---
title: "AI Hardware Going Global: The Survivors Are All Doing Subtraction"
author: "Unique Research"
sourcePublication: "Unique Research Substack"
originalPublishedAt: "2026-06-02T11:01:06+00:00"
canonical: "https://ffcap.cn/en/research/src-20260602-01html"
source: "https://uniqueresearch.substack.com/p/src-20260602-01html"
language: "en"
---

# AI Hardware Going Global: The Survivors Are All Doing Subtraction

_Original · Unique Research · 2026-06-02_

_Editor's note: This is a complete English rendition of the original Chinese article, including the editorial analysis and full roundtable transcript. All statements by named panelists are presented as their own views. Market share figures, revenue numbers and legal case details are source-reported claims. The author's analytical judgments are preserved as original-author commentary._

AI Industry Observation

AI Hardware Going Global: The Survivors Are All Doing Subtraction

Sometimes shackles force out better things than freedom does.

AI glasses cut the camera. AI rings cut health monitoring. An AI plush toy has a factory price of 5 yuan, and the founder wants to sell a 150-yuan AI module—the entire supply chain upstream and downstream can't understand what he's talking about.

These are four real slices from a roundtable at the 2026 Shenzhen Unique Awards. Four completely different forms of AI hardware companies—one making glasses, one making rings, one making desktop pets, one making plush toys—sat together to discuss product layout in overseas markets. After the discussion, you'll find that their ways of surviving are strikingly consistent: not doing addition in their products, but desperately doing subtraction.

Put simply, this is a small breathing space that startups have earned under the triple shackles of resources, technology, and market education.

But sometimes shackles force out better things than freedom does.

Collective Subtraction: When "Standard" Becomes "Burden"

Deng Xudong, co-founder of Gyges Labs (Vocci), took the stage wearing a pair of ordinary-looking glasses.

"You probably can't tell at all that these are smart glasses with a display," he said. These glasses carry what they claim is the world's smallest wafer display system, DigiWindow—but no camera.

In a track where almost all AI glasses are competing on cameras, this is a counterintuitive choice. Deng Xudong's explanation was blunt: "If you add a camera, it gets heavy. If it's heavy, you don't want to wear it. If you don't want to wear it, it gathers dust."

In one sentence, he laid out the industry's "impossible triangle."

He gave this strategy a name: "restrained innovation," saying he has "reverence for hardware." It sounds philosophical. But think about it—the Vision Pro is powerful and has a good experience, "the only problem is you can't take it out." Deng Xudong himself admits that the fundamental reason for cutting features is "to solve the problem of being able to take it out, you have to make some trade-offs."

In other words: it's not that they don't want to add features—it's that adding them means death.

The same logic appears in Tang Chang, founder of Nuoyi Innovation (诺一创新). He made an AI voice ring called Spark, selling on lightning-fast recording of inspirations and meeting content—but no health monitoring. In a health-ring track where Oura holds 74% market share and sold 5.5 million units in 2025, Tang Chang chose to completely detach.

"The people who buy our rings and the people who buy health rings are not the same demographic," he said directly. Health ring users wear them at night to monitor sleep and blood oxygen; his users wear them during the day to record meetings and capture inspirations. Scenarios don't overlap, users don't overlap.

But this isn't entirely a story of "I discovered a blue ocean." Tang Chang's entrepreneurial experience was quite brutal: after leaving ByteDance, his team shrank to just himself, sustained by startup subsidies. Early prototypes even had a version with a camera, which was later cut—this was called "strategic compromise" by outsiders, but essentially, what qualifications does a one-person startup team have to go head-to-head with Oura?

I think we should be honest: the so-called "restrained innovation" is half forced, and half survival wisdom groped out of being forced. In an early market where even Meta and Apple are still trial-and-erroring, if a startup doesn't do subtraction and goes into the red ocean to compete on extreme operations and low cost, the outcome will most likely be being strangled. The predicament of "can't do addition" has instead forced them to find differentiated cracks.

Deng Xudong said something worth pondering: "We want to be leaders, not followers."

That sounds inspiring when said. But a more accurate description might be: beside the track crowded with followers, they've stepped out a small path no one has walked yet. Whether it leads anywhere is another matter.

