A Conversation with Rokid's Misa Zhu: The AI Glasses Battleground Is China Alone

"Spring for an 11-Year-Old Company"

By Muxin Xu

Edited by Jing Liu

Suits poured into the office building in droves — some from local governments, some channel distributors, some livestreamers. The man they all wanted to see, Rokid founder Mingming Zhu, was debugging glasses at his desk. A month earlier, he'd used this very pair to demonstrate projecting speech notes onto the lenses, a stunt that went viral. That same day, Zhu mentioned another detail: not long before, on one evening, Feng Ji of Game Science, Xingxing Wang of Unitree, Wenfeng Liang of DeepSeek, Bicheng Han of BrainCo, and several other founders had come to his home for dinner. After the meal, these Hangzhou founders — all white-hot lately — signed their names on the same blackboard, which now sits in Zhu's meeting room. Rokid has become one of the most sought-after companies in Hangzhou's Dream Town.

But this isn't a new company. At 11 years old, Rokid had previously focused on consumer AR glasses and B2B verticals in energy, industry, and cultural heritage. In 2014, Zhu left Alibaba to found Rokid, backed by top-tier investors including Vision Plus Capital, Linear Venture, and Fosun, reaching unicorn valuation. Zhu's office matches the man — thoroughly geeky. His desk sits in the corner of a gridded open floor, ringed by a moat of original Apple computers and vintage keyboards. In interviews he always brings up Jobs — a reference point that seems to belong more to the previous generation of entrepreneurs.

Last week, Anyong Waves sat down with Zhu at Rokid's office. That same day, news broke that another AR glasses company backed by Alibaba, Singularity Near, had run into operational trouble by late 2024 — its founder hit with spending restrictions, operations now paralyzed.

Zhu was surprised to hear it: "Why collapse right before dawn? Rokid finally got the industry hot. If they'd just gritted their teeth a little longer." The day felt like a microcosm of AR's decade-plus history — bubbles and troughs, some crowned with flowers, others not making it to the finish line.

Below is the conversation between Anyong and Mingming Zhu (Misa):

Part 01

AR or AI

"Anyong": What did you talk about at that "legendary" dinner?

Misa: It was February 15. Wenfeng was going to the private enterprise symposium the next day, so he wanted to chat with us about technical topics.

"Anyong": Hangzhou's "little dragons" all come from different fields. What do you discuss?

Misa: It's all the same — technology is universal. Didn't I also get Black Myth: Wukong running on my AR Lite glasses?

"Anyong": Rokid's previous flagship products were AR glasses, but what went viral was AI glasses. Do you think the latter represents the real technical direction?

Misa: There's an impossible triangle in this industry: display capability, battery life, and compute power. These two product types each occupy two corners, but the endgame is a complete triangle. You'll find that AI and AR glasses are evolving to resemble each other more — AR needs to get lighter, AI needs better display. They'll eventually merge into one.

"Anyong": What obstacles stand in the way before that endgame?

Misa: Fusion will be possible in five years; until then, it's trade-offs. Right now AI technology has matured first, while AR hardware still isn't ready. So I'm putting 70% of my energy into spatial computing and software, 30% into refining AR technology with the industry.

"Anyong": You currently have both AR and AI glasses product lines. Which do users buy into more?

Misa: AI glasses orders are 10x AR's right now — and that's with F-codes restricting purchases. Without limits, it'd probably be 20-30x.

"Anyong": But IDC data shows overseas market share leaders are mostly on the AR track.

Misa: I basically don't look at that. Pure AR glasses aren't mass-market products, not real demand drivers. They can be a nice small business, a toy. AI glasses are a completely different tier of赛道 [track/sector].

"Anyong": How do AI glasses avoid being seen as toys and become real demand?

Misa: Anything you can't wear all day isn't real demand. Meta-Rayban's experiments these past few years, plus AI capabilities — I believe within a few years AI glasses will become essential. The most important thing for PMF is that your product definition can't be too far off. You need to make users pull out their wallets decisively, not convince them to. For example, Rokid's hot right now because of the teleprompter, but most people won't pay for a teleprompter. That's a viral moment, not a selling point. The selling point is AI, is essential demand, is a pair of glasses you'll still want to wear even when it's out of battery.

Part 02

To B or To C

"Anyong": Was this speech-reader glasses your first AI glasses product?

Misa: First to C one. In 2020 we made a pure B2B monocular AI glasses, but cost was too high for consumer. Still, a lot of tech in our consumer glasses — waveguides, micro-displays — came from experiments in industrial and cultural heritage settings. B-to-C makes sense in an early-stage industry. Apple did education and publishing early on too.

"Anyong": What was the decisive turning point for B-to-C?

Misa: First, AI. AR tech and optical tech kept developing, but AI pulled the final trigger. Second, our experience doing AR glasses guides for museums. Cultural heritage was the crucial bridge from B to C — a lot of people undervalue this space.

"Anyong": What feedback did cultural heritage users give you?

Misa: There's always been a debate about smart glasses: to display or not to display? The reason we stuck with the display route was cultural heritage users' acceptance of head-mounted AR. And users gave specific feedback too — get rid of the cable, prescription lenses, and so on. A lot of companies haven't gone through the B-to-C process, so copying Rayban becomes the easier choice.

"Anyong": What's your current revenue split between B and C?

Misa: C-side revenue scale is definitely higher, but B-side margins are better. We have huge B-side influence. Cultural heritage is about 100 million RMB a year, our ecosystem partners do tens of millions, several million people visit museums annually, and Rokid is the only player in this赛道.

"Anyong": But they might not recognize it as a Rokid product.

Misa: Doesn't matter. Don't think of branding like pushing a thumbtack into users' heads. I'd rather users suddenly realize, "Oh, this is the glasses I wore at the museum."

