RMB 124 Million for a Chief Scientist, RMB 100,000 a Month for an Intern

The most expensive input in Chinese robotics this year is not compute, not reducers and not rare earth magnets. It is people, and the market has started pricing them like scarce components.

Graduate students with backpacks walking across a university campus, the entry point to China embodied AI hiring
Campus recruiting is now a bidding war for embodied AI graduates in China (Source: OFweek)

In April 2026, UBTech, the Hong Kong-listed humanoid maker, opened a global search for a chief scientist in embodied intelligence. The advertised total package started at RMB 15 million and reached RMB 124 million. Divided across working days, that is close to RMB 500,000 for a single day of work.

Shortly afterwards, according to internal reports in mid-April, Dreame’s robot subsidiary Mofa Yuanzi floated a RMB 200 million package for a chief scientist, resetting the ceiling again.

The demand curve is vertical

Recruitment data makes the shortage concrete. According to the Maimai spring hiring report, AI job postings in China rose 8.7-fold between January and April 2026. Embodied intelligence postings within that pool rose 15-fold, at an average monthly salary of RMB 62,000.

Liepin’s data for January to July shows robot-sector postings up 60.17 per cent year on year, with the humanoid segment up 147.02 per cent.

What has changed is the shape of the demand. Two years ago the scramble was for algorithm engineers to build the brain. This year the fastest-growing roles are the ones that make a product shippable: industrial engineers up 102.33 per cent, electrical engineers up 98.94 per cent, solutions engineers up 85.98 per cent and automation engineers up 85.51 per cent, against 43.73 per cent for algorithm roles.

That shift follows the hardware. UBTech put its Walker series into automotive and smart-manufacturing pilots in 2025, delivering more than 500 industrial humanoids, and has set a target of 10,000 units of annual capacity for 2026. AgiBot reached its 10,000th general-purpose unit in March 2026, roughly three months after the 5,000th.

Once volume becomes the constraint, the bottleneck stops being the model and becomes yield, cycle time, supplier quality and field service. None of those can be solved by hiring one famous professor.

The campus is now a bidding war

For 2026 autumn recruitment, a RMB 500,000 package for a strong master’s or PhD graduate in embodied intelligence has been described, without irony, as a cabbage price. Some startups now offer full-time pay to doctoral students in their third and fourth year on the condition that they join after graduation, with monthly salaries reaching RMB 100,000.

The pipeline cannot catch up. The Ministry of Education added embodied intelligence as an undergraduate major for the first time in 2026, supporting nine universities including Harbin Institute of Technology and Beihang. The first cohort reaches the job market in 2030.

A formula that prices people as valuation

The strangest artefact of this market, reported by TMT Post, is a headcount valuation rule circulating among investors. Recruit a top professor from Tsinghua or Zhejiang University as an adviser or team member, and the company’s valuation rises RMB 200 million to RMB 500 million. Poach a senior executive from a leading company, and it rises RMB 50 million.

That formula only works while the next round is priced by narrative. It stops working the moment an investor asks what the hire shipped.

The global comparisons are not reassuring for anyone hoping this is a local bubble. In May 2025, OpenAI paid USD 6.5 billion for an AI hardware team of 55 people, over USD 100 million per head. Boston Dynamics paid more than USD 2 million a year to take Tesla’s robot programme lead.

Who pays for the bidding

The funding is there, for now. IT Juzi puts first-half embodied intelligence funding at RMB 935 billion, up nearly fivefold, across more than 300 deals, with 22 companies valued above RMB 10 billion by mid-year.

UBTech’s own numbers show what the other side looks like. Its 2025 accounts recorded RMB 821 million of revenue from full-size humanoids on 1,079 units, a net loss of RMB 790 million, and RMB 4.92 billion of cash. The company can afford a RMB 124 million scientist. Most of its competitors cannot, and will pay for one anyway.

Salaries are the one cost that does not depreciate into a fixed asset. If the product does not ship, the payroll stays on the books and the valuation that justified it disappears.

Editor’s note: This is an adapted translation of the original OFweek report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://robot.ofweek.com/2026-10/ART-8321205-12003-30708922.html.

Leave a comment