Who's Paying: 5-Yuan Toys, $3.3 Million Crowdfunding, and Men Who "Don't Know They're Lonely"

When it comes to finding the first batch of paying users, the four companies' paths differ so much they seem like four different industries.

Zhou Yixu, founder of Saibo Chuangli (赛博创力), told an experience that left a deep impression on me. In March 2024, he took AI modules to the Shenzhen Toy Fair, with a simple idea: "Just spend 50 yuan buying my module, and your toy becomes an AI toy."

The result was two body blows.

First, "When I talked to everyone in the supply chain upstream and downstream, I had to start by explaining 'what is open-ended dialogue' and 'what is a large model.'"—in 2024, the Shenzhen toy industry still needed science popularization on large models.

Second, "A lot of toys have factory prices of 5, 8, 15 yuan. You ask them to spend another 150 yuan on an AI module, and their entire business model changes."

Two months later, Zhou Yixu changed his approach. He went to the Canton Fair to face overseas buyers directly. Many overseas customers were already using ChatGPT, could understand, and were willing to pay. "When I went back to toy manufacturers with money in hand, everything went smoothly."

This story carries a lot of information. It illustrates a counterintuitive fact: in AI hardware going global, "overseas buyers who can use ChatGPT" are easier to communicate with than "Shenzhen toy factories that have been in business for twenty years." It's not that the supply chain isn't good—it's that the cognitive gap hasn't been bridged.

Gyges Labs took another path. Their AI glasses, co-produced with partner brand Haliiday, raised $3.3 million on Kickstarter, setting a crowdfunding record for the AI glasses category. This money wasn't just financial validation—more critically, it captured a batch of early adopters "willing to pay for novel and unusual products."

When it came to making the Vocci ring, the strategy changed. Because the ring is positioned as a productivity tool—helping you record meetings, capture key points, manage time—the target users are no longer novelty-seeking geeks, but "Pro users who run ten-plus hours a day, have five or six meetings, and forget what was discussed as soon as the meeting ends."

"Our founding team is typical users," Deng Xudong said.

For reference, Plaud, which also makes AI recording devices, reached $250 million in annual revenue in 2025, shipped over 1 million units, with half of revenue coming from AI subscription services. This shows that the demand for "helping you remember what was said in meetings" has been validated. Vocci's bet is: the same demand, in a more frictionless hardware form (a ring), can it unlock a new batch of users?

The story Wu Qilin, founder of Lingxiao Technology (灵蛸科技), told surprised me the most.

Her company makes AI companion desktop pets, with a team from Harbin Institute of Technology and the University of Waterloo in Canada. Mentioning Waterloo, she directly said something all international students know but few publicly discuss: "One thing the University of Waterloo is known for in the Canadian international student circle is being a 'suicide university'—every year someone jumps off a building due to academic pressure."

She originally wanted to enter the female companionship track—"by conventional understanding, women are more emotional and need emotional companionship more." But what changed her mind was precisely the group of male engineers on her team who graduated from Waterloo.

"On the contrary, it's the men on our tech team who feel they need emotional companionship. They need a desktop pet to remind them to drink water, remind them they should take a break—you can't get stuck in academic pressure, you have to go out."

So she made a decision: don't segment users by gender, segment by stress level. People in long-term high-pressure environments, remote workers—regardless of gender—are target users.

The value of this insight is that it pulls "emotional companionship" from a somewhat emotional concept back to a measurable dimension—how high is your daily stress? Much more precise than "are you female?"

Where Is the Moat: Paving the Way for Giants, or Rushing to Build Barriers Before Giants Arrive?

This is the question I kept thinking about while listening to this roundtable.

Zhou Yixu said something very honest. He found that the object users are willing to chat deeply with "must be a character they already really like." If you give users an original, never-before-seen cartoon character, "they might not even be willing to hang it on their bag." But if it's an IP or character they like, "they're willing to hang it on their bag, put it on their desk, and chat deeply with it."

This leads to a cruel inference: IP is the key to companionship, and startups almost cannot own top-tier IP.

Zhou Yixu himself said: "If Disney enters the market, everyone might quickly be pulled over."

So where is his moat?

