BYD has no plans to sell its robots yet

The robot track has been quite hot in the past two days.

On June 1st, Unitree Technology passed the review for listing on the Science and Technology Innovation Board and submitted for registration the next day. Immediately afterwards, Li Ke, the executive vice president of BYD, appeared in an interview and personally admitted that BYD is working on robots. Before her words faded, Estun Automation's stock price hit the daily limit that day.

Then on June 5th, Green Harmonic's stock price hit the 20% daily limit, and the entire sector took off. The Huaxia Robotics ETF rose nearly 4% in one day.

01

Many people asked: What's BYD's purpose in getting involved? The online rumor is that BYD has been working on it for four years, with the code name "Yao, Shun, and Yu". The prototype has reached the seventh generation, and 150 prototypes are running 24/7 in Shenzhen and Changsha.

A reporter from Jiemian News verified this statement with BYD, and BYD replied with four words: All untrue.

The interesting part is here; what was denied were the details, but the direction remained unchanged. Li Ke personally said in the interview that BYD is making robots. In the same Jiemian News report, an insider close to BYD also confirmed this.

BYD's equity investment records in Pacini and Zhiyuan are clearly stated in the industrial and commercial information; the establishment of an embodied intelligence team in 2022 and the recruitment of talents in the field of embodied intelligence from universities in 2024 are all public information.

So, BYD doesn't recognize the specific code name, the iteration generation, or the number of units deployed; but there is no doubt that it is working on this. For listed companies, the data generally doesn't deviate much.

I checked BYD's annual report for 2025:

By the end of last year, the total number of employees was 869,600; compared with 968,900 at the end of 2024, it decreased by nearly 100,000.

The main reduction was in production workers, from 762,900 to 665,800, a decrease of 97,000. The number of technical staff, on the other hand, increased from 122,900 to 127,700, an increase of 4,800.

One area is being cut while the other is being increased. The directions are completely opposite.

Now let's look at the money. The total salary of the whole company increased from 117.1 billion to 130.5 billion, an increase of 13.4 billion; with fewer people, more money was spent. The average annual salary increased from just over 120,000 to 150,000. To put it simply, it's about hiring fewer people at a higher cost, a major reshuffle.

Let's do the math; for 660,000 production workers, with an annual salary of 150,000 per person, the labor cost on the production side alone is close to 100 billion a year.

I also found a set of comparative data. BYD's personnel expense ratio is 15.7%. What does this mean? For every dollar of cars you sell, nearly 16 cents goes to personnel.

What about its peers? Geely's is 5.2%, and Great Wall's is 7.4%.

For the same car sales, BYD spends two to three times more on personnel than others. This is the cost of its business model.

BYD manufactures nearly 70% of the parts for its cars by itself; it makes its own batteries, motors, and designs its own chips. The more work it does, the more people it needs, and the higher the personnel expense ratio.

Vertical integration gives it cost advantages and supply chain security, but it also burdens it with the heaviest labor load in the industry.

So, in the past year, it laid off 100,000 production workers and invested over 150 billion in automation.

The effect of integrated die - casting is quite obvious; for a rear floor, in the past, 74 stamping parts had to be welded together, but now it can be formed in one die - casting. The welding process is gone, so naturally, the welding workers are no longer needed.

The reduction of these 100,000 people is due to traditional automation methods such as integrated die - casting and intelligent warehousing, and has nothing to do with humanoid robots. Currently, the specific number of humanoid robots deployed in BYD's factories has not been disclosed. It is known that they are doing tasks such as handling, labeling, and inspection.

Now comes the key point. Traditional automation can handle the tasks that robotic arms on the assembly line can do; the remaining tasks that require two - hand cooperation, moving around, and adapting to different workstations are what humanoid robots are really meant to do.

My analysis and judgment:

Humanoid robots are the next step on this path. With an annual labor cost of nearly 100 billion and a personnel expense ratio two to three times that of its peers. Traditional automation has completed half of the work, and there is no choice but to take the remaining path.

02

In this track, there is an account that everyone in the circle agrees on.

