai Digital (澜海数码). His company makes AI companion robots for the elderly — the kind of device that sits on a nightstand, asks if you took your medicine, and calls your daughter if you don't respond.
At first glance, this doesn't sound like the future of AI hardware. There's no neural engine, no spatial computing, no $2,000 headset. But Li Feng's point is: the best AI product isn't the one that makes you go "wow"; it's the one that becomes invisible — like a toaster, like a light switch, like oxygen. You don't celebrate it every morning; you just use it.
This was the premise of a roundtable at the Unique Grand Awards titled "The AI Hardware Inflection Point." The five panelists came from wildly different corners: companion robots, smart sensors, elderly care devices, AI wearables, and interactive toys. The host was Unique Research's Wang Yan (王艳).
Over two hours, the conversation zigzagged through form factors, battery life, data privacy, eldercare, and why most AI gadgets on Kickstarter will be dead in 18 months. But the throughline was consistent: AI hardware stops being a gimmick when it stops demanding your attention and starts serving it.
Below is a lightly edited transcript of the full discussion.
The Moment You Notice the AI Gadget, It's Already Failed
Li Feng started with a provocative framing: "I'll tell you when AI hardware has truly arrived. It's the day a grandmother who has never used a smartphone can operate it without reading a manual, without asking her grandchild, and without throwing it in a drawer after three days. When that happens, she's not using 'AI hardware' anymore — she's just using her alarm clock, her medication reminder, her companion. The AI disappears into the background. That's the inflection point."
Zhang Ning (张宁), founder of Zhixiang Technology (智象科技), makes smart sensor devices for industrial and home use. He agreed but from a different angle: "In industrial settings, we've seen this. The best sensor is one that a factory worker doesn't think about. If every shift starts with the worker checking the sensor battery, calibrating it, and troubleshooting it, it's a failure. The best sensor just measures and transmits; the worker forgets it's there. Same principle in consumer AI hardware: if the device requires daily attention — charging, updating, troubleshooting — it's a toy, not a tool."
Zhao Chunhua (赵春华), co-founder and CEO of Shidong Technology (视洞科技), works on AI-powered elderly care devices. She told a story: "We tested a device with a family. The grandmother took to it immediately because it had a big button and a loud voice. After a week, her daughter told us: 'Mom doesn't talk about it anymore.' At first I was worried — did she stop using it? Then the daughter said: 'She uses it every morning and every evening. She just doesn't mention it. It's become part of her routine, like brushing her teeth.' That's when I knew we had something."
The "forgetting test" became the group's unofficial criterion. If users stop commenting on how "cool" or "smart" the device is, and start using it without thinking, it's crossed from gadget to utility.
Most AI Gadgets Are Solving Problems Nobody Has
Wang Yan asked the obvious question: if the standard is invisibility, why are so many AI hardware products solving problems that don't exist?
Liu Wei (刘伟), founder of Hanyun Intelligent (涵云智能), makes AI wearables — earbuds with real-time translation, a clip-on device that transcribes meetings. His answer was blunt: "I'll be honest. Half the AI gadgets on the market are solving a problem the founder has, not the user. The founder thinks 'wouldn't it be cool if...' and builds it. The user's reaction is 'that's cute' and then never opens the app again."
He gave a self-critical example: "We built a clip-on recording device. The pitch was 'record your meetings, transcribe them, summarize them.' On paper, everyone wants this. But in practice, lawyers don't use it because they can't record privileged conversations; salespeople don't use it because their clients complain; students don't use it because it's another thing to carry. We pivoted to a specific use case: medical transcription for small clinics. Doctors see 30 patients a day, spend 20 minutes typing notes after each one. Our device listens to the consultation and auto-generates the note. That's a real problem with a real budget. The clip-on meeting recorder? Nobody cares."
Wang Zong (王总), CEO of Shengshi Interactive (盛世互动), makes AI interactive toys for children. She said: "In the toy space, this is even more extreme. Every quarter, a new AI toy launches with a big marketing push — 'chat with your stuffed animal!' Parents buy it because it's educational. The child plays with it for three days. Then it's in the toy box. Why? Because the toy doesn't understand the child. It's a generic large model wearing a teddy bear costume. The child wants the toy to remember their name, their favorite color, the game they played yesterday. If the toy forgets, the child loses interest. We learned this the hard way: memory, not intelligence, is the killer feature for kids."
Zhao Chunhua added: "For elderly users, it's the same logic. An elderly person doesn't care if the device runs a 70-billion-parameter model. They care if it remembers that their son lives in another city, that they have diabetes, that they take pills after breakfast. The AI is invisible; what's visible is continuity. When the device remembers who you are and what you need, it earns trust. When it forgets, it's discarded."
