In Honor's view, there are no barriers for robots.

At the south gate of Nanhaizi Park in Yizhuang, Beijing, the 2026 Yizhuang Half Marathon started with the sound of a gunshot. There were more than 100 robot teams and over 300 robots in total, almost five times more than last year.

At 8:18 a.m., the first robot crossed the finish line. It was called "Lightning" and was from Honor's "Jueying Chitu Team". It used the remote - control mode and took a net time of 48 minutes and 19 seconds, but it wasn't the champion.

The champion was another "Lightning", the autonomous navigation version, which finished in 50 minutes and 26 seconds. It was from Honor's "Qitian Dasheng Team".

01

What does 50 minutes and 26 seconds mean?

I checked. The world record for the men's half - marathon by a human is 57 minutes and 20 seconds, which was just set by a Ugandan athlete on March 8th this year. "Lightning" was about 7 minutes faster than that.

The runner - up and the third - place finisher were also called "Lightning". The top three were all from Honor. Later, Honor said that the top six were also from it.

The previous champion, Tiangong Ultra, had a system disorder and fell into the grass after crossing the finish line this year. The veteran H1 from Unitree also fell after finishing the race. Last year, it took 2 hours, 40 minutes and 42 seconds to complete the race. This year, Honor's "Lightning" only took 50 minutes and 26 seconds. This mobile phone company, which participated in the race for the first time, cut the record by two - thirds in one year.

There were two things done right on this robot, one obvious and one hidden.

The obvious one was the legs.

"Lightning" is 169 cm tall with an effective leg length of 0.95 meters. It sounds like the legs are made long, and there is an engineering detail here. According to Honor's public statement, the 0.95 - meter length is achieved by moving the rotation center upwards.

What does it mean?

The starting point of the thigh is raised upwards. The rotation position of the hip joint is raised inside the body, so the part of the leg that can swing out becomes longer. The result when running is a larger stride, and with the same stride frequency, a faster speed can be achieved.

Engineer Yao from Honor said in a post - race interview that this idea was inspired by the body structure of long - distance runners.

By comparison, Tiangong Ultra is 180 cm tall, 11 cm taller than "Lightning", but the official has not disclosed its effective leg length. The veteran H1 sent by Unitree is relatively short. "Lightning" is not the tallest, but it has the largest leg - to - body ratio.

Combined with a self - developed integrated joint module with a peak torque of 400 N·m, the peak speed of "Lightning" in the 100 - meter run reaches 6 m/s. The running speed of the remote - control version throughout the race exceeds 7 m/s.

The legs only solve the problem of running fast. Running fast means generating more heat. The next difficult thing is what really makes the difference, because the hidden heat problem is one that the entire industry has not solved well.

I looked through the reports of last year's half - marathon. The scene was quite vivid. Many robots "had heatstroke and stopped working" halfway through the race. The staff had to run alongside and spray coolant on the robot joints.

When a humanoid robot runs at high speed, the temperature of the lower - limb joint motors can reach 120°C. Beyond this temperature, the motor torque will decrease, the positioning accuracy will decline, there is a risk of demagnetization of the permanent magnets, and the material aging will accelerate.

To protect the motors, the system usually automatically reduces the power. Most of the time, the robot runs out of energy, slows down, or even stops because of this.

"Lightning"'s solution is liquid cooling.

According to Honor, the liquid - cooling pipes penetrate deep into the motors like capillaries. The high - power liquid pump has a heat - exchange flow rate of more than 4 liters per minute. Each of the four drive motors in the lower limbs is equipped with an independent liquid - cooling circuit.

There is a detail worth mentioning. The core liquid - cooling pump of "Lightning" comes from a supplier called Huake Lengxin. Its HD01 suspension pump has a rotation speed of more than 20,000 revolutions per minute, a flow rate of 6 liters per minute, a volume of about 30 millimeters, and a weight of less than 100 grams.

So, Honor's "self - developed liquid - cooling system" should, more accurately, be a joint development where "Honor sets the plan and Huake Lengxin provides the core components".

To be honest, liquid cooling is one of the two paths in the humanoid robot industry.

Unitree uses air cooling. Its patented solution is a multi - stage heat - dissipation duct combined with two fans working in tandem. One fan covers the thigh motor and the drive board, and the other is specifically responsible for the calf motor. Unitree has said that air cooling has no risk of liquid leakage and is more suitable for lightweight design.

ZhiYuan and CloudMinds follow the same liquid - cooling path as Honor, and they started much earlier. ZhiYuan's PowerFlow joint module has long integrated an integrated liquid - cooling circuit, and CloudMinds' Expedition A1 also uses liquid cooling.