The answer lies in an inconspicuous detail. He mentioned that through the user agreement, chat data "belongs to the user," but users allow them to use it to "improve model quality." That is, although Saibo Chuangli doesn't own IP, by making AI agents for IP, it has accumulated dialogue data like "what content users are most interested in about this IP."

"In the future, if this IP owner wants to choose the best dialogue agent model, we might be the better choice. Because the technical differences may not be big, but we know what questions your users want to ask."

This is a "water seller" strategy—not competing for IP, but competing for the position of "the person who best understands this IP's users." Can it hold up? It depends on two things: first, after large model technology homogenizes, how valuable is scenario data; second, will big IP owners do this themselves. If Disney decides to train its own dialogue model, Saibo Chuangli's data advantage could vanish overnight.

My judgment is: these startups are indeed "paving the way for giants" at this stage, but they're trying to use the time gap to rush and build a thin barrier. The barrier isn't in hardware, isn't in IP, but in dialogue data and operational know-how for specific scenarios. How thick this barrier is, honestly, nobody knows.

Deng Xudong's thinking is completely different. He doesn't obsess over the moat of a single product, but draws a longer line: "What we want to build is the next-generation display interaction computing platform."

Glasses solve display, rings solve interaction, and next is computing. Their company, Gyges Labs, is registered in Singapore, raised a Pre-A round led by GSR Ventures (金沙江创投), with teams distributed in Shenzhen and the US—"using the capabilities of both bay areas." Shenzhen provides supply chain, Silicon Valley provides cutting-edge AI capability, and Singapore serves as the globalization operations hub.

This kind of global architecture is itself a moat: most Shenzhen hardware companies can't achieve it.

Tang Chang, on the other hand, bets his moat on the concept of "AI Native interaction." His company's full name is Natural Operation Interface—"natural operation interface." The ring is just the first carrier. "First serve these few thousand users well, iterate step by step, and eventually something innovative will fall back into the general consumer market."

Three completely different moat strategies: data accumulation, platform architecture, interaction paradigm. What they have in common is—none have been validated yet. This track is still too early—early enough that even though Ray-Ban Meta sold 7 million glasses in 2025 (82% of the global AI glasses market), Deng Xudong feels that "in consumer electronics, a product volume under 100 million makes it hard to say the industry has reached a mature stage."

The Reefs of 1-to-N: The Shadow of Character.ai and the Truth of the "Loneliness Economy"

When discussing going-global compliance, Wu Qilin directly mentioned the elephant in the room.

"Previously Character.ai was sued for inducing a minor to commit suicide. That minor seems to have actually died by suicide, and his mother sued Character.ai."

This isn't hearsay. In February 2024, 14-year-old Sewell Setzer III died by suicide after developing deep emotional dependence on a Character.ai chatbot. His mother, Megan Garcia, filed a lawsuit in October of that year. In May 2025, a federal judge rejected Character.ai's First Amendment defense, allowing the lawsuit to proceed. In January 2026, the two sides reached a settlement.

This incident directly drove the legislative process of the US Senate's CHATBOT Act and GUARD Act, serving as a wake-up call for all AI hardware companies involving emotional interaction.

Wu Qilin said that after her team saw this news, they "immediately thought about how to reduce risk." Specific measures include sensitive word monitoring, high-frequency negative emotion warnings, and when danger signals are detected, "persuading users to seek more real external help." Technically, she mentioned processing data on-device as much as possible and using federated learning for model updates.

Are these measures useful? Frankly, for a startup team founded only in September 2024, achieving comprehensiveness in data security, content moderation, and cross-border compliance is extremely difficult. But at least it shows they're aware of the severity of the problem.

What's more worth discussing is a fundamental question: has the "emotional companionship" track actually been validated?

Wu Qilin's judgment is optimistic: "Overseas users have no upper limit on what they'll pay for emotional companionship." She analogizes AI companionship to an affordable alternative to psychological counseling—overseas psychological counseling is itself expensive, and AI companionship can enter the market as "adjunctive therapy."

But she also admits: "The companionship track isn't a single hardware sale—it depends on users' long-term willingness to pay, and generally requires subscription fees to keep a company alive long-term."