The Cost of Forgetting: Battery, Privacy, and the Physical World
The discussion moved to the practical constraints that prevent AI hardware from becoming truly invisible.
Zhang Ning: "Battery life is the silent killer. In our industrial sensors, we need a device that runs on a coin cell for five years. If it dies every two weeks, the maintenance cost exceeds the device cost. In consumer AI hardware, the same principle applies but no one talks about it. An AI companion that needs daily charging is a burden, not a companion. We're seeing a split: devices that always need to be plugged in (smart displays, home robots) vs. devices that run for months on a charge (sensors, wearables). The plug-in ones will live on a desk or a nightstand; the battery ones will move with the user. Both have their place, but the design constraints are completely different."
Liu Wei: "Privacy is the other silent killer. An AI device that's always listening, always watching — you can't ask users to trust it without a visible mechanism. We found that a physical switch that disconnects the microphone isn't just a feature; it's a trust signal. Users need to see, touch, and confirm that the device is off. Software switches in an app don't work. The physical toggle is worth its weight in gold."
Li Feng: "In elderly care, privacy is even more sensitive. If the device records the grandmother's daily conversations, the family needs to know what's stored, where it goes, and who can access it. We built a local storage option: everything stays on the device unless the user explicitly chooses to share with family. The cloud version is opt-in, not default. This slowed down development, but it's non-negotiable for this demographic."
Wang Zong: "For kids, privacy has another dimension: parents want control. They want to see what the toy knows, reset it, delete it. If the toy is an opaque black box connected to some server, parents won't buy it again. We designed a parent dashboard that shows exactly what the toy remembers, what it says, and what it sends. Transparency is a selling point."
What's the Business Model When the Device Is Invisible?
The hardest question came last: if the best AI device is invisible, how do you build a business around it?
Li Feng: "In elderly care, the business model isn't the hardware; it's the service. The device itself is a loss leader — we sell it at near cost. The recurring revenue comes from the monitoring service: if the grandmother doesn't respond to the morning check-in, the device alerts the family and the monitoring center. That service costs 30 RMB a month. Multiply by thousands of subscribers, and that's the business. The hardware is the pipe; the service is the water."
Zhang Ning: "In industrial sensing, it's the opposite. The hardware is the product; the service is thin. A factory buys 5,000 sensors, installs them, and they run for five years. Our revenue is in the device sale and occasional maintenance. The data platform is included but not separately priced. The model is B2B and transactional, not subscription."
Liu Wei: "For our medical transcription device, it's a mix. The hardware is sold to clinics at a reasonable price. The AI transcription service is subscription: 200 RMB per clinician per month. The clinic pays because the device saves 15 hours of documentation time per week. That's a clear ROI. We don't need the device to be sexy; we need it to pay for itself within two months."
Wang Zong: "For children's toys, the business model is still maturing. We sell the toy upfront, then offer a content subscription — new stories, new games, new characters. The key insight is that the toy must keep growing with the child. A toy that stays the same from age 4 to age 7 gets boring. The subscription adds new content and new conversations. But the price point has to be low — parents won't pay $30 a month for a toy. We found $5–$8 a month is the sweet spot."
Zhao Chunhua: "In elder care, we're also exploring insurance partnerships. Some Chinese health insurance policies cover monitoring devices as part of senior care benefits. If insurance covers the device and the service, the barrier to adoption drops dramatically. That's a channel we're investing in. It's slower than direct-to-consumer but more durable."
Wang Yan closed: "The common thread across all these models is that the device itself is rarely the whole business. It's either a pipeline for a service, a tool that pays for itself, or a platform for content. The companies that win aren't the ones with the flashiest hardware; they're the ones that figured out the recurring revenue model before they designed the first prototype."
Written at the End
I left this conversation thinking about the toaster.
Nobody talks about their toaster. It's on the counter, it makes toast, it's always there. Nobody posts about it on social media, nobody reviews it as a "revolutionary kitchen device." It's just... there.
That's what AI hardware needs to become. Not a conversation piece. Not a Kickstarter sensation. Not a company that raises $50M and disappears in 18 months. A device that people forget to notice because it's doing its job quietly.
The panelists weren't dismissing flashy AI gadgets. They were saying: the inflection point isn't measured by how many people are talking about your product. It's measured by how many people stop talking about it — because it's just part of their lives.
"When users forget your device exists, you haven't failed. You've succeeded."
