Some people earn 10,000 yuan a day, while others promise not to take any money.
In this round of AI wave, talents are the first group to be repriced.
From large model companies competing for algorithm engineers with high salaries, to tech giants offering “sky-high” compensations to top researchers, the price of talents has been constantly refreshing market perceptions. At the stage of rapid iteration of technical paths and extreme scarcity of core talents, a key employee may even affect the R&D rhythm and product direction of an entire company.
While employees are being fought over at high prices, how much should the CEO standing at the top of the company’s decision-making hierarchy be worth?
This question cannot be answered solely by a single payroll. In technology companies, the CEO is not only an operator and manager, but may also be the founder, core technical leader and major shareholder, whose returns are usually scattered in cash compensation, equity value and future value-added expectations. As a result, some CEOs receive high annual salaries, while others voluntarily waive their salaries. Judging solely by the level of salary is not sufficient to determine their actual returns.
“Future Tech Circle” sorts out the compensation, personnel efficiency and equity holding value data of 18 leading technology companies across four tracks: AI large model, autonomous driving, chip semiconductor and embodied intelligence, trying to answer one question: How exactly do technology companies price their CEOs and core talents, and what kind of resource allocation and future expectations do these prices reflect?
The largest divergence in the chip track? The highest CEO salary hits 7.2 million yuan, while the lowest is “zero salary”
Overall, the CEO salaries of the 18 companies show a clear gradient distribution: 2 companies have annual salaries above 5 million yuan, 4 companies between 3 million and 5 million yuan, 9 companies between 1 million and 3 million yuan, 3 companies below 1 million yuan, with an average annual salary of 2.572 million yuan and a median of 2.098 million yuan. If calculated on the basis of 250 working days a year, the average daily salary is about 10,000 yuan.
Calculated by track, autonomous driving companies have the highest average annual CEO salary, reaching 3.673 million yuan; excluding the special CEO Zhu Yiming, chip semiconductors follow closely with 3.365 million yuan; AI large model and embodied intelligence are 1.835 million yuan and 1.48 million yuan respectively.
Among them, the chip semiconductor track has the largest compensation span. Zhang Jianzhong of Moore Threads receives an annual salary of 7.2 million yuan, ranking first among the 18 companies, about 1.2 times that of Han Xu of WeRide, the second place; Zhu Yiming of Changxin Technology, which also belongs to the chip track, promises not to receive any salary before the company makes a profit, becoming the only “zero salary” CEO on the list. However, Yan Junjie of MiniMax also publicly announced in July 2026 that he will not receive any compensation before the company achieves AGI. If AGI cannot be achieved ahead of schedule, he will most likely join the “zero salary” group in 2027 as well.
The salary contrast between Zhang Jianzhong and Zhu Yiming reflects the different choices of the two companies in founder incentives. Moore Threads is at a critical stage of sprinting for the capital market, accelerating product iteration and commercial expansion. Zhang Jianzhong’s cash compensation not only includes the pricing of the founder’s management responsibilities and industry experience, but also reflects the company’s emphasis on the stability of the core management team. The memory chip industry where Changxin Technology is located is characterized by heavy assets, long cycles and high upfront investment. When the company has not yet achieved profitability, Zhu Yiming chooses to give up cash compensation, binding his personal returns more closely to the company’s long-term development and equity appreciation.
As the explosion of AI computing power demand drives a surge in memory demand, this promise has reaped returns in the first half of 2026. Changxin Technology achieved operating revenue of 150.31 billion yuan, a year-on-year increase of 873.6%, and the net profit attributable to shareholders reached 77.61 billion yuan, completely turning losses into profits and becoming the domestic leading player in memory chips.
Different from the head-to-tail divergence in the chip track, the CEO compensations of autonomous driving companies are mostly concentrated at high levels. Han Xu of WeRide gets 5.799 million yuan, Yu Kai of Horizon Robotics 3.929 million yuan, and Peng Jun of Pony.ai about 3.35 million yuan, all exceeding 3 million yuan, ranking 2nd, 3rd and 5th respectively among the 18 companies, while Cao Xudong of Momenta gets 1.612 million yuan, significantly lower than CEOs of other companies in the same track.