This is precisely the biggest question mark right now. Plaud achieved $250 million in annual revenue through the "hardware + subscription" model, with 50% from subscriptions—but it sells an essential function of "helping you remember what was said in meetings." The long-term subscription logic for emotional companionship is completely different: users' novelty will fade, AI's long-term memory and emotional depth are far from mature, and once a safety incident occurs (as with Character.ai), the entire category could be hit hard by regulation.

My judgment is: the emotional companionship track has successfully run through "traffic acquisition and ice-breaking"—users are willing to pay the first sum for a fresh emotional experience. But large-scale long-term subscription repurchase is currently still a beautiful story built on expectations. Everyone is waiting for that one qualitative change in technology and business model that turns "companionship" from a novelty into a habit.

No iPhone Moment, Only Individual Cracks

The final question of the roundtable was: if in 2026 you had to all-in all resources on one product form for going global, what would you choose?

The four people's answers were strikingly consistent—they all chose what they're currently making. Deng Xudong still does wearables, Tang Chang still does rings, Wu Qilin still does desktop pets, and Zhou Yixu said "might expand to more forms."

No one said "when AI glasses mature I'll make AI glasses," and no one said "when large models get stronger I'll pivot to software."

This indicates two things.

First, they have conviction in the path they chose—or rather, they've invested too much to pivot.

Second, perhaps the "iPhone moment for AI hardware" is itself a false proposition. There won't be a single device that unifies the world. A more likely future is: glasses occupy one scenario, rings occupy one scenario, desktop pets occupy one scenario, plush toys occupy one scenario—everyone has three or five AI hardware devices scattered around them, each managing its own.

Deng Xudong said something I think can serve as the ending of this roundtable: "When AI is on your computer, it replaces people; but when AI is on ourselves, it makes people into superheroes."

Is that too big a statement? Maybe. But in a 2026 where everyone is anxious about whether AI will replace them, a group of people in Shenzhen are wearing AI on their bodies, hanging it on their bags, putting it on their desks, trying to prove that AI can also make people stronger—that fact itself makes you want to keep watching.

As for who survives long enough to be seen, that's another story.

More Conversation Details

Guests:

Saibo Chuangli Founder & CEO — Zhou Yixu

Nuoyi Innovation Founder — Tang Chang

Gyges Labs (Vocci) Co-founder — Deng Xudong

Lingxiao Technology Founder — Wu Qilin

Moderator: InnoVoxa Technology Brand Going-Global Partner — Michael

Michael: My name is Michael. I'm at a Singapore company called InnoVoxa that helps Chinese consumer brands enter North American offline retail channels, and I'm also the moderator of today's roundtable. Here in Shenzhen we talk about growth, and especially for hardware, not talking about going global doesn't feel sexy enough. So the four roundtable guests in this session are representatives of hardware enterprises in different forms across various industries. Before our roundtable, let's first ask the four guests to take one minute to introduce themselves and the business behind their companies.

Zhou Yixu: I'm Zhou Yixu, founder of Saibo Chuangli from Beijing. We're making AI companion hardware, and we're more focused on the AI trendy toy field, more oriented toward young people's scenarios. Of course, hardware is just a carrier—what's more is the software behind it, the companionship, the growth of its relationship with users, user stickiness, etc. We started this direction in early 2022, and should be considered relatively early in the industry to enter AI companionship. And overall, as Michael just mentioned going global, we're also a relatively early AI companion hardware company to go global.

Tang Chang: I'm Tang Chang, you can call me Charlie. My company is called Nuoyi Innovation (abbreviation of "Nothing's Impossible"). We're a hardware company targeting AGI. Our main product now is called Spark. It's a wearable, voice-driven smart ring with vibration feedback that builds the minimal interface for human-AI interaction. It's currently doing initial market promotion overseas. If you're interested in our company, you can search online for more cooperation opportunities. Welcome to communicate with me.

Deng Xudong: Hello everyone, I'm Deng Xudong, co-founder of Gyges Labs. Our company is a Wearable AI company—simply put, we want to wear AI on people. The ways of wearing can be various, and ultimately through AI, people's capabilities become stronger. So our company slogan is "To make everyone superhero." Our company was founded in 2022, and in 2025 we co-released our first smart glasses brand called Haliiday, and in 2026 we independently released Vocci, a voice-interaction ring. I see today's theme is quite focused on overseas and going global—actually our company also has very deep going-global experience, including four founders who have lived and worked overseas for many years. I myself am a going-global veteran, having been doing smart hardware-related things around the world since 2010.