More Conversation Details
Speakers
Li Feng, Founder, Langhai Digital (澜海数码)
Zhang Ning, Founder & CEO, Zhixiang Technology (智象科技)
Zhao Chunhua, Co-founder & CEO, Shidong Technology (视洞科技)
Liu Wei, Founder, Hanyun Intelligent (涵云智能)
Wang Zong, CEO, Shengshi Interactive (盛世互动)
Host
Wang Yan, Unique Research
Wang Yan: Welcome everyone. Today's topic is the AI hardware inflection point. We have five founders from very different corners of the hardware space. Let's start with quick introductions and what you're building.
Li Feng: I'm Li Feng, founder of Langhai Digital. We make AI companion robots for elderly people in China. Our device sits on a nightstand, reminds people to take medication, checks in on them regularly, and alerts family if there's no response. It's voice-first, designed for users who may never have used a smartphone.
Zhang Ning: I'm Zhang Ning, founder and CEO of Zhixiang Technology. We make smart sensor devices for industrial and home applications. Our sensors monitor temperature, humidity, vibration, and occupancy — things that used to require manual checks. They run on batteries for years and transmit data wirelessly.
Zhao Chunhua: I'm Zhao Chunhua, co-founder and CEO of Shidong Technology. We focus on AI-powered elderly care devices. Beyond companion robots, we do fall detection, remote health monitoring, and family communication tools designed specifically for senior users and their adult children.
Liu Wei: I'm Liu Wei, founder of Hanyun Intelligent. We make AI wearables — originally we tried meeting transcription earbuds, but we pivoted to medical transcription for small clinics. Our device listens to doctor-patient consultations and auto-generates clinical notes.
Wang Zong: I'm Wang Zong, CEO of Shengshi Interactive. We make AI interactive toys for children ages 4 to 8. Our toys combine physical stuffed animals with AI conversation, storytelling, and educational games that grow with the child.
Wang Yan: Great. Let's dive in. The framing for today is "the inflection point begins when users forget it exists." What does that mean to each of you, in your own product context?
Li Feng: For me, it's simple. The day a grandmother uses our device without asking anyone how to use it, without reading a manual, and without throwing it in a drawer — that's the inflection point. When she stops saying "this AI thing" and starts saying "my clock" or "my reminder," the product has succeeded. The AI should be invisible.
Zhang Ning: In industrial sensing, this is already the standard. The best sensor is one that a factory worker doesn't think about. If every shift starts with checking the sensor, troubleshooting it, or replacing batteries, it's a failure. The best sensor just measures and transmits; the worker forgets it's there. Same principle in consumer AI: if the device demands daily attention — charging, updating, troubleshooting — it's a toy, not a tool.
Zhao Chunhua: We tested with a family. The grandmother took to the device immediately — big button, loud voice. After a week, her daughter said "Mom doesn't talk about it anymore." I was worried, then the daughter said: "She uses it every morning and evening. She just doesn't mention it. It's become part of her routine, like brushing her teeth." That's when I knew.
Liu Wei: I agree. The best wearable is one you forget you're wearing. If users are constantly adjusting it, charging it, or taking it off, they won't wear it long enough for it to be useful. Our medical transcription clip-on has to be light enough that the doctor forgets it's there during a 20-minute consultation. If it's distracting, the doctor stops using it.
Wang Zong: For kids, it's different but similar. The best AI toy isn't the one that impresses the parent on day one. It's the one the child keeps coming back to. When the child stops saying "look what my toy can do!" and starts saying "my toy remembers that I like dinosaurs," it's working. The AI becomes a friend, not a gadget.
Wang Yan: If invisibility is the goal, why do so many AI hardware products solve problems nobody has?
Liu Wei: I'll be self-critical. We built a clip-on meeting recorder. The pitch was "record, transcribe, summarize." On paper everyone wants it. But lawyers can't record privileged conversations; salespeople's clients complain; students don't want to carry another device. We pivoted to medical transcription for small clinics. Doctors see 30 patients a day, spend 20 minutes typing notes after each. Our device listens and generates the note. That's a real problem with a real budget. The meeting recorder? Nobody cared.
Wang Zong: In toys, every quarter a new AI toy launches — "chat with your stuffed animal!" Parents buy it because it's educational. The child plays for three days. Then it's in the toy box. Why? Because the toy doesn't understand the child. It's a generic large model wearing a teddy bear costume. The child wants the toy to remember their name, favorite color, the game they played yesterday. When it forgets, interest dies. Memory, not intelligence, is the killer feature for kids.
Zhao Chunhua: For elderly users, same logic. They don't care if it runs a 70-billion-parameter model. They care if it remembers their son lives in another city, that they have diabetes, that they take pills after breakfast. The AI is invisible; what's visible is continuity. When the device remembers who you are, it earns trust. When it forgets, it's discarded.
