The battle for talent among robotics companies has reached what one industry insider called a “manic” stage. Several people familiar with the matter said a Peking University associate professor was approached with “an astronomical offer of 10 million to 20 million yuan.”

Some Tsinghua students are being “reserved” by companies for 3 million yuan as soon as they choose their major, on the condition that they work for the company for 10 years after graduation.

Capital, meanwhile, has created a distorted “headcount valuation formula”: a professor from Tsinghua, Peking University, Harbin Institute of Technology, Shanghai Jiao Tong University, Beihang University or another top robotics school can add 200 million to 500 million yuan to a company’s valuation;

a core executive from UBTECH, Unitree or AgiBot can add 50 million yuan; and a PhD from a prestigious university can add 10 million yuan.

This “valuation game,” defined by capital, is dragging robotics into a death spiral of cash burn without real deployment...

The Talent War

Cheng Yue, co-founder of a robotics company, noticed that in 2025 everything began showing signs of spinning out of control.

Starting salaries for some undergraduate algorithm roles have already been bid up to 400,000 yuan a year, while master’s graduates start at 600,000 yuan and PhDs need 1 million yuan a year before talks can even begin.

For context, data from MyCOS’s 2025 employment blue book shows the average annual salary for new graduates is only 74,000 yuan.

In other words, salaries in robotics are more than five times higher than those of ordinary college graduates.

A Manufacturing Talent Development Planning Guide released by China’s Ministry of Industry and Information Technology and Ministry of Education predicts that in the niche but cutting-edge field of humanoid robots, conservative estimates put the shortage of core R&D talent at about 300,000 people.

That shortage of 300,000 is not evenly distributed. It looks more like an inverted pyramid.

The closer you get to core technologies, the more acute the shortage becomes. Top industry talent, including PhDs, professors and experts, is where companies are fighting hardest.

Several people familiar with the matter said a Peking University associate professor was recently “poached,” with the leading robotics company offering “an astronomical annual salary of 10 million to 20 million yuan.”

The talent war has become so intense that some companies have had to find other ways in.

Several people familiar with the matter said many outstanding students in relevant majors received signing offers from a leading company shortly after entering Tsinghua University: “a 3 million yuan scholarship, but after graduation they have to join the company and commit for 10 years.”

Cheng Yue’s company was founded in 2024 and has only raised a Series B round, leaving it without enough money to compete for talent.

With no real alternative, he recruited some students from ordinary second-tier undergraduate colleges and junior colleges in 2025.

“But within three to six months, other companies poached them with roughly double the salary,” Cheng said.

They had to double salaries, only to see employees poached again a few months later by rivals offering even more.

As offers rose across the market, salaries for core roles in robotics roughly tripled in 2025.

“Noncompete agreements have completely stopped working in this circle,” Cheng said.

Worried that core technology would leave with departing employees, he rolled out blockchain-based storage technology at the company, with access to materials tiered by job level.

“We can see clearly in the backend who downloaded which files.”

But that tactic only keeps honest people honest. It cannot stop bad actors. If someone simply photographs the screen, what can you do?

This red-hot sector is becoming surreal and distorted...

Who Is Being Poached?

At present, the most critical and hardest-to-hire roles in the industry are robotics motion-control algorithm specialists.

“What is most lacking now is algorithm talent that can make robots walk steadily, keep their hands from shaking and handle complex movements like humans,” Cheng said. There is still no fully aligned major, he added. “Some universities have launched related programs, but their students will not graduate for several years.”

The closest-fit disciplines are control science and engineering, robotics engineering, multibody dynamics and related fields.

Data from platforms such as Zhaopin and Liepin show that in 2025, job postings for “motion-control algorithm roles” in humanoid robotics grew by more than 400%, with supply and demand badly out of balance.

The second tier is embodied AI algorithms, which connect the virtual and physical worlds.

This type of talent is mainly responsible for applying virtual simulation data to real-world control.

Computer science and technology, artificial intelligence, especially in imitation learning, and other majors are also targets in the hiring scramble.

Within the industry, UBTECH is known as the robotics sector’s “Whampoa Military Academy.”

As the “first humanoid robot stock,” UBTECH has trained a large pool of core technical talent.

Sara, a headhunter focused on robotics, said: “For a startup, poaching one core person from UBTECH can lift its valuation by 40%. If it poaches a core team, the valuation can double outright.”