I looked through several industry reports, and the numbers are quite similar. A humanoid robot that can work in a factory, including procurement, deployment, operation, and maintenance, costs 500,000 to 600,000 in total.

Who is it going to replace? A production line worker with an annual salary of about 100,000. It takes about 5 years to recoup the cost.

It sounds okay, right? Don't be in a hurry. There is another number. The design life of the current mainstream humanoid robots is 3 to 5 years. What does this mean? The robot may break down before it recoups the cost.

No matter who does the math, the conclusion is the same: At this stage, selling humanoid robots is not profitable.

I also looked into more detailed data; for the assembly accuracy of precision components, it needs to be controlled at the 0.01 - millimeter level, and the industry yield rate is generally less than 80%. Some enterprises have disclosed that the batch - production defect rate of the core component, the planetary roller screw, is as high as 12%. The high cost is largely stuck at this point.

This is the current situation in the industry.

BYD's situation is different. It has a high self - research rate for core components. It manufactures RV reducers, servo motors, etc. by itself instead of purchasing them externally. The cost is much lower than buying from outside.

What's more crucial is how to use them. For general robot companies, they need to sell the robots to customers; selling means having a complete set of things: a sales team, customer education, after - sales network, and remote operation and maintenance. For a 500,000 - yuan robot, less than half of the cost may actually go to the hardware.

BYD doesn't need this set. It manufactures and uses the robots by itself, without middlemen or sales expenses, and doesn't need to spend time explaining what the robot can and cannot do.

The iteration speed is also different. What's the normal path? Manufacture the robot, hand it over to the customer. The customer uses it for half a year, finds problems, feeds them back, and then the next version is modified.

BYD's feedback loop is zero - distance. If there is a bug in the robot at the Pingshan factory, the engineers in the next building can see it and fix it on the same day. The fact that the prototype can be iterated from the first generation to the seventh generation is directly related to this feedback speed.

There is also a detail worth mentioning separately.

In April 2025, BYD invested over 100 million yuan to become the second - largest external shareholder of a company called Pacini Perception Technology. Pacini makes 6D Hall array tactile sensors.

I checked the background. Before BYD's technical team made the investment, they reviewed all the mainstream tactile sensing solutions in the world; visual - tactile and capacitive ones, and finally locked in Pacini's technical route.

This investment direction says a lot. It invests in the "hand", not the "brain".

In the factory, for tasks such as handling, labeling, and assembly, the tasks are determined, and the routes are fixed. All the robot needs is to feel what it is holding, whether the force is enough, and whether it is placed correctly.

To put it simply, BYD wants a tool that is sufficient for fixed workstations. I think this idea reduces both the technical threshold and the cost.

People in the industry often say that the commercialization of humanoid robots still needs to wait for 5 years. That's for the external sales account; BYD is calculating another problem, which is to make a robot cheaper than a worker at its own workstations.

03

Additionally, there is a set of numbers worth looking at separately. I checked how many humanoid robots were shipped globally in 2025?

According to a report from Omdia, the total global shipments exceeded 13,000 units. Zhiyuan Robotics alone accounted for 5,168 units, nearly 40%; Unitree Technology shipped 4,200 units, more than quadrupling compared to the previous year.

The data from the GGII (Gongkong Industry Research Institute) is even higher. In China alone, there were 18,000 units, an increase of more than 650% compared to 2024. Morgan Stanley's figure is 14,000 units, only counting China.

The three numbers have different scopes. Some count the whole machines, and some may include quadruped robots. However, the following comparison holds true regardless of which number is used.

There is a number circulating online: BYD's internal deployment target for 2026 is 20,000 units. If this number is true, the quantity that one company needs for its own use is more than the total global sales last year.

BYD has denied this number. Let's put the specific number aside for now. Look at this from another perspective.

BYD has confirmed that it is making humanoid robots and using them in its own factories. It has 660,000 production workers and is one of the car companies with the largest number of production employees globally.

Even if it deploys only a few thousand units in the first year, this quantity is already a very large single - customer order in an industry where the global annual shipments have just exceeded 10,000 units.

