Editor's Note
When we want a Coke, we usually only have two choices: Pepsi and Coca-Cola. When choosing a mobile phone, there's a 90% chance we'll be switching between Apple and brands like Huawei, Xiaomi, OPPO, and Vivo. When buying sportswear, the first brands that come to mind are most likely Nike and Adidas.
But the world is so colorful because, beyond these giants, there are companies that defy tradition, strive to create something different, focus on design and functionality, and look to tomorrow.
They have unconventional business models, and their designs and products offer unique user value and ample social talking points. Crucially, they are not burdened by the constraints of large corporations and dare to advance recklessly. They are "diversity companies."
Diversity is key to an open world. ifanr believes that only companies that truly value and understand diversity can foresee the future sooner than most. In this column of the same name, ifanr will share exclusive interviews with these diverse companies, witnessing how they are reshaping the future and defining the new normal.
This is the 10th article in the "Diversity Companies" column.
In 2017, Xu Chi resigned from Magic Leap and returned to China to found XREAL (formerly Nreal). At that time, the entire XR industry was celebrating Magic Leap's whale flip demonstration, and everyone thought that this was the future, but no one actually sold a single consumer-grade AR glasses.
Nine years have passed, and this field has experienced the bubble and ebb of the VR metaverse, the high-profile entry and lukewarm reception of Apple Vision Pro, the subsidized expansion of Meta Ray-Ban, and the re-examination of all terminal forms by the AI wave.
XREAL survived and became one of the first hardware strategic partners for Google's Android XR platform. According to IDC data, XREAL has maintained its position as the world's number one AR glasses market share for four consecutive years. And just recently, XREAL officially submitted its listing application to the Hong Kong Stock Exchange.
This smart glasses company, which has been dormant for nearly a decade, is about to enter a new phase of business.
This exclusive interview with iFanr was conducted before XREAL filed for its IPO on the Hong Kong Stock Exchange. In the conversation, Xu Chi did not shy away from any pointed questions—from "Why Apple's Vision Pro was destined to have problems," to "Chinese manufacturers are using supply chain integration to fight a first half of the war," and then to "No company in the eyewear industry has ever truly made money." But throughout, a clear judgment was consistently maintained:
Glasses are the best medium for AI because only they can provide the model with the highest quality context.
Xu Chi, Founder and CEO of XREAL
Surviving in an industry where no company makes money
Q: When you left Magic Leap to start your own business, you were working on a very cutting-edge product. Why did you decide to start such a company?
A: When I was at Magic Leap, the first few months were amazing. Suddenly, you're standing at the beginning of a new era and have the opportunity to be at the forefront of it. If you're lucky, you can even participate in defining it. That feeling is fantastic.
At the time, my judgment was that this was the next big opportunity, and it would definitely materialize by 2020. I came back in 2016 because I felt that if I didn't return soon, it would be too late. My thought at the time was that someone knowledgeable in this industry would definitely come back from abroad—like Robin Li and Charles Zhang back then. That person could be me, or it could be one of my colleagues, because there were only so many people who understood this area at the time. So why not you? You can't come back fully prepared. If things don't go well, I'll just go back. That was my simple thought at the time.
It's been a tough journey, with the industry experiencing ups and downs. But I've always stuck to one thing: we've never strayed from our original intentions. This is also a test of what each entrepreneur's inner drive truly is—is it for fame, for success, or for wealth?
We genuinely believe that eyeglasses are the next big thing, and that shouldn't be easy. Coincidentally, we entered this industry very early, almost with a sense of mission, eager to see what the final answer would be, and even wanting to stay with the industry until that answer is revealed.
Q: XREAL just celebrated its ninth anniversary earlier this year. In nine years, have you met your expectations?
A: First of all, it definitely didn't meet expectations; the whole industry doesn't meet expectations, but I'm still quite satisfied.
Given our understanding and enthusiasm at the time, we were quite lucky to get to where we are today. Along the way, we met many good people, kind-hearted individuals, upstream and downstream partners, and our own team, which is why we are where we are today.
Of course, if we were to go through it all again with today's mindset, we would definitely do it even better. That's the process of growth. I often tell my colleagues that if XR had a museum that recorded every step of XR's history, XREAL would have already left its own significant mark.
Q: Industry trends are constantly changing. Have you ever experienced your darkest moments? How did you overcome them?
A: Definitely.
Before an industry truly takes off, every darkest moment is often accompanied by some shining moments. The most memorable was probably when the pandemic first emerged. At that time, our overseas business was at its best because people needed such products while staying at home, and all overseas operators wanted to cooperate with us. Our CES debut was a great success.