"Anyong": Among AR companies you have relatively heavy state investment, including Hefei and Wuhan, and lots of local government partnerships — also because of cultural heritage?

Misa: B2B business is the bridge to government engagement. However remote a local government might seem, they always have culture, tourism, and heritage.

Part 03

60 Points or 90 Points

"Anyong": We're now defined as on the eve of the "hundred glasses war," with many new competitors emerging — big tech like Xiaomi and Baidu, "outsiders" like moody and Sharge. Which type worries you?

Misa: It's a land-grab phase right now; competition isn't the main story. Even when big companies enter, they're working in a snail shell — even if they go all-in, their per-unit-area investment ratio is about the same as startups', so not too worried.

"Anyong": As you said before, we're in a phase where "big companies are raising the volume, but product capability hasn't caught up to demand" — a rare time window for you and other startups.

Misa: Right, about two years.

"Anyong": What's most important in these two years?

Misa: Competition has six dimensions: product, innovation, ecosystem — we're already ahead there. Brand, channels, supply chain — big companies have advantages. We've unexpectedly completed the brand part, so lately we're mainly working on channels. Going forward we need to solidify supply chain. This is total war; can't have weak links.

"Anyong": But Xiaomi went from zero to one in car-making in four years. If Lei Jun really pushes on AI glasses, would Xiaomi be a rival to watch?

Misa: He'll succeed. But AI glasses market capacity will exceed phones, because they're more personalized. We just need to stay at the table.

"Anyong": What will Rokid's competitive edge be?

Misa: At core we're a software, operating system company. Jobs spent a lifetime telling people Apple was a software company, but because its hardware was so outstanding, it kept getting overlooked. Why so many Apple fanatics? Hardware differences across phones are already minor; in the end users can't leave Apple's software and ecosystem. The most criticized things about AR glasses — battery life, unstable Bluetooth — have nothing to do with hardware. My first startup, Mammoth Tech, was also an OS company. A company's DNA doesn't change.

"Anyong": If you're doing OS, Google's Android XR would also be a rival.

Misa: I agree — in the future our biggest rival might be Google. But not too worried now; Google is still a second-tier player. The biggest AR glasses market is China. Whoever masters the China market will truly win.

"Anyong": Why China?

Misa: Simple. I'll give you one number: the US has about 400 million people, only 7.9% myopic. China has 1.4 billion, and high school graduates are about 80% myopic.

"Anyong": But many AR brands are making overseas moves now. Why?

Misa: Avoiding domestic competition — strategically correct. Early days aren't for internal卷 [involution/infighting]; everyone fighting to the death on narrow terrain just lets big companies come in and harvest everything later.

"Anyong": Rokid's so hot now — what if many consumers come for the name, then find the glasses experience doesn't meet expectations?

Misa: That's what we worry about most. The comfortable startup rhythm is: ship a 70-point product, then say let's refine together. But Rokid doesn't have that luxury anymore. By June we have to ship an 85-95 point product. If we can't, the traffic will definitely turn on you.

"Anyong": What score was the engineering sample we just tried?

Misa: 60 points.

"Anyong": How do you get to 90 in three months?

Misa: My definition of 90 is: I toss you the glasses, no engineer standing next to you explaining how to operate it. Like when everyone first saw the original iPhone's "slide to unlock" — you just naturally swiped with your finger. Our biggest challenge right now isn't AI or hardware. It's how users learn the AI assistant wake word. The simplest way is subtitles saying "please say 'Leqi'" — but that's too mediocre. A user's first 10 interactions with the product form their core impression. What we need to crack has nothing to do with AI or hardware.

"Anyong": Sounds like a product manager capability.

Misa: It's definitely a product moment now.

Part 04

Idealism or Groundedness

"Anyong": When you decided to start Rokid in 2014, were you inspired by Google Glass?

Misa: At the time I thought I could make something better than Google Glass. I posted on Moments saying I'd self-fund it — looking back, truly didn't know the height of heaven or depth of earth. I put in over 10 million RMB myself, far from enough.

"Anyong": But Google Glass turned out to be a failure.

Misa: Google Glass's problem was being too early. Of course Google itself probably never intended it as a mass product — just let a good idea sprout, whether it grows into a big tree or not, doesn't matter. Vision Pro has a classic problem: it has too many capabilities, like Zhang Wuji's internal martial arts — ends up as a powerful, expensive, heavy, gorgeous monster. It's an interesting contrast: you have to be able to handle your own overflow of capability. Elon Musk did it. Wenfeng Liang did it too — his quantitative trading capability overflowed, which made DeepSeek possible.

"Anyong": Rokid's two best-selling products historically, Lite and Max — you rejected both. Why?

Misa: At the time I only wanted to make what became today's Rokid Glasses, a mass-essential AI glasses product. I didn't want to waste time on AR. But later facts proved the team decision was correct. These weren't my ideal products, but they were important products for a startup.

"Anyong": As CEO of a startup in an early-stage industry, do you ever feel you're too idealistic?

Misa: Maybe this has to do with me barely having had a job. I work from big to small — if I believe AI is the future, I set the goal then work backwards. I observe people who've had jobs: their characteristic is planning the future from existing resources — what can I deliver today, what can I deliver in six months. Innovative companies probably need a CEO like me.

"Anyong": Rokid is 11 years old now. What was the hardest moment?

Misa: Rokid hasn't fought a single fully prepared war in ten years, because the whole industry wasn't mature. But Jobs said: always choose spring technology. If you choose autumn or winter tech, you'll find you can prepare everything thoroughly, but you'll be starved of imagination.

"Anyong": Are you mainly responsible for technology now, or commercialization?

Misa: I'm responsible for product.

Image source: IC photo

Recommended Reading

Where Money Flows, People Rise and Fall