The key lies in how many units it can absorb. The factory system with 660,000 production workers provides not only orders for humanoid robots but also a large enough test field and a long enough iteration runway.

There is a consensus in the industry that for the humanoid robot supply chain to really take off, the annual production capacity needs to exceed 100,000 units so that the component costs can be spread out by scale.

How many units can the whole industry achieve in 2026? GGII's estimate is 62,500 units, and Morgan Stanley is more conservative, at 28,000 units. There is still a long way to go to reach 100,000 units.

Most robot companies face similar difficulties. After manufacturing the robots, they have to find customers one by one to sell them. They sell ten units to one company and twenty units to another, gradually accumulating. No one knows when the supply chain will be fully utilized.

BYD doesn't need to accumulate. It is its own big customer. Once the internal demand is released, the orders for the supply chain are guaranteed.

This matter is even more interesting when viewed in the context of 2026. This year is called the "Year of Mass Production" by many people. I checked the progress of several companies. Tesla's Optimus V3 will start a dedicated production line at the Fremont factory by the end of this year. The plan is to produce 50,000 to 100,000 units in 2026 and start selling them externally in 2027.

Elon Musk mentioned in the earnings conference that China is Optimus' biggest competitor.

XPeng's IRON will be unveiled in the third quarter and mass - produced by the end of the year; He Xiaopeng's plan is to first deploy them in XPeng's own stores. More than 500 units are already running in the factory, and they will be delivered to commercial customers early next year.

Ubtech is even more direct. It started the pre - sale of consumer - grade humanoid robots on JD.com in early June.

When comparing the paths of these three companies, the differences are clear; Tesla promotes the robots externally as soon as they are made, relying on external customers to scale up; XPeng uses them in its own stores first and then sells them externally after everything runs smoothly. Ubtech is targeting consumer customers.

BYD has taken neither of these paths.

I checked, and as of June this year, it has not released any robot models for consumers.

This creates an interesting contrast. While others are trying to sell the robots, BYD is still trying to make the robots do more work in its own factory.

It may seem that BYD is slower than others, but from another perspective, it is safer.

Tesla and XPeng need external customers to be willing to pay for the scale to increase; BYD doesn't need to persuade anyone. Whether it deploys 2,000 units or 20,000 units this year, its own production line is the largest internal test field in the industry.

I think what really deserves attention is who can survive the vacuum period before commercialization; while the industry is still waiting for the market to mature, BYD has skipped this waiting period with its own scale.

04

It's not the first time that BYD has worked on something quietly.

In 1995, Wang Chuanfu registered BYD in Shenzhen. It made nickel - cadmium batteries and did OEM work for mobile phones.

At that time, Japan's Sanyo monopolized the market. An automated production line cost 20 million yuan, and BYD couldn't afford it. Wang Chuanfu split the production process into individual manual workstations and used human labor to replace equipment, managing to reduce the cost to less than half of that of Japanese manufacturers.

While making batteries, he found a problem; when selling batteries to others, the fate was in the hands of the customers. If the customers cut the orders, you would be in trouble.

In 2003, BYD spent 270 million yuan to acquire Qinchuan Automobile; the entire industry thought he was crazy. Why would a battery factory make cars? The stock price dropped on that day.

What he wanted was to find a way out for himself. First, use the batteries in his own cars. After the scale and cost were brought down, Fudi Group became independent in 2020.

Fudi Battery changed from an internal division of BYD to a company that can supply batteries to external car companies; now, Fudi supplying batteries to external car companies has become an independent business.

The same thing happened with chips.

BYD Semiconductor was also an internal department at first, making IGBT chips for its own cars. When its capabilities exceeded the internal needs, it was spun off in 2020 to serve external customers. In 2022, it passed the first - round listing review but then postponed it. However, the framework for independent operation has been established.

Looking back at the matter of robots, the timeline is clear.

Xia Zuoquan, a co - founder of BYD, started investing in Ubtech in 2013 and made several rounds of investments. This shows that the BYD system has been paying attention to robots much earlier than the outside world thought.

In 2022, BYD established an embodied intelligence research team, which was officially disclosed by BYD itself; in 2023, it