But then the pandemic hit, people couldn't leave the country, financing was disrupted, the team was unstable, and internal and external conflicts erupted all at once. Internally, there were strategy and management disputes, and externally, some companies that were doing well suddenly dropped their partnerships.
Looking back now, I feel much more at ease because these were all processes that should have been taken for granted.
Q: In my opinion, Vision Pro replicated Magic Leap's technology, and even surpassed it. But Vision Pro didn't meet expectations. Was that a blow to you at the time?
A: We were actually quite disappointed at the time. I remember very clearly that once when I went to meet Mr. Xing from Meituan, he was also paying attention to this field. After we finished talking, he asked, "What is Apple doing?" I said at the time, "The product that Apple is making is probably not going to work."
At the time, many people in China believed that "everything Apple does has a reason," and you couldn't convince them. It's difficult to use a product that hasn't even been released yet to provide evidence. Later, if you wanted to say that Apple did something wrong, you'd attract criticism.
We can only go with the flow. But actually, I've felt there was something wrong with this Apple product for quite some time now.
Q: What is the reason?
A: I think this is the first Apple product ever that wasn't cut.
When Steve Jobs was at Apple, it was all about extreme tailoring—"I don't know what you want, I give you what you want." But the Vision Pro is clearly "I don't know what you want, so I give you everything"—it adds this and that, it's a product that piles on features.
It's said that this is indeed Apple's internal product logic. They are repeating the path taken by the Apple Watch—the first-generation Watch wasn't successful, but it gave them the opportunity for subsequent success, showing them that focusing on health monitoring and exercise was the right direction.
The initial idea behind AVP was also to avoid making judgments and try to add as many features as possible to see what users would prefer. However, their mistake was that adding too many features to the headset made it too heavy and uncomfortable to wear.
As a result, the first-generation product did not provide Apple with any feedback on "which direction the next generation should take," because the sample size was too small. Therefore, they will likely be more conservative in their next step.
Q: Your main products currently being shipped are actually large-screen mobile devices. When did you decide to forgo spatial calculations and instead focus on developing the large-screen mobile devices first? Why do you believe that this positioning of large-screen mobile devices is the right one?
A: This wasn't something I judged; it was something I was proven wrong by. Our current situation is truly a result of our journey. Exploring uncharted territory, genuine user feedback is extremely important.
Our first-generation product was designed to be smaller, cheaper, and better. The idea at the time was to partner with telecom operators, who had local influence, brand endorsement, channels, and ecosystems, while we provided the technology, handling both the hardware and software.
We once created what we considered the most complete commercialization loop in South Korea: pre-installed apps on phones, glasses bundled with phones, 5G contracts to drive down prices, sales through carriers and Samsung/LG channels, and LG finding local content to build an ecosystem. This is the most complete ecosystem we've seen so far, but it wasn't successful—because neither we nor the carriers had real platform appeal.
Only then will you start to reflect: who is truly capable of building a platform?
I'm making a bold prediction: only Apple and Google will be included. Not even Meta, not even OpenAI.
Because of their momentum and accumulation in the mobile phone ecosystem over the past 20 years, they were the only ones capable of building a platform. At that time, my thinking was very simple—Don't do it.
Because if one day you develop a system, and Google releases a completely different system, you've essentially misled all your developers. What if the interaction logic is completely different?
So we have to go back and simplify. We come from a technical background, and cutting back in technical fields is the most painful. You have to tell people doing SLAM, "Sorry, we used to do six degrees of freedom, now we have to do three." You might say, "Anyone can do three degrees of freedom, right?" But there's no other way.
However, our original intention remained unchanged—although we focused on display technology, another line of thought never went unbroken. Until Google found us.
Q: How did you manage to get this collaboration with Google?
A: We've always maintained open-source connections with Google. They've been keeping an eye on us internally, including some Apple executives, who buy our new products as soon as they're released. The attention from your peers is probably the greatest recognition you can receive.
Until Apple released AVP, Google immediately made a decision to follow suit. However, they suddenly discovered that AVP was unsuccessful. There were two major takeaways to its failure: it was too expensive and too heavy. Because it was expensive, developers weren't interested, believing there wouldn't be enough market share within three to five years. Because it was too heavy, consumers had no intention of wearing it long-term or continuously.
The real solution lies in making it affordable and lightweight. XREAL has been focusing on lightweight design and modular construction from day one. Leveraging our long-term accumulation of core technologies in space computing and our excellent domestic supply chain capabilities, we are also more competitive in terms of price. Thus, this matter became a natural progression.