UBTECH has two innovation brains: one in Beijing, known as the “Northern Research Institute,” and one in Shenzhen, known as the “Shenzhen Research Institute.”

Both are among China’s top early technical teams in humanoid robotics.

Sara once rented offices near both sites to “camp out and poach people.”

According to publicly available data, UBTECH has lost nearly 50 core backbone employees since 2021, with total departures reaching 300 to 600 people. (List of core personnel who left UBTECH) Before the old Whampoa Military Academy has even been picked clean, new targets have already become prized prey for headhunters.

Sara is preparing to set up a branch office in Hangzhou, with the main goal of poaching employees from Unitree and Deep Robotics.

In this hunt, no company is safe.

The Weight of Labor Costs

Many people assume the biggest cost in robotics is manufacturing the robots.

In fact, the biggest cost is labor, Cheng said. To secure a valuation of 1 billion to 2 billion yuan in the primary market, a team has to clear a hard threshold: 50 to 100 people.

But among those 100 people, they cannot all be low-paid interns. There must be a high share of PhDs and graduates from top universities, or investors will think the company lacks enough technical depth to justify a high valuation.

A distorted cash-burning model has begun to emerge.

It is easy to calculate how fast a robotics company burns money. A star professor costs more than 10 million yuan, dozens of PhDs earn million-yuan annual salaries, and dozens more engineers earn 500,000 yuan a year. A standard robotics startup that has not even begun selling products can burn through 100 million yuan a year on labor costs alone.

If a robotics company needs 100 million yuan a year just to operate, then with robot production costs and office expenses added, the total can reach “as much as 150 million yuan.”

UBTECH’s financial reports offer a glimpse into the industry’s reality.

In 2024, UBTECH had 2,191 full-time employees and total compensation costs of 941 million yuan. Full-year revenue was 1.305 billion yuan, meaning labor costs amounted to 72.1% of revenue.

Another noteworthy detail: UBTECH’s average annual compensation per employee was about 430,000 yuan. The average in China’s tech industry is only 200,000 to 300,000 yuan a year, underscoring just how high average pay is in robotics.

Behind all of this is the push from capital.

Sara said investors now choose projects mainly by whether the story “sounds compelling.”

But robots are not actually very useful yet. In factories, they are worse than robotic arms; coffee-making robots are worse than coffee machines.

“Sorry, but everyone in the industry is still hand-building things. There is no real use case to speak of,” Cheng said. To support the capital-market story, companies can only stage “performances” that will not go wrong, such as dancing or martial arts routines.

Getting onto the Spring Festival Gala is an even bigger boost.

That is probably why robotics companies are willing to spend 100 million yuan just to fight for a spot on the Spring Festival Gala.

To tell a high-valuation story well, companies have to maintain an “all-star” R&D team and “Hollywood-style” demo effects.

Poaching talent has become inevitable.

Several investors said they use a basic internal “headcount valuation formula.” For example, the top robotics schools in the industry right now include Tsinghua, Peking University, Harbin Institute of Technology, Shanghai Jiao Tong University and Beihang University.

If a professor from one of these schools joins the team, the company’s valuation can rise by at least 200 million to 500 million yuan. If a senior executive from a leading company such as UBTECH, Unitree or AgiBot joins, valuation can increase by 50 million yuan. A PhD from a prestigious university is worth 10 million yuan in valuation. But this is a full-blown death spiral.

We visited more than 20 robotics companies, and dozens of practitioners told us they all fear the same point in time: 2026.

This year, they believe, the primary-market bubble in the industry will burst. Ninety percent of companies will fail to raise money, and annual labor costs in the hundreds of millions will drain them to death.

As a result, every founder has been pulled into a gamble that amounts to drinking poison to quench thirst.

“If you do not poach talent, the team’s credentials look weak, you cannot raise money, and the company dies immediately,” Cheng said. Bringing in sky-high-priced talent and performing for capital might help secure funding.

But if a company burns money wildly and still cannot produce a real product, the story will no longer hold, and the end result is still death.

No matter how compelling the story, it ultimately has to face the judgment of the financial statements.

Talent is a moat, but also a shackle; valuation is a halo, but also a cliff; capital is fuel, but also poison.

UBTECH founder Zhou Jian has also admitted that over more than a decade of entrepreneurship, he thought countless times about giving up.

The sadness of being carried along by capital and unable to dismount may be something only those inside the game can truly understand.