The average CEO salary in the AI large model track is 1.835 million yuan, the most balanced among the four tracks. This forms a contrast with the public impression of the large model industry as “high investment and high salaries to grab talents”: The main expenditure of large model companies is concentrated on computing power infrastructure and R&D talents, and the proportion of CEO cash compensation in the overall cost is relatively limited. Especially when the founder holds a large number of shares at the same time, his long-term returns do not necessarily need to be reflected through high salaries.
The average annual CEO salary of embodied intelligence companies is 1.48 million yuan, the lowest among the four tracks. Zhu Qiuguo of Deep Robotics gets an annual salary of 872,900 yuan, the lowest-paid CEO except for Zhu Yiming who is “zero salary”. Compared with chip companies that have formed a relatively mature revenue model, embodied intelligence is still in the stage of parallel product verification, production capacity construction and scenario expansion. Enterprises need to invest limited funds in R&D, supply chain and market implementation, so the cash compensation of the management team is generally more restrained.
Employee Salary and Compensation Multiple: Commercialization Stage Affects Salary Gap
If CEO salary reflects the incentive strategy at the corporate governance level, then the average employee salary can better show the real money investment of each company in the talent war.
Overall, the average employee salary of the 14 companies that disclosed data has a very wide span, and both extreme values appear in the autonomous driving track: The highest value Momenta reaches 1.041 million yuan, while the lowest value WeRide is only 222,000 yuan. Excluding 4 companies with incomplete data disclosure, the median average employee salary of the 14 calculable companies is about 803,000 yuan.
The employee salaries of the other three tracks are relatively concentrated, with an internal difference of 250,000 yuan for AI large models, 189,000 yuan for chip semiconductors, and 25,000 yuan for embodied intelligence. Only the autonomous driving track has a difference of 819,000 yuan, with a span of nearly 5 times, and the internal difference far exceeds that of other tracks.
The core of this gap lies in the different stages of commercialization. Momenta takes the mass-produced intelligent driving route and cooperates deeply with car manufacturers. By the end of 2025, the number of mass-produced vehicles equipped with Momenta’s system has exceeded 680,000 units, with annual revenue of 2.413 billion yuan, of which license revenue accounts for more than 40%. With a relatively large revenue scale, it has the ability to pay higher salaries.
WeRide is still in the early stage of Robotaxi commercialization. Its total revenue in 2025 was 685 million yuan, net loss 1.655 billion yuan, and R&D expenditure 1.4 billion yuan. The revenue is far from covering the R&D investment. More critically, the personnel structure: WeRide has 3,523 R&D personnel, accounting for 92.69% of the total number of employees. The number of R&D personnel of WeRide alone is more than twice the total number of employees of Momenta, and the huge R&D team leads to a significant dilution of per capita compensation.
Another noteworthy point is that the average employee salaries of the other three tracks are all above 700,000 yuan, while the two embodied intelligence companies are only just over 300,000 yuan. This gap also stems from the difference in the industry stage: Embodied intelligence is still in the transition stage from technical verification to large-scale commercial use.
According to a report by Interact Analysis, although the global production of humanoid robots exceeded 20,000 units in 2025, only about 10% of them have actually entered real operation scenarios such as factories and logistics, and most of the rest are used for scientific research, data collection and entertainment. This pattern has not fundamentally changed so far. At the same time, the assembly of production lines and delivery links require a large amount of manpower, so the per capita salary is naturally diluted. In contrast, large model and chip companies sell “designs” and “algorithms”, with extremely low marginal costs, so the per capita salary is higher.
Then look at the compensation multiple between CEO and employees. The multiple distribution of the 14 calculable companies is concentrated on the whole, but the tail is very long: the vast majority of companies are concentrated in the range of 1 to 5 times, totaling 12; while WeRide ranks first with 26.1 times, more than 3 times that of the second-ranked Moore Threads, and 6 to 20 times that of most companies.
