AI will replace the phone, but the winner may not be a phone maker
Zhu Mingming says AI will definitely replace the phone, but glasses are not a winner-take-all business — at most ten players can sit at the table; what he is betting on is the two or three years when the big companies can't see it, don't understand it, don't care about it, and it's too late.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Four thousand yuan in the account, payday in two weeks
At the worst moment of his first startup, Zhu Mingming called everyone together and gave them two choices: keep going, and he would carve out his own shares for them, but he said plainly that they might be worthless paper; or take the computers on the table back as wages, so he wouldn't owe anyone. He said he wouldn't come in that Saturday or Sunday, and wouldn't lock the door, because he couldn't bear to watch people carry the computers away. When he came back on Monday, not one person had left. Less than two weeks later, Alibaba invested, and he never missed a paycheck.
— Zhu MingmingAlibaba OS lost on closedness, not on technology
There are two ways to build an operating system. Alibaba chose the relatively closed one, controlling everything from hardware to software directly and polishing it as a whole, a bit like Apple's approach. In the end, that path proved slower than Android's open route, and Alibaba didn't have that kind of pull at the time. Zhu Mingming calls it a decent start and a not-so-great outcome; the system is now mainly used in cars. By then he had already been reassigned to work on AI.
— Zhu MingmingAI and AR were never two separate paths
Outsiders thought Rocket suddenly switched from AI to AR in 2019, but Zhu Mingming says that's completely wrong: internally, both legs were always running. As early as 2013 he posted that he wanted to build a product like Google Glass, and when he left Alibaba he told Jack Ma that AI and AR would change the world and would become one thing. But at the time an AR product couldn't be productized, so they first shipped AI products to land, giving AR enough time to develop, and then switch when the time was right. The switch came a year earlier than planned.
— Zhu MingmingThe hardest part of the pivot wasn't the direction, it was the people
In the 2019 restructuring, more than half — about 60% — of the staff left within two weeks, and one of the two buildings was basically emptied. Zhu Mingming says the direction decision itself wasn't painful, because he had figured it out long ago; the pain was that many of them he had personally interviewed and hired, they were excellent, and they had done nothing wrong. He set himself two preconditions: first, he had to be able to give N+1 severance, so no one leaving would feel wronged; second, the account had to keep 10 months of cash to validate the new product. So this wasn't a strategic decision, it was a decision calculated from the books, dragged out until the cash-planning deadline.
— Zhu MingmingFive days later, this company wouldn't exist
The restructuring happened in October 2019, and the first product came out in January 2020. The glasses' first dealer conference was on January 15, 2020. Five days later, on January 20, the pandemic broke out, and no suppliers or channels were allowed to move. Precisely because that launch event happened, after the pandemic arrived they found that the glasses, combined with infrared sensing, could measure body temperature non-contact and at scale, and suddenly sold to 80 countries. Zhu Mingming says, if we had delayed just 5 days, Rocket wouldn't exist.
— Zhu MingmingOf Jack Ma's four noes, glasses hit none
Jack Ma summed up four opportunities for startups competing with giants: can't see it, can't understand it, don't care about it, too late. Zhu Mingming says that with AR glasses, from day one big companies will definitely see it and definitely understand it; the only question is whether they care about it now. So the window left for Rocket is to use the time when the giants don't care to get to the point within two to three years where it's too late for them. He admits equal competition is impossible — you can never compete on equal terms with a company like Tencent; being able to compete on par is already good.
— Zhu MingmingHardware's steps can't be crossed by effort alone
Zhu Mingming says hardware growth is never linear or continuous; at a certain point it enters a plateau, and no amount of effort helps, because that's the boundary of the entire industry, nothing to do with your effort. At that stage big companies slowly catch up. So after you finally jump one step, the next jump might be a year away, and that year is the time others use to catch up. Only software, experience, and ecosystem can grow linearly or even exponentially — that's Rocket's real know-how: battery life more than an hour longer than Meta, translation results within 0.5 seconds, all achieved by squeezing power consumption at the OS level, not by application work.
— Zhu MingmingProduction can't keep up because last year was too conservative
Zhu Mingming says this year's biggest difficulty is production — whether the industry can meet capacity is not easy. He reflects that this came from not being bold enough in financing: in the middle of last year he judged the macro environment might not be so good, so he changed the supply chain to build-to-order, adding capacity only as orders came in. It seemed reasonable, but this year when traffic came, they couldn't handle it. If funds were ample, he would keep at least two to three months of inventory, and at worst treat inventory as a cost. He says at Alibaba he could spend several billion a year at most, so his concept of capital is actually more solid.
