“ What is the result of high wages?
Are the salaries of talent in the robotics industry a bit too high?
In the past few days, the topics of the 100,000 tonne monthly salary of the Ph.D. students have surged in popularity, making people feel the vibrancy of the industry. But the robotics conference also heard different voices when interviewing some of the top companies.
01.
Distance between high pay and AI
One interviewer cited an industry-shaking incident where a publicly-listed robotics company in April publicly recruited a physically intelligent chief scientist with an annual salary of $15 million, up to $124 million, consisting of three parts: cash, benefits and equity.
He transformed the tunnels into robotic units, with most people's annual salaries already exceeding the sales price of a full-size humanoid robot, and ultimately the ROI of both robots and humans is uncertain.
Another algorithm we've heard in interviews is to compare human output to AI models.
With eight A100s and a few 4090's and 5080's of GPUs, the entire computing resource is invested in millions of dollars, and it takes a year to reinforce a base model, and it costs about $5,000 a month to call on a vertical model to optimize such a task. For example, a decade ago, hundreds of thousands of doctors in the AI industry invested billions of dollars in image recognition.
This means that, to some extent, there is no shortage of talent. Ordinary undergraduates, even interns with the strongest AI, can also reach their abilities. This leads to some people in the past who may have been considered second-class talent, in some unknown companies, made demos or real-world experiences worse or even better than first-class companies, the bubble is broken.
02.
Start-up recruitment is setting the limit price for an industry
The Hunting for Big Data Institute's Hunting 2026 Robot Talent Supply and Demand Trends Insights Report shows that from January to July 2026, the number of new jobs in the robotics field grew by 60.17%, including 147.02% in the human-robot segment.
The Pulse High-Employment Talent Think tank's Autumn 2026 Recruitment Workplace Insights Report also shows that from January to July 2026, the number of newly hired people with a smart body grew six to seven times as much, with the number of newly hired jobs in some humanoid robotic enterprises growing as much as 831.88 percent.
Figures from a Beijing University of Technology survey show a more direct gap of up to 1 million people with physical intelligence. The same study also notes a smaller proportion, with demand for high-performance computing engineers dropping from 0.99 in 2025 to 0.26, even higher than the demand for physical intelligence research and development stations of 1:10.
So how much is a person worth? In this bubble of all robots with physical intelligence, this supposedly the simplest question becomes a difficult question to quantify.
Because the price does not come from the demand, the demand just pushes the price to the ceiling.
According to the vice president of one of the world's leading robotics companies, the cost of hiring a physically intelligent graduate is already very high, a doctor needs at least 1.5 million annual salaries, and good people need to pay 3 million to 5 million annual salaries.
During the spring semester, the demand for AI-related positions increased by 15 times, with an average monthly salary of 62,000 yuan. During the fall semester, an offer of 500,000 yuan per year for an M.A. and Ph.D. in the field of AI-related positions has been laughed at as the price of green tea.
In order to lock up the pupils in advance, some start-ups have introduced a policy of distributing pupils on a full-time basis during their internship, directly giving up to 100,000 yuan per month to the trainees.
A hunter-gatherer refused to be interviewed, and she gave the shortest and most accurate description to the media, saying that she had recently been busy digging people at school.
Because universities are moving faster than businesses. In 2026, the Ministry of Education for the first time added the specialty of embroidery and intelligent embroidery to the undergraduate specialty directory, and supported the addition of related specialties to nine universities such as Harbin Industrial University, Beijing Aerospace University and others.
However, these nine schools are not large in size. Six of the nine schools only admit 30 people, the most of the Beijing Technical University admit 120 people, a total of about 360 people a year. And most universities are focused on the training direction to continue to deepen, but the gap in the industry is one million, and the first undergraduates should not enter the market until 2030 at the earliest, so the half-time difference is calculated by decade.
As the time to develop talent is far from meeting the needs of today's smart companies. From 2025 to 2026, smart companies are not eligible for recruitment, only to go to the smart driving company.
A senior intelligent driving product manager recalled that in those two years, he received more than 40 company invitations every week. One company's recruitment manager estimated that intelligent driving data people are adaptable, less expensive to handle and cheaper than big model companies.
These people are being digested. According to statistics, nearly 40 core executives and technical backgrounds in the domestic field of intelligent driving have crossed the border into the field of physical intelligence, involving more than 20 startups, more than 70 percent of which were founded in the last two years.
Within two years, a profession that was not on the talent market list was listed and a 15 million jump in recruitment started to push the industry price cap to seven digits. This is hard to imagine in the original robot market.
03.
Who is paying for this price?
These figures are correct, and the problem is that putting them back into the company's accounts and investing in the proportion of output does not look very good.
It means that the cost structure of robots is shifting from material to human beings, costs outside of robots are shifting to human beings, and human cost and output, an industry that does not yet have a set of external settlements.
The company, mentioned in the introduction, paid top scientists, although ultimately not open, was also affected by certain research and development costs.