Tesla's Optimus is still under evaluation. A Japanese research team has even developed a "motor sweat gland" that allows water to seep out of the porous structure and evaporate for heat dissipation.

So, Honor is not the first to use liquid cooling, but it is the first to make liquid cooling work in a real - world race. It may sound like a small difference, but in fact, it is the difference between mass production and a sample.

02

For a robot team that has only been established for one year, where did all these things come from? The answer lies in the supplier list of "Lightning".

In addition to Huake Lengxin, Lingyi Zhizao supplied 159 types of core metal structural parts, covering the hips, legs, feet, and torso.

Lanshi Technology supplied 132 types, involving the head, arms, hips, and legs. AAC Technologies not only made the key structural parts of "Lightning"'s body but also was responsible for the head and legs of "Yuanqizai".

I specifically checked the backgrounds of these companies. Lingyi has been making iPhone structural parts for more than a decade. Lanshi initially made glass covers for Apple, and AAC is a leading company in the mobile phone acoustic module field.

Honor didn't look for specialized robot suppliers. It just took out its original supply - chain list, selected dozens of partners, and made this robot. This is quite abstract.

Based on my understanding, I've sorted out that there are at least five capabilities that a mobile phone factory can directly transfer to robots.

The first is thermal management. Mobile phone factories have accumulated more than a decade of experience in stuffing SoCs into a one - square - centimeter space and doing heat dissipation. The processes, materials, and simulation capabilities of the liquid - cooling solutions are all ready - made.

The second is batteries and BMS. The teams in mobile phone factories are probably the people who understand charging and discharging curves, temperature protection, and battery - life management best in the world.

The third is SoC and visual computing power. Mobile phone SoCs are originally designed to handle tasks such as multi - camera, SLAM, and night - scene noise reduction simultaneously, and these capabilities are exactly what robots need for autonomous navigation.

The fourth is supply - chain negotiation. For companies like Lingyi, Lanshi, and AAC, Honor can just make a phone call to invite them to cooperate. For a startup, making ten phone calls might not even get them on the waiting list.

The fifth is mass - production technology. Mobile phone factories can control the yield rate, production rhythm, and cost when producing hundreds of thousands of mobile phones a day. Startups like Unitree, Songyan, and Tiangong are still slowly exploring these aspects.

These five capabilities together form a very strong hand. Personally, I think that in the future, when we see a mobile phone factory announcing that it's getting into the robot business, we can use this list as a reference. If a company has more than three of these capabilities, it has a systematic advantage; if it only has one or two, it's just following the trend.

03

However, this list also has limitations. The things outside the list, such as motors, reducers, force - control algorithms, and simulation training, are the real foundation of a robot company.

Did Honor do a good job in these key areas in the past year?

I checked the timeline of Honor's robot business. In April 2025, Honor established a new - industry incubation department, which also set up secondary departments such as the Embodied Intelligence Laboratory and the Bionic Ontology Research Laboratory.

In May, at the launch event of the Honor 400 series, they first announced their intention to enter the robot field. By April 2026, when they won the race, it had only been 12 months at most.

What does it mean to build a robot company in one year in this industry? I compared it with several other companies.

Unitree was founded in 2016. It took until 2023 to release its first humanoid robot, H1, after spending a full 7 years specifically accumulating technologies in motors, reducers, and motion control.

The Beijing Humanoid Robot Innovation Center behind Tiangong was founded in November 2023. It only took 6 months to release the first version of "Tiangong".

But it should be noted that this is an institution jointly established by Ubtech, Xiaomi, Jingcheng Jidian, and Yizhuang Robot, which is equivalent to directly using the existing technological accumulations of the four companies.

Songyan Power was founded in September 2023. The average age of its team is less than 30 years old. It took two and a half years to get 9 rounds of financing, perform on the Spring Festival Gala, and win the second place in the half - marathon. This is already recognized as the "ultra - fast speed" in the industry.

Generally, it takes at least two to three years to build a humanoid robot company in the industry, but Honor only took one year.

How exactly this one - year period was spent is not fully clear from the information publicly released by Honor.

The parts they disclosed are the integrated joint module, high - dynamic motion - control algorithm, and multi - sensor fusion technology, all of which are achievements. The parts not disclosed are: the motor model, the reducer supplier, the simulation - training duration, and the reinforcement - learning framework.

Do companies in the industry usually disclose these things? I don't know.