Li Feng: I'd add: many AI gadgets are designed by young engineers for other young engineers. The use case is "I want to show off to my friends that I have an AI thing." That's a vanity purchase, not a need. The products that stick solve a boring, repetitive, stressful problem that someone already has. If your product doesn't replace a chore, it's a decoration.
Wang Yan: What are the practical constraints — battery, connectivity, privacy — that prevent these devices from becoming truly invisible?
Zhang Ning: Battery life is the silent killer. In industrial sensors, we need a coin cell lasting five years. If it dies every two weeks, maintenance cost exceeds device cost. In consumer AI hardware, same principle but nobody talks about it. An AI companion that needs daily charging is a burden, not a companion. We're seeing a split: always-plugged-in devices (displays, home robots) vs. long-battery devices (sensors, wearables). The plug-in ones live on a desk; the battery ones move with the user.
Liu Wei: Privacy is the other silent killer. An always-listening AI device — you can't ask users to trust it without a visible mechanism. We found a physical switch that disconnects the microphone isn't just a feature; it's a trust signal. Users need to see, touch, confirm it's off. Software switches in an app don't work. The physical toggle is worth its weight in gold.
Li Feng: In elderly care, privacy is even more sensitive. If the device records daily conversations, the family needs to know what's stored, where it goes, who can access it. We built local storage as default; cloud sharing is opt-in, not default. This slowed development but is non-negotiable.
Wang Zong: For kids, privacy has another dimension: parents want control. They want to see what the toy knows, reset it, delete it. An opaque black box connected to a server won't be bought again. We built a parent dashboard showing exactly what the toy remembers, says, and sends. Transparency is a selling point.
Zhao Chunhua: We also face connectivity issues. Many elderly users live in homes with poor Wi-Fi. If the device needs a cloud connection to function, and the Wi-Fi drops, the device becomes useless. We designed offline mode: basic functions — medication reminders, emergency calls — work locally without cloud. The cloud only adds value (family alerts, health reports) when connected. Graceful degradation is essential.
Wang Yan: Business models — how do you make money when the device is supposed to be invisible?
Li Feng: In elderly care, the hardware is a loss leader. We sell at near cost. Recurring revenue is the monitoring service: if Grandma doesn't respond to morning check-in, the device alerts family and the monitoring center. That's 30 RMB/month. Multiply by thousands of subscribers; that's the business. The hardware is the pipe; the service is the water.
Zhang Ning: Industrial sensing is the opposite. Hardware is the product; service is thin. A factory buys 5,000 sensors, installs them, runs for five years. Revenue is in device sale and occasional maintenance. Data platform is included but not separately priced. B2B transactional, not subscription.
Liu Wei: Medical transcription is a mix. Hardware sold to clinics at a reasonable price. AI transcription service is subscription: 200 RMB per clinician per month. The clinic pays because it saves 15 hours of documentation time weekly. Clear ROI. We don't need the device to be sexy; we need it to pay for itself within two months.
Wang Zong: Children's toys are still maturing. Sell the toy upfront, then content subscription: new stories, games, characters. The toy must keep growing with the child. A toy that stays the same from age 4 to 7 gets boring. Subscription adds new content and conversations. Price point must be low — parents won't pay $30/month for a toy. We found $5–$8/month is the sweet spot.
Zhao Chunhua: In elder care, we're exploring insurance partnerships. Some Chinese health insurance covers monitoring devices as senior care benefits. If insurance covers device and service, adoption barrier drops dramatically. Slower than D2C but more durable.
Wang Yan: Let's go around the table — one piece of advice for a founder building AI hardware today.
Li Feng: Don't build for yourself. Build for the person who will actually use it. If your target user is a 75-year-old grandmother, you don't get to design the interface the way you want. She doesn't care about your architecture; she cares if it works when she needs it.
Zhang Ning: Obsess over reliability. The best AI hardware fails silently — meaning it doesn't fail. If your device needs daily troubleshooting, you're in the software business, not the hardware business. Hardware is about mean time between failures, not feature count.
Zhao Chunhua: Design for the family, not just the user. In elder care, the elderly person uses the device, but the adult child pays and worries. You need to serve both: simple enough for Grandma, transparent enough for the daughter.
Liu Wei: Find the ROI. Before you build anything, ask: who pays, how much, and why will they renew? If you can't answer that in one sentence, the hardware is a hobby, not a business.
Wang Zong: Build for the child's growth, not the parent's day one. Parents buy on day one based on marketing; children stay on month six based on whether the toy grows with them. Design for month six.
Wang Yan: Thank you all. Final thought: the best AI product is the one people stop talking about because it just works. That's the inflection point.