Some media noted that the company spent 303 million yuan on R&D in the first half of the year, with 1,103 R&D personnel. If you put the R&D costs on a person's head, it is about 275,000 yuan per person in the first half of the year, 549,000 yuan per year, and about 600 million yuan per year after the year spent on R&D, most of which are people. This does not take into account the cost of recruitment itself, nor does it take into account the overall human cost of R&D personnel.
This money could eventually be borne by the customer's purchase price, which customers are less willing to accept now.
While an industry is giving robots units of measurement, a given set of formulas cannot be used for settlement.
So why are people still willing to give exaggerated high wages? Some people think it's because they're robbing people with foreign companies.
But in interviews, more investors revealed that within the smart industry, there is actually a rumor circulating about a formula for the valuation of smartphones. The industry is rumored that a publicly traded home appliance company hired a research and development team for a bankrupt robot in Shanghai to use this valuation method.
He also said that the company's valuation could rise as long as it was hired by academic experts, key executives from well-known companies, or top professors from universities as consultants or team members.
In human valuation equations there is no output, only the addition of a value.
One hunter said that when you get out of the circle of companies that do better, even if you don't have a good degree, but you've been involved in the 0 to 1 process early on, you're willing to pay.
When the price of a job is talked about how much valuation pump he can bring instead of how much effective work pump he can produce, the job has gone from cost to narrative.
But the cost of telling a story is always a price to pay.
There is another group of people who have endured this pressure to treat highly talented people: high school instructors.
A doctoral student who has not yet graduated can get up to $100,000 a month for his internship if the direction of his research is right. The backdrop of this statement is that others in the same laboratory, as well as engineers from other companies with similar qualifications, have their offers reinvented by the $100,000.
When the annual salary of $500,000 is called the "white price" by the circle, it becomes $500,000. It's hard to turn around and it's hard to keep people in the lab once you open your mouth.
One investor's judgment on the industry's salary is that it's too easy to make money and it's easy to overlook the most important basic functions of life.
The supply end is completely lacking in elasticity and is one of the most embarrassing parts of the entire pricing system. Job demand grows six to seven times a year, the gap is one million, and the industry can really take out almost a fixed pool of people.
When all the companies go to get the same batch of people in the same pool, there is a gap, but the actual output of the industry is zero, and the only output is price.
This is the place where human inflation and other inflation are the most similar, the molecules have collapsed, the fractions have not moved. The other bubbles are more money changed, money changed hair, human inflation is less money, the price itself goes up.
Another layer of cost is calculated where it is not visible. When the recruitment competition is fierce, there are companies in the price package for a PhD, a large part of which is equity and benefits, which are not included in the current profit or loss table, but will become diluted in future financing, and more complex allocation when withdrawing.
The money came out of the pockets of the investors, passed through the company's cash accounts, and ended up on the payroll, not on the market orders.
But the salary table is the most expensive one, because once it is signed, it is deducted monthly, unbalanced by volume of production, unbalanced by order, unbalanced by refund.
04.
Putting people's prices into the company's accounts
The normal way for an industry to price a scarce product is to first tell the person what they can produce in a year, then divide the product into scarce quantities, and finally get the price.
On October 29, a leading AI company released a measure of the value of physical AI (Worken), a formula for calculating the number of active robots multiplied by the amount of time they worked autonomously and then by the production efficiency ratio. The official definition of the production efficiency ratio is that the qualified quality created in unit time is compared to skilled industrial workers in the same position.
In other words, a robot will eventually be compared to a human and a skilled worker.
So, are humans really better than robots, or the new era of AI?
But at the very least, how much a person is worth is ultimately determined by how much calculable output he has left over the year, rather than how much the valuation will decrease when he leaves.
This set of human valuation statements should not be written in any public financing terms or book of shares. Written in, it shows that it has become a pricing mechanism from rumors and actually generated a bubble.
05.
Link to the Future
Workers are not opposed to high wages, because there is always a way.
A developer who has transformed the automotive industry has said directly: "You have to make a quick buck before you go back to the car".
Some experts have spoken in public about a number that can actually be used to evaluate smart people. He said that in 20,000 hours of data, less than 3,000 hours could be really effective.
This 15% ratio is not only applicable to data, it should also apply to people, and currently only a very small number of people have taken this link seriously.
Now that the R&D conversion link has been skipped, the scarcity level is directly linked to the price, and there is no intermediate amount in the middle that can be charged, nor any direction that can be switched.
Money flows in and out, jobs fall in and out, and people who can produce real deliveries are not evenly matched. And the research and development costs of this round of salary increases, the cost of which will eventually fall on the purchase price of customers, is no other way out.
The motion of this machine is not on the turbine, but on the person. Although we all know that this condition will not last for too long, and as the bubbles are squeezed out, the higher you stand on the waves, the more it may hurt to fall. But no one wants to be the first to leave the field.
Some respondents said that in the future, we can go back and see how many of those people who have crossed the border from smart driving will be left behind when the repeatedly-cited dividing line arrives in 2027.
Staying here shows that the road crosses the border; the large area flows backwards, showing that the gap between technology and transformation is larger than imagined.