After removing WeRide whose multiple is significantly amplified due to employee salary, the distribution of the remaining 13 companies can better reflect the general law of the industry. The average values of chip semiconductors and embodied intelligence are both around 4 times, autonomous driving is about 2.9 times, and AI large model is about 2.3 times, all in a relatively gentle range.
This shows that after excluding extreme values, the cash compensation gap between CEOs and employees of hard technology companies is generally at a relatively controllable level. Whether it is chip semiconductors, autonomous driving, AI large models or embodied intelligence, the average value of each track is between 1 and 5 times.
Average Personnel Efficiency: Chip Companies Take an Overwhelming Lead, Autonomous Driving Is Still Climbing Up
In addition to employee salaries, another noteworthy dimension is personnel efficiency, which measures the output efficiency of each employee.
Overall, the average personnel efficiency of the 18 companies has a very wide span: The highest value Cambricon reaches 6.226 million yuan, while the lowest value WeRide is only 199,000 yuan, the difference between the head and the tail exceeds 31 times, and the overall median is about 1.36 million yuan.
The reason why Cambricon’s personnel efficiency is so outstanding is that while its revenue in 2025 increased by more than 450% year-on-year, the total number of employees only increased by 127, and the per capita revenue generation was sharply amplified.
In addition to the company’s personnel and revenue rhythm, the level of personnel efficiency is also highly related to the business model. Chip companies sell “designs”, once the tapeout is successful, the marginal cost is extremely low; autonomous driving companies sell “the future”, and they are still in the stage of “burning money for R&D with revenue lagging behind”, a large number of high-paid R&D personnel have not yet been converted into large-scale revenue; the AI large model and embodied intelligence tracks are in the middle, with large internal differences, reflecting different commercialization progress.
There is another quite contrasting conclusion: The average cash compensation of autonomous driving CEOs is the highest among the four tracks, but the average personnel efficiency is the lowest among the four tracks. Perhaps at the stage when commercialization has not been fully realized and technical routes are iterating rapidly, retaining founders with both technical background and financing ability is the top priority expenditure for these companies.
Equity Value: Behind High Valuations, Capital Is Paying for “the Future”
The founders of hard technology companies generally have high net worth. Among the 14 companies that disclosed equity value, the overall median is about 8.2 billion yuan, and the highest value Chen Tianshi of Cambricon reaches 185.91 billion yuan. This level is already top-tier in the hard technology track.
But the logic behind the high net worth of technology company founders is different from that of traditional industries. Wealth accumulation in traditional industries often relies on profit dividends and asset appreciation, while the equity value of hard technology companies comes more from the capital market’s pricing of the future. A chip company may not be profitable yet, but as long as its technical path is recognized, its valuation can be realized in advance.
And this pricing logic stems from the imagination space for the future. Whether it is AI, autonomous driving, chip semiconductors, or embodied intelligence, there is huge “story premium”, and people are willing to pay high valuations in advance for the upcoming technological revolution.
This logic of “paying for the future” is most vividly embodied in the chip track. Chen Tianshi of Cambricon has a net worth of 185.91 billion yuan, ranking first by a huge margin, twice that of Zhu Yiming of Changxin Technology, the second place. As the first stock of domestic AI chips, Cambricon’s stock price keeps rising, and Chen Tianshi, as the founder, holds a relatively high proportion of shares, so his net worth naturally rises accordingly.
This also explains why CEO’s cash compensation is often not positively correlated with equity value. Zhang Jianzhong of Moore Threads ranks first with 7.2 million yuan in cash compensation, but his equity value is about 18.54 billion yuan; Chen Tianshi of Cambricon only gets 1.464 million yuan in cash compensation, but his equity value is as high as 185.91 billion yuan. This mismatch between “cash compensation” and “equity value” is a typical feature of the founder incentive structure of hard technology companies – some founders are willing to sacrifice current cash compensation in exchange for long-term equity appreciation of the company.
But capital’s patience has its boundaries. The future of hard technology companies must ultimately land on their commercialization capabilities. As tracks such as Robotaxi, embodied intelligence and large models gradually enter the verification period, the valuations of those companies that cannot turn “stories” into revenue and profits will eventually return to rationality.
At that time, whether the equity value