ZhiYuan believes that a reasonable explanation is probably this: The core motion system of "Lightning" is more like a joint development where Honor sets the plan and then finds suppliers in the industry to provide the core components. The liquid - cooling pump from Huake Lengxin is an example. It's very likely that the motors and reducers also follow this approach.

This is not a criticism. Under the limitation of only one year, this is the only way to make it happen.

Unitree, Tiangong, and ZhiYuan follow the heavy - asset route of self - developing from the motor. The foundation they've built takes a startup ten years to accumulate.

Honor follows the integration route. It selects the already - mature core components in the industry and then uses the system - integration capabilities of a mobile phone factory to assemble these components into a robot.

There is no such thing as one route being better than the other. They have completely different characteristics. One is to build a robot company steadily, and the other is to solve robot problems with the thinking of a mobile phone company.

04

I'm wondering, if Honor doesn't want to be a robot company, what's its purpose?

The answer lies in another robot called "Yuanqizai". It is 136.9 cm tall, silver - colored, and has a very amiable design. It won the Best Gait Award in this half - marathon and focuses on human - robot interaction capabilities.

Honor itself has said that "Yuanqizai" will be used in the smart retail of Honor stores in the future to attract customers and act as a shopping guide.

In addition to these two humanoid robots, Honor is also planning a four - legged robot dog called "Anzai" and a dexterous hand called "Luban". Looking at this product matrix, the position of "Lightning" becomes clear. It is the flagship showcase, while "Yuanqizai" is for mass production.

So, Honor is probably building a brand for "Yuanqizai".

Using the title of the half - marathon champion to deeply imprint the words "Honor Robot" in consumers' minds, and then letting "Yuanqizai" work in its own stores, this is its whole logic. But just having this logic is not enough. The most interesting thing is that only Honor can make this path work.

I've sorted out at least four structural conditions.

The first is the self - owned offline channels. I specifically checked, and Honor currently has about 5,000 experience stores and exclusive stores in China. Putting one "Yuanqizai" in each store would result in 5,000 orders right away. This is a ready - made demand.

Unitree, Songyan, and ZhiYuan don't have such self - owned channels. They either enter factories, research institutions, or exhibitions. They have to negotiate for each order, which is very labor - intensive.

The second is that Honor doesn't need to make money from selling its robots. For startups, the selling price of each robot has to be higher than the cost. This is the basic principle of doing business.

Honor is different. When "Yuanqizai" is placed in Honor stores, it is equivalent to a fixed asset of the store, just like the display cabinets and the lights in the store. The cost can be amortized slowly.

Its value lies in helping to sell more mobile phones in the store every day. The way of calculating the accounts is completely different.

The third is that the brand budget can be directly transferred for use. Honor's annual brand - marketing budget for its mobile phone business is in the hundreds of millions. Allocating a part of it to robot R & D doesn't put any pressure on the group's finances.

Startups can't do this because every cent they spend is from financing, and they have to account to the investors for the money spent in the market.

The last one is user feedback. When a robot is placed in a store, hundreds of consumers come in and out every day, greet it, ask it questions, and take pictures with it. These interactions are the most real data.

It's equivalent to "Yuanqizai" doing user testing every day, and this testing scenario is closer to the real consumption scenario than any enterprise tour provided by B - end customers. Startups like Unitree and Songyan can't get such high - density feedback.

These four conditions together form an extremely wide moat that startups simply can't cross.

It's also worth taking a look at the paths of other companies.

Unitree aims to cash out through an IPO. It sold more than 5,500 humanoid robots in 2025. In March 2026, it disclosed its IPO application materials, planning to raise 4.2 billion yuan, and its valuation has reached 12.7 billion yuan.

Songyan Power relies on financing and emotional value. It got 9 rounds of financing in one year, with a total of more than 1.5 billion yuan. The "Xiaobumi" it launched at 9,998 yuan sold out the first batch of 500 units in two days. Tiangong relies on a state - owned platform. In early 2026, it completed the first round of market - based financing of 700 million yuan, following the path of a parent platform plus an open - source community.

These three companies have three different paths, all of which are good, but none of them is Honor's path. The core difference is that they all need to get money from the outside, while Honor doesn't. It has its own stores.

Finally, here's a little detail for you to think about:

The robot companies that appeared on the 2026 Spring Festival Gala were Unitree, Songyan, ZhiYuan, and Galaxy Universal. Tiangong and Honor were not there.

Actually, the reason is very simple. The Spring Festival Gala is for the market to gain exposure; the marathon is for the industry to showcase engineering capabilities; the stores are for Honor itself to practice real - world implementation.