Wu Qilin: I'm Wu Qilin, co-founder of Lingxiao Technology. Our company makes companion robots. We were founded in September 2024. Our company is mainly a startup formed by a team from Harbin Institute of Technology and a student team from the University of Waterloo in Canada. We've currently made some IPs for provincial-level cultural tourism, and for Australia and New Zealand we've also made some going-global AI toy products.

Michael: So our roundtable today, as Mr. Deng just mentioned, is about emotional companionship monetization—especially how to find our real users, paying users, overseas. The first question I think has three perspectives. The first perspective is of course because our four guests represent different product forms, so first we can discuss some questions from the direction of product form: why did your company choose this entry point? First I'd like to ask Mr. Deng—we know that, as you just said, you position as a Wearable AI technology company, right? You now have two products already on the market, one is the AI glasses Holo that was very popular at CES last year, right? Including your AI ring that should now be on the market—I don't know if I'm pronouncing it wrong, is it called Vocci?

Deng Xudong: Very accurate, very accurate.

Michael: Actually, in the past we've seen many AI glasses makers competing on cameras, and ring makers might be competing on health, but you haven't done either. That is, you've abandoned these standard features. I'd like to understand—is this your product philosophy, or is there some commercial consideration behind it?

Deng Xudong: That's a very good question. Actually, the Wearable AI we want to build, in its ultimate form, is seeking a device that is the next-generation display and interaction computing platform beyond the phone. What exactly is it? We've also seen many big companies do a lot of exploration—for example, the Vision Pro came out with such a product, it's big and comprehensive with very powerful functions and actually quite a good experience. But the only problem is you can't take it out. So to solve the problem of being able to take it out, you have to make some trade-offs. So one of our company's philosophies is, because we've been in Shenzhen smart hardware for many years, we actually have reverence for hardware. This reverence guides what we do: "restrained innovation." Although our first-generation glasses had the world's smallest display system, which is very innovative, at the same time we're very restrained. So as the moderator just said, we didn't do the camera function that everyone is doing—instead we cut the camera function, and also made the voice function very weak. What we focus on is a concealed display and anytime-anywhere display. Like the glasses I'm wearing, you probably can't tell at all that they're smart glasses with a display, because we made the optical waveguide module and this system very, very small. So back to your question—why don't glasses have cameras? Because if you add a camera it gets heavy, if it's heavy you don't want to wear it, if you don't want to wear it it gathers dust, right? So our philosophy is restrained innovation, and after making the first product successful, we continued this philosophy into the thinking for the second product.

Because when we made the first pair of glasses, we found that many consumer usage scenarios were for meeting minutes. But do you have to use a pair of glasses to do meeting minutes? Definitely not necessarily—a ring might be more convenient. Based on consumer insight, we made the ring. For rings, you see maybe over 95% of companies will make something health-related, and everyone crowds in. But what we thought about was bringing consumers a very frictionless interaction method, a fingertip agent anytime anywhere, so we made this voice-interaction ring. This is also a product made under the guiding ideology of restrained innovation and maintaining reverence for hardware and products. As for the ultimate form, I think in another 10 years, we definitely won't only use phones for our display, interaction, and computing. On the road to the next-generation display interaction computing platform, I think everyone here is also a contributor, and we very much look forward to such products bringing consumers a better user experience.

Michael: Mr. Deng's product philosophy just now should embody very long-term thinking.

Deng Xudong: That's right.

Michael: In a commercial society with such big temptations, doing restrained innovation is easier said than done—we should learn more from Mr. Deng. Another small question I'd like to ask Mr. Deng: do your glasses currently have a teleprompter?

Deng Xudong: I see I can turn it on, but just now it wasn't on. Because these things may have been said 100 times during roadshows.

Michael: This doesn't even need a teleprompter. But if, say, I went to do a stand-up comedy last time, it could be used—but you can't stare at it, right? If you stare, you lose eye contact with the audience.

Deng Xudong: You can still type out these key words as prompts, it's quite concealed.

Michael: I'm thinking next time at this roundtable event maybe I don't need to prepare in advance either—just whatever question is asked, tell me how to say it, and I'll just read along.

