AI enterprises are poaching top talents with 10 million yuan annual salaries, and headhunters can earn enough to buy a full apartment from a single deal.

AI companies are offering sky-high salaries to poach talents, which also greatly benefits headhunters.

Recently, Tech Planet reported that an embodied intelligence company is recruiting AI algorithm experts with a total annual compensation package reaching 10.5 million yuan. With a 26% headhunter service rate, the corresponding service fee hits 2.73 million yuan.

"I have never taken such a high-value order," Yang Hongbo, co-founder of headhunting firm INR, told Pencil News. Among the orders he has actually handled, excluding the future value-added of equity, the highest-tier overall offer is around more than 7 million yuan.

If calculated at a rate of one quarter, the firm can get 1.75 million yuan from a single order.

Yang Hongbo is witnessing industry milestones almost every day.

In June this year, a fresh PhD graduate from University of Science and Technology of China had already received a 620,000-yuan offer and was preparing to join the company. Yang Hongbo recommended him to another position that better matched his underlying technical capabilities, and the hiring party directly raised the compensation to around 850,000 yuan.

Only 3 days were left before he was supposed to report for duty, a third company came with a new offer.

1.03 million yuan.

The talent was poached away again.

"I poached someone else's candidate, and then my candidate was poached by a third party."

In just a few weeks, the PhD graduate's technical ability did not improve abruptly, but the market price for him rose from 620,000 yuan to 1.03 million yuan.

01

How much profit can one single order make in the end?

"High-value orders do exist, but a 2.73 million yuan commission is extremely rare in the headhunting industry, which is not very representative as a reference,"

Yang Hongbo told Pencil News that based on the market he has actually engaged with, the service fee for most headhunting orders still hovers around 100,000 yuan, and an order exceeding 300,000 yuan is already considered a relatively large one.

It is not uncommon for senior Agent algorithm talents to earn 2 million to 3 million yuan per year, and the common headhunter service rate is about 20% to 25%.

However, there is another difference when it comes to actual settlement.

The "3 million yuan annual package" that many AI talents receive may include bonuses, stocks and options. A large number of orders in the headhunting industry are still calculated based on fixed cash salary.

For example, if a candidate's monthly base salary is 100,000 yuan, even if he has several months of year-end bonus and a large amount of equity, the headhunter fee may only be calculated as "100,000 yuan × 12 months × service rate".

"The vast majority of orders we are in contact with basically follow this logic," Yang Hongbo said.

Therefore, the 2.73 million yuan is the service fee of the headhunting company, which is not equal to the money that a single headhunter can take home.

In another report on employees of relevant headhunting groups, the insider confirmed the existence of this order, and estimated that after the 2.73 million yuan is divided between the company and the team, the actual income that falls to the individual is about 400,000 to 500,000 yuan. The common industry rate is about 20% to 25%, the annual common performance of traditional headhunters is about 700,000 to 1 million yuan, and the personal annual income mostly falls in the range of 200,000 to 400,000 yuan.

Of course, the price of top-tier talents has indeed been pushed to another magnitude.

In April this year, Ubtech publicly recruited a "Chief Scientist of Embodied Intelligence", with an annual salary starting at 15 million yuan and a maximum of 124 million yuan. The compensation consists of cash, benefits and equity. This position is responsible for defining the company's technical roadmap in humanoid robots and embodied intelligence.

But this kind of position is almost the top technical leadership of the entire company.

A case handled by Yang Hongbo can better explain how the ordinary high-end talent market raises prices.

A technical leader from a large tech firm originally had an annual income of about 3.6 million yuan. Later, a much smaller AI application company came to poach him.

The cash part was basically the same as his original salary, less than 2 million yuan a year.

What the company really bet heavily on was equity, granting about 4.8 million yuan worth of equity at one time. Calculated based on the standard of the entire offer at that time, the total compensation was close to 7 million yuan.

The reason why small companies dare to raise salaries to this level is backed by a revolution taking place in the AI industry.

The most obvious feeling Yang Hongbo has this year is that "AI application positions have begun to outnumber basic model positions, which is different from last year. In the second half of this year, the market and talents will become much calmer, and more attention will be paid to the landing of practical applications and the selection of application tracks."

When companies are willing to raise salaries, it usually means that the talent is closer to products and revenue.

Headhunters' commissions also rise accordingly.

But this money is not easy to earn.

Candidates reneging at the last minute, original companies raising salaries to retain talents, competitors poaching, and even the newly formed team of the client being suddenly disbanded, will directly make an order that has been followed for several months worth zero.

In the headhunting circle, this situation is called order jumping: the boom drives up the unit price, and also drives up the risk at the same time.

02

3 days before joining, the candidate was poached away again

The obvious change in this year's AI talent market is that the value of many people is being recalculated by positions.

"In June this year, I met a newly graduated PhD from University of Science and Technology of China."

He has experience publishing papers in top journals and top conferences, and his main research field is computer graphics.

Following the traditional career path, the most obvious jobs he could consider are game development and graphics R&D. When he graduated, a company gave him an offer of 620,000 yuan.

He was already prepared to take the offer.