XREAL's Android XR glasses project, Aura, in collaboration with Google.
Glasses are the best medium for AI.
Q: Whether it's spatial computing devices or AI hardware, what should the ultimate form of smart glasses be? Some people in the industry have mentioned a division from L1 to L5, do you agree with that? Because in the field of glasses, the current L1 experience is far better than L5, which is quite strange.
A: I previously gave a definition from L1 to L5, mainly a classification of intelligence levels—in the early stages, they could be used occasionally, but later they became more and more like your own personal assistant. But why are lightweight glasses destined not to replace everything? Because of the physical limitations of display and computing power.
If you want to add a display, the most common approach now is optical waveguides. However, even at its best, the display capabilities of optical waveguides are only comparable to those of in-car head-up displays (HUDs). It's fine for translation and navigation, but you wouldn't use an in-car HUD for watching movies or playing games. Furthermore, we've become spoiled by Retina displays—while Retina displays are the foundation of a display, they also require a lot of GPUs to render more pixels. If this were done on a lightweight, always-on device, the battery life wouldn't be sufficient.
Therefore, we must make trade-offs: there is a lighter device that can be worn all day, but with a weaker display; and there is a relatively heavier device, but with a portable form factor and display capabilities on par with today's Retina displays. These two are inherently separate.
Q: So you think that in the future, a pair of glasses won't solve all problems?
A: When people think of glasses, they might think of different forms. The Meta Ray-Ban is one form, what we're working on now is another, and the large helmet is yet another. It's not a matter of choosing one of three. Just like today you have a mobile phone, tablet, laptop, and desktop computer, they meet different scenarios and have different priorities.
AI glasses are meant to be worn all day, so they must be lightweight. The second form is our current mobile form, which is portable rather than always worn. The advantage is that it can be slightly heavier, but it can be worn during work and displays richer content. On the other side is a large helmet, including AVP, which offers an absolutely fantastic experience, but it might be more like a dedicated home device.
We believe these three forms will coexist for the next 10 years or even longer, and no single device will replace everything. Just like in science fiction movies where watches were envisioned replacing phones, unfortunately, we still wear both watches and phones today. Some things have physical boundaries.
Q: I have a mobile phone and a computer, why do I need to use glasses to replace them?
A: I used to think that today's computers and mobile phones have compressed the entire world of internet information into a two-dimensional rectangular grid. True three-dimensional perception, three-dimensional display, and the fusion of the virtual and real worlds are inevitable. But recently I've had a new thought—maybe that alone isn't strong enough, not enough to make users feel, "I have to do this."
This is the new answer we've come up with after more than a year of reflection: we should thank AI, as it may bring us a completely new way of interacting. In the past, whether it was a computer or a mobile phone, it was essentially a human controlling a machine. Keyboards are efficient but have a high learning curve, while touchscreens are relatively efficient and have a low learning curve, but they still haven't escaped the paradigm of "human controlling machine." Apple uses eye tracking for 3D interaction on AVP, which is extremely inefficient; it's essentially interacting on a 3D canvas.
When AI emerged, it was a revelation. The true next generation of interaction will no longer be about humans controlling machines, but about humans communicating efficiently with an intelligent agent, just like we do now. In the future, your phone, computer, and glasses will all have an intelligent agent, communicating through the five senses in a way that connects people.
Q: Many AI hardware devices nowadays, such as headphones and pendants with cameras, also serve as AI input. How do you view the competition with these devices? They are cheaper and have even wider application scenarios.
A: Let's go back to first principles. Why are glasses the best platform for AI? Because when you add eye tracking in the future, glasses may be the only device that can know your focus point.
For example, whether it's headphones or other devices, if they want to take a picture and analyze it—for example, if there are three people sitting in front of you, who are you looking at?—uploading the entire picture involves a huge amount of computation. But with eye tracking, I can detect that you're looking at a specific person; I can even crop out their silhouette and only upload that one image to the cloud. Humans are naturally like this too; when I'm focused on talking to you, I might only notice your facial expressions and not pay attention to the trees behind you. Only glasses can do these things.
Essentially, this is very similar to the principle of LLM—the attention mechanism. Glasses are the easiest terminal to obtain the highest quality context.
Q: I tried Project Aura yesterday, and I felt that with a truly usable display, many productivity scenarios become possible with the help of the AI Agent. For example, I can do without a computer—as long as I can give commands, clearly receive output results, and determine whether the Agent's delivery meets expectations, that's enough.
A: You've made an excellent point. Now imagine you're the CEO of a company, and the AI Agents are the various employees. How can you make these employees understand your instructions more and more accurately?