Deng Xudong: That's possible, but behind it someone still needs to read it with emotion.

Michael: The second question I'd like to ask Mr. Tang. Actually, the restrained innovation Mr. Deng just talked about—what Mr. Tang is doing also falls into this category. Mr. Tang, because your current product form has a ring on the market, and rings—as just mentioned, whether overseas Oura or domestic RingConn—everyone is competing in the health track. Is your ring also detaching from this? I'd like to ask a direct question: why would overseas users be willing to buy another ring, but this ring can't monitor my health, can't monitor things like my sleep or exercise?

Tang Chang: Because many people ask why someone would buy an extra ring. Actually, this question depends on whether the two types of people buying different rings are the same demographic. Actually, people buying our type—like lightning recording, inspiration recording, or informal short-meeting recording—may use it more during the day; health rings are more for at night. We validated this in a very typical way—going to overseas CES to communicate offline with our users, also collecting emails, sending them some questionnaires to see their different purchase willingness for the two types of rings. The people who contact me more are our type of daytime users, who may value the recording itself more, and they value structured organization, value reminders. These may be the insights we've gained from overseas users—it's decoupled from health. Domestic users, like me, am actually also a heavy user. I'm actually like Mr. Deng—many meetings a day, frequently shuttling around, and I feel some thoughts I haven't captured are very important, so I need low-friction, low-resistance recording, and I resonate with this type of user, which is different from health.

Michael: Actually, these are two although the form is the same—both rings—but their users basically don't overlap. This is a different track, right? Everyone's user overlap is basically zero, right?

Tang Chang: Their overlap isn't that high, but in terms of spending power they're both a group with relatively strong spending power, so there may be some overlap. But now with AI, everyone's differences are actually quite large. Like when I first arrived, I originally wore an Oura wanting to do multimodal recording, but now some of my devices somehow disappeared, so I might also wear an Apple Watch, and I can still record sound. Actually, one person can have many AI devices. Maybe right now, every geek has N multiple such devices on them—not necessarily only one voice recording device.

Michael: Right, now everyone is doing subtraction. Maybe what the ultimate form will be, I think it all takes time, and future views differ. Another question I'd like to ask Ms. Wu—because you do AI companionship, and now companionship forms are actually very diverse, like some desktop pets, including Mr. Zhou's plush toys, and even some AI pendants, which I think also belong to this AI companionship track. What user profile did you choose? Also, what is this group's willingness to pay overseas, or roughly what level is their psychological price ceiling?

Wu Qilin: Actually, our company internally doesn't distinguish our intended users as male or female. Because in our research process, we actually found that many male users have a strong need for emotional companionship, but they don't realize they have this need—you have to implicitly uncover this point. So based on this, our company's initial definition was to go the female companionship track, because by the most conventional understanding, women are more emotional and need emotional companionship more. But actually in the real research process, including within our team—because our tech R&D personnel all graduated from the University of Waterloo in Canada, and the University of Waterloo has a very famous point in Canadian international student circles—it's a "suicide university." It's famous because every year someone jumps off a building because academic pressure is too great. On the contrary, it's the men on our tech team who feel they need emotional companionship. These men need a desktop pet to remind them to drink water, remind them it's time to take a break, take a break, it's time to go out and see the outside world. You can't get stuck in academic pressure or whatever, you have to go out.

So because of this point, our company stopped doing gender distinction. We more judge whether this user is in a long-term high-pressure environment, whether this user is in a work-from-home type of situation. These two are actually what we target more, moving away from gender boundaries. And including what you just mentioned about overseas companion users' psychological price ceiling for this AI companionship—we actually feel overseas users have no ceiling for emotional companionship. Many users define this emotional companionship as like a pet, like a psychological therapy, it's a bit like an adjunctive therapy concept. As we all know, overseas psychological counseling is itself expensive—whether in China or overseas it's an expensive existence. Of course, we're not trying to replace professional therapy; we want to make an adjunctive entry into this market.