— Zhu MingmingIn their own words · checked verbatim
But at that time I actually had less than 4000 yuan left in the account, and payday was in two weeks. Think about it — that was the first time in my life I felt despair.
但那个时候其实我账上还有不到4000块钱 但还有两个星期就发工资了 你想想看那个时候 就是我人生中第一次绝望
Zhu Mingming8:07
There's another option: the computers on your desks are just about enough to cover everyone's wages, so take them back, and I won't owe anyone.
还有一种情况就是 你们桌上的电脑 正好差不多可以抵大家的工资了 你们就搬回去 这样我也不欠大家
Zhu Mingming9:07
First, you have to be able to give N+1 severance; you can't let the employees who leave feel wronged. That's the first. Second, you have to ensure this transition — because it's a product direction transition, and a direction transition isn't something you can validate overnight just by turning the corner. You also have to give 10 months for the company to validate this thing.
第一你要能够给出N加1的赔偿 你不能让你离职的员工受委屈 这是第一个 第二要保证这个转型 因为那是一个产品的方向的转型 方向转型不是说 我今天转过去 明天就能够 汤桌子条路就能够验证 你还要给出10个月的时间 让公司去验证这个东西
Zhu Mingming1:00:45
So if we look back at this decision, if I had just delayed 5 days, Rocket wouldn't exist.
所以如果这个决策 我们回头来看 我只要delay5天 对Rocky就不存在
Zhu Mingming1:05:50
The first is called can't see it, the second is can't understand it, the third is don't care about it, the fourth is too late. Many startup opportunities come exactly this way.
第一叫看不见 第二叫看不懂 第三叫看不上 第四叫来不及 很多创业公司的机会都是这么来的
Zhu Mingming1:18:57
Its growth is never linear or continuous; at a certain point it reaches a plateau, and no matter how hard you try it's useless, because that's the boundary of the entire industry, nothing to do with your effort.
它的成长从来都不是线性连续的 它是一定程度到一个平台期 你再努力也没用 因为那个是整个产业的边界就在这 跟你的努力没关系的
Zhu Mingming1:34:02
But hardware is exactly the opposite — hardware really allows everyone to say, I dare to be a follower.
但硬件恰恰相反 硬件是真的可以 所有人都可以说 我感为人后的
Zhu Mingming1:35:02
My requirement for him is that the company's brand and product heat should far exceed my personal heat. I think that's my most ideal definition of success.
我对他的要求 就是公司的品牌和产品的热度 要远超我个人的热度 我觉得这是我对成功最理想的定义
Zhu Mingming2:04:18
Figures
| External investment before Alibaba's acquisition | 1 million USD | 10:08 |
| Zhu Mingming's personal investment in Rocket | 2.3 million USD | 23:14 |
| Zhu Mingming's subsequent additional investment | 100-200 million | 1:43:09 |
| Rocket Air Lite average daily usage | Over 2 hours | 43:29 |
| Number of Rocket developers | Over 10,000, including over 3,000 enterprises | 43:29 |
| Rocket product recommendation index | 60% of sales come from repeat purchases and referrals | 1:20:59 |
| Rocket glasses battery life comparison | Meta about 3-4 hours, Rocket about 4-6 hours | 1:33:01 |
| Myopia rate among Chinese high school graduates | Over 92% | 53:39 |
| Rocket employee count | About 400 | 1:55:13 |
Glossary
- Always on
- A device that is worn all the time and available at any moment — Zhu Mingming sees this as the precondition for an AI carrier to work.
- System Cost
- The steps that must be taken along the GUI operation path before reaching the target function.
- PMF
- The point where a product matches market demand; Zhu Mingming says this road is very long.
- Dare to be a follower
- Waiting for mature supply chains and solutions before following up — Zhu Mingming sees this as the optimal strategy for hardware giants.
How to listen
Founders building smart hardware and AI terminals, investors watching the hardware track, and anyone interested in the competitive window between big companies and startups.
After 1:48:32, the part about Hangzhou entrepreneurs and personal preferences — information density drops noticeably.