It's not about you paraphrasing in words—because words might distort some background information—but rather that the person involved has been involved in many of the scenarios in your work. When you repeat an idea to them, they might say, "Oh, you came up with that idea in that scenario, you mentioned it while chatting with someone," because they have more background information and are more likely to complete the task more accurately.
Therefore, I need to elevate the input of the AI Agent, to turn it into a contextual input, rather than just abstract text.
Q: If you were to develop AI glasses in the future, what would you like them to look like?
A: I hope it can truly provide me with insights from a third-party perspective, insights I might not have noticed myself. I'm still looking at it from the perspective of a personal assistant. I hope it can help me review my work at the end of the day, offering angles and things I hadn't considered from a first-person perspective. Therefore, it needs to be available 24/7 and multimodal.
Q: Doesn't this contradict your current direction in developing displays? Your technical expertise is more focused on displays, but the scenario you just mentioned seems to be possible without a display.
A: What XREAL does well today is that when we solve a problem, we go back to first principles and then use a more difficult approach to solve it. Just like we don't make chips for displays. Like why Tesla can build cars—someone who used to work in payments can build cars? Why can someone who makes cars build rockets? It's not because "this is the closest thing, so I'll do it." What's impressive is that they consistently follow first principles—how to solve a problem using a seemingly complex, but actually the most direct, method.
Q: In your opinion, what are the first principles of XREAL?
A: A multimodal, all-weather AI device—with at least eight hours of battery life, plus long-term memory—is a highly monetizable AI personal assistant.
Our core goal is to create an AI personal assistant. The question is whether to prioritize 24/7 operation, display output, or multimodal functionality. Each step is a necessary step towards becoming the ultimate personal assistant. This idea truly took shape after multimodal AI matured. Multimodal AI expanded the boundaries of what we believe are the capabilities of this field. My initial vision was for a smaller, lighter, and cheaper device.
Long-termism in the Chaotic Era
Q: What do you think is the core value of smart glasses?
A: The core value of glasses lies in the fact that they represent the best form of sharing high-quality context and attention with the model. Today's context is similar to a CPU cache, a kind of short-term memory. Long-term memory, on the other hand, is a completely new memory system. This will emerge within the next two to three years, and it's something that everyone in the agent field has been researching.
Q: Is this an industry consensus, or are many people who make eyeglasses just interested in making eyeglasses?
A: When the iPhone came out in 2007, it wasn't a consensus. We've actually entered a chaotic era today. Just like back then, no one could predetermine the answer; we only say Musk was brilliant and Jobs was amazing in hindsight. But that period was a chaotic era to some extent; everyone was searching for the answer.
But what I want to say is that when this industry is about disruptive innovation, it's unlikely that a situation like in a martial arts novel will occur where a mysterious master suddenly appears and wipes out everyone. This industry places great emphasis on research and development. The ultimate product in this chaotic era, the iPhone Moment, is very likely not something that happens in the middle of the supply chain.
Q: Many domestic manufacturers have already entered the 1000-yuan price range, and there are more and more noisy products on the market. How do you maintain your brand image? What is the fundamental difference between you and companies that integrate the supply chain?
A: If we keep emphasizing originality, but actually can't outsell companies with integrated supply chains, then it probably means that our original elements lack differentiation. I believe our products are differentiated, but difficult things take time.
Since XREAL's successful display glasses business in 2022, I've been constantly thinking about where our brand should be positioned. We aspire to produce mid-to-high-end products, and brand recognition needs time to develop. Time is the biggest enemy for startups, and we must be patient.
JK from Insta360 once said: "A brand is the trust that consumers place in you when they don't have enough information."
We especially value this trust. You might need several generations of products to build it, but just one bad product can destroy it. So in this process, we are no longer just pursuing high-speed growth, but high-quality growth.
For years, what we've been doing is ensuring we lead the industry in changing user experiences: chips, wide-viewing-angle optical engines, and real-time 2D to 3D conversion. I believe these will gradually solidify in consumers' minds. Naturally, some will try to take shortcuts through marketing, attempting to create an impression of "I'm similar to you," but I believe time will prove everything.
Q: The AI industry is changing almost daily this year. As a hardware entrepreneur, do you feel anxious?
A: The logic is the same as stock trading. If you're always in the market, watching the fluctuations every day, it's easy for short-term volatility to affect your judgment and mood. If you take a long-term view and broaden your perspective, you might have a clearer picture.