We don't think there's a particularly big gap in willingness to pay between domestic and overseas, but relatively speaking, domestic users value more what value this thing can bring me. But overseas users may feel whether this thing is meaningful to me, whether it's good, whether it truly accompanies, whether it truly solves the pain point I bought it for initially—and they're willing to pay. And the companionship track, I think, isn't a single hardware purchase profit model—it depends on users' long-term willingness to pay, and generally everyone still needs subscription fees to keep a company alive long-term. And we also need to continuously update, iterate, and learn, and we also need to do things like data protection—it's not a single one-time transaction thing anymore.

Michael: And Ms. Wu, we may be talking about the next step of going global, from 1 to N, some cultural compliance, privacy directions. My last question is for Mr. Zhou. Actually, making IP content for AI plush toys is, in my personal opinion, a world-class challenge. You see, for example, some domestic IPs are actually hard to go global—like "Pleasant Goat and Big Big Wolf" (喜羊羊与灰太狼), right? Domestic audiences like it, but overseas audiences' perceptions are still different. So when you make this kind of IP going-global content, do you rely on your own hardware platform as you just mentioned, plus localized IP open-mode access cooperation, right? In this model, how do you balance the narrative, copyright, cost, and your bargaining power over local content?

Zhou Yixu: For us, as just mentioned, we actually prefer to cooperate with localized IPs. Of course, since this is an IP business, everyone knows that IP may take the larger share of the profit. But there's actually no way around this, because we believe in the early stage people may choose some new AI-native IPs, or even ones without much backstory, just a cute look, to buy this kind of toy. Mainly because the real IP giants haven't entered the market yet—for example, if Disney enters, everyone wow, they might quickly be pulled over. For bargaining, right now we're taking two approaches: first, cooperating with IP owners; second, we're also trying to use AI to create some new IPs. Of course, this actually sounds more like two paths—one is using AI to try to create some new IPs, and if successful, we'll consolidate these IPs into our AI + hardware offering. But at the current stage, we're still cooperating with more IP owners.

Michael: Because the music just started playing to rush us again, last question—each guest, if the product form you're currently making, if in 2026 you had to all-in all resources on one product form for going global, which form would you choose and why? Let's start with Mr. Zhou, okay?

Zhou Yixu: I think the future form may be hard to determine as one specific form, because companionship—I more think it will appear in various places in our lives. Like the current IP or companionship thing, we can divide it into two parts: one is creating a sense of companionship, the second is real companionship interaction. Creating a sense of companionship is now already in our lives through IP co-branding—you buy a snack and might choose an IP package, buy an appliance and might also choose an IP co-branded edition. Then in the future, companionship interaction will also blend into all aspects of our lives.

Michael: So Mr. Zhou will still choose the product form he's currently making. Understood.

Zhou Yixu: Might expand to more forms.

Michael: Okay, Mr. Tang.

Tang Chang: For me, because I'm personally more drawn to efficiency and minimalist style, I'll keep making my ring. I think the ultimate answer in the future I can't predict right now. Many people are making various AI companionships now—could be toys, could be glasses, could be all kinds of things—I can't say. But personally, I'll continue investing in my current product direction. I believe it still has relatively good prospects.

Michael: Right, AI rings are relatively a frictionless kind of wearable, right?

Tang Chang: Right, frictionless, ultra-portable.

Michael: Okay, Mr. Deng.

Deng Xudong: Because I just saw news that Meta seems to have laid off another 5,000 people.

Michael: Everyone says AI will replace people.

Deng Xudong: As for AI replacing people—actually when AI is on your computer, it really does replace people. But when AI is on ourselves, running on what we wear, it's to make people—to make everyone superhero. So I'll keep making wearables—we make humans irreplaceable.

Michael: So still wearable products.

Deng Xudong: Wearables make ourselves stronger. Including actually I think wearables, including what's now called exoskeletons, are also very interesting—maybe at some stage we can consider that, but in short from the wearable perspective, making people stronger.

Michael: Not necessarily glasses, not necessarily rings, could be pendants etc.

Deng Xudong: We make all kinds of Wearables.

Wu Qilin: Our company will also continue to hold this direction, because we actually believe AI desktop pets have a probability of becoming the concept of a second screen beyond the phone. We think long-term, because everyone's excessive eye use is also a big problem, and when people get used to having such a creature on their desktop, having an AI companion may just persist.

---

Original publication: https://uniqueresearch.substack.com/p/src-20260602-01html
On-site reading page: https://ffcap.cn/en/research/src-20260602-01html