Yang Hongbo later got in touch with this candidate, found that his underlying technical capabilities could be applied to another type of R&D position, so he recommended him to another company.

After knowing that the original offer was 620,000 yuan, the hiring party directly added more than 200,000 yuan to the compensation.

Around 850,000 yuan.

The candidate thought the position was more suitable and the income was higher, so he decided to change jobs.

There were only 3 days left before he was officially supposed to report for duty.

The third company suddenly showed up.

The position was AI Infra, with an offer of 1.03 million yuan.

In the end, the candidate joined the third company.

"So I poached someone else's candidate first, and then my candidate was poached by someone else in the end."

The interesting part of this story is that the candidate did not publish several new papers abruptly in these weeks, nor did he suddenly gain three more years of work experience.

What changed is that his capabilities were placed in different market scenarios.

Graphics R&D, 620,000 yuan.

A more underlying R&D position, 850,000 yuan.

Entering the then more scarce AI Infra track, 1.03 million yuan.

Yang Hongbo found that many fresh master's and doctoral students have the problem that their skills cannot be maximized for monetization.

"They know what they have learned, but they don't know what else they can do. This is exactly where the value of headhunters lies: headhunters can help candidates build a basic perception of their own market value and positioning."

The most intense part of the current AI talent market is exactly here.

What many companies are competing for is not standard talents for standard positions, but a group of people with transferable underlying capabilities.

This shortage can also be observed from the external market.

In this year's graduation season, the NLP Lab of the Department of Computer Science of Tsinghua University announced the destinations of 14 graduates, among whom 4 joined ByteDance and 1 joined DeepSeek. Liu Zhiyuan, a professor in the Department of Computer Science of Tsinghua University, revealed that the highest annual salary for PhD graduates in the large model direction of this lab can reach more than 6 million yuan, and master's students can also get compensation at the million-yuan level.

With few talents and many competing companies, offers will naturally keep rising.

The most exaggerated candidate Yang Hongbo has ever met received nearly 20 offers in one month.

At that time, he was already preparing to leave his original job, had very little work to do, and sat in the phone booth every day to attend interviews.

One or two interviews a day, for half a month in a row.

He finally accumulated nearly 20 offers in hand.

However, Yang Hongbo reminded that a large number of offers do not reflect the individual scarcity of the talent, but may just be the "change in supply and demand relationship" of the business track.

If a person happens to step into the popular direction of the year, such as AI applications, Agent or Workflow, even if his monthly salary is only 30,000 to 50,000 yuan, it is possible that more than a dozen companies will reach out to him at the same time.

When it comes to top talents, companies will compete on more dimensions.

Salary, equity, and position are just the most intuitive items.

Candidates will also ask: What is the progress of business commercialization? What about the company's cash flow? What is the background of the investors? How much authorization does the boss give to this position? And how patient the decision-maker is for this project.

Sometimes HR can no longer persuade the candidate, and the boss has to step in personally.

On the surface, the competition for AI talents is just raising salaries, but in essence, it is a competition for a small group of people who can determine the success or failure of the project.

03

The stronger the AI industry develops, the more high-quality headhunters earn

The headhunting industry is also impacted by AI, but there is no need to be overly anxious. Yang Hongbo also believes that AI will take over a large number of repetitive tasks, but the real value of high-end headhunters still lies in connection and career cognition.

Now AI can already search resumes, match positions with talents, and screen candidates, and dedicated Agent products have emerged in the recruitment industry.

But when an enterprise encounters a candidate who has more than a dozen offers in hand, how to persuade him to join is the most difficult part.

Corporate HR may contact dozens or even hundreds of candidates a day, send company introductions, continue the conversation if the candidate is interested, and move to the next one if he is not.

The work of headhunters is more like sales.

When a truly suitable candidate appears, headhunters need to communicate repeatedly: why he wants to leave, why he does not want to leave, what he is worried about, how much salary he expects, and whether he values the business prospect more or the position more.

Yang Hongbo believes that there is another often underestimated value in this process — career planning.

What opportunities and risks will there be if he joins Company A; if he joins Company B, he will get more money, but is it possible that the business will disappear in two years; the candidate is doing computer graphics now, are there other types of companies in the market willing to pay a higher price for his capabilities.

All these factors affect the final deal.

The hotter the AI industry is, the more important this kind of capability becomes.

In the Tech Planet report, some headhunters mentioned that some enterprises will assign one position to a large number of headhunters at the same time in order to recruit people quickly, and there is even a situation in the market where "hundreds of headhunters hunt for one single position". For the same candidate, the one who first recommends him to the enterprise talent pool may get the order.

At this time, simply finding a candidate's phone number no longer has much barrier. The real business is to poach the candidate and then make him stay stably.

This is very interesting: recruitment is clearly one of the industries that is most likely to be improved in efficiency by AI, but the business of AI-related headhunters has become better this year.

The headhunters that can stay in the end are those who can provide high-quality judgment for candidates, help them negotiate, and complete the connection between people.

This article is from the WeChat official account "Pencil News" (ID: pencilnews), written by Song Ge, edited by Huang Xiaogui, and authorized for release by 36Kr.