The core challenge is testing your long-term strategic resolve. Before DeepSeek's meteoric rise, the names people in China heard of were Kimi and Doubao. DeepSeek didn't choose to advertise alongside other companies at that time; instead, it quietly focused on its own development until one day, overseas observers discovered it had even shaken Nvidia's stock price. We probably feel that doing the same thing is more appropriate for us.
Our previous foundation gives us some leeway to wait. Many companies today are forced to release glasses, to create glasses only in presentations, because they need to survive until the next stage—just like when they were building cars, everyone was still making glasses based on PowerPoint presentations. But I think it's good that we can take a step back and think more long-term.
Q: Google did a lot of promotion at CES, but didn't launch any products. Are you worried that your platform's development is too slow? Will your product compete with Google's?
A: Actually, Google's CES event is a small-scale, closed-door invitation event. They invited many people to attend, including us, who spent half a day in their meeting room meeting different partners. I'm not afraid of them being slow; I'm afraid of them being fast. Because a platform needs a rhythm; it's not enough for just the platform to be launched. There also needs to be key content and an ecosystem. We are very satisfied with the current situation.
Furthermore, I feel that there's a bit of a rush in AI development in China today. Everyone seems to be racing, feeling that if they're six months late, they'll miss out. But I don't think defining this next-generation interaction paradigm in AI is about getting a head start; it's a long-distance race, and getting on the right track is far more important than rushing ahead.
Google will do what it did with Android. I believe that at some point they will have their own Pixel, but they will definitely focus on building the platform first. This is our very clear strategy. So we are not worried about competition in the short term. They may be our best partner—they do what we can't, and what we do happen to be what they need most.
Q: Glasses will most likely go through a process similar to that of mobile phones and new energy vehicles, going from the first half to the second half. Where do you think we are now?
A: The eyewear market will likely follow a similar path to smartphones and new energy vehicles: leading manufacturers will continuously invest in R&D, achieve breakthroughs, rapidly iterate on their products, and establish industry rules. Then, downstream players in the supply chain will reduce costs and empower more manufacturers. Most Chinese manufacturers are familiar with the latter half – making small iterations, incremental innovations, and large-scale manufacturing on products already defined by others. But the eyewear industry hasn't even reached that second half yet.
What I least want to see is this industry using supply chain integration and marketing to fight a battle in the first half.
Because the first half of the game still requires technological innovation and iteration. Personally, I don't think any of today's products have reached the wow factor of the original iPhone 1. And that iPhone Moment is unlikely to have come from a fourth-rate company that only integrates the supply chain.
While eyewear is currently booming, no Chinese eyewear manufacturer has yet achieved sales of over one million units for a single product. Globally, only Meta stands out, but Meta relies on subsidies. The true turning point for this industry will be assessed without subsidies.
Hand-drawn posters from XREAL users
Q: Is your ultimate business model still selling hardware?
A: Of course not. Even today, model manufacturers haven't figured out their business model. What you really want to ask is, when a new terminal, a new interaction paradigm leading to a new terminal, emerges, what will the value chain distribution look like?
I believe we will definitely have a place. And because the device side is getting closer and closer to you, the attributes of the hardware or entry point side will become stronger and stronger. In the future, you may not be buying hardware, but rather paying a monthly subscription fee to have this assistant serve you.
If this assistant has been with you for three years, attended almost all of your meetings, and not just recorded data, but formed its own judgments and abstract long-term memories as if attending meetings, then you won't be able to do without it.
Q: Who owns the data? What does this mean for the future value chain?
A: There has always been a question in this industry: who owns the data?
Today, Samsung is giving its data directly to Google, using your data to monetize through advertising. But the data ownership originally belongs to the user. Furthermore, long-term memory will be decoupled from AI—just as CPU and memory can be decoupled.
When you have a large number of devices, you have more control over who you choose to give your data to.
Q: When Android XR or multimodal AI matures, all the major manufacturers will enter the market, leaving little time for startups?
A: You understand, right? Just like when we were making phones alongside Android, all the hardware manufacturers came in. You moved from one table to another, and everyone's chips changed. Time may be running out for startups, so maintaining differentiation and a fast iteration pace is crucial.
Everyone says they want to be like Apple, but Apple's greatest achievement is solving three problems: hardware manufacturing, system development, and how to connect the hardware and software in a complete interaction paradigm.
But many people might only associate it with Lenovo, or even Oracle. Different levels have different roles and earn different incomes. As long as I can secure a place in this ecosystem, that's enough, but it's too early to talk about a specific position now.
#Welcome to follow iFanr's official WeChat account: iFanr (WeChat ID: ifanr), where more exciting content will be presented to you as soon as possible.
Related reading: 
