There are so many robots that there aren't enough human workers...

“How much labor can this robot save for me?”

At the recently held 25th China International Industry Fair (hereinafter referred to as the “Industry Fair”), more than 350 robot enterprises gathered. The new species created by AI + robots, embodied intelligence, became the “star”. They are not necessarily humanoid, but are embodied intelligent agents, which are adding “eyes” (visual perception) and “brains” (overall cognition) to the “hands” (robotic arms). No matter how robots evolve, “replacing humans with machines” remains an eternal topic.

“There are many robots, but there aren't enough human workers.” The owner of a metal processing factory said that using “machines to replace humans” is not just about saving money, but about the shortage of workers. For example, it's very difficult to recruit grinding workers. “It's too hard, tiring, and there are occupational hazards. It takes several years to become a skilled worker, and young people are not willing to do it.”

The labor shortage caused by factors such as an aging population is accelerating the “replacement of humans with machines”. According to the “World Robotics 2024” report released by the International Federation of Robotics (IFR), the robot density in China has reached 470 robots per 10,000 employees. In 2023, it exceeded that of Germany and Japan, ranking third in the world, only after South Korea and Singapore.

How to calculate the cost - benefit of replacing humans with machines will directly affect the implementation speed of robots. However, in the actual application process, new problems such as “affordable to buy, unaffordable to use, and not easy to use” have emerged. The key lies in safety, controllability, and autonomy.

With the injection of large AI models, the era of “single - combat” embodied intelligence is over. Instead, there are assembly lines composed of “super production teams”.

Trend 1: From “Single - combat” to “Super Production Team”

The new species of industrial embodied intelligent robots was born at the 2024 Industry Fair. It has gradually become a trend to equip industrial robots with “brains” using AI. However, they can still only be responsible for a single process, such as a certain type of work in the quality inspection, grinding, loading and unloading processes.

At the 2025 Industry Fair, the biggest trend in the application of embodied intelligence in industrial scenarios is from “single - combat” to “super production team”, and from a single process to full - process collaboration.

At the booth of Weiyi Intelligence, a “Creation TRON New Production Line” was attracting a large crowd. This is a “super production team” composed of 8 industrial embodied intelligent robots. They are producing 4 different products and can complete the production line switch in a few minutes. If new materials are added to the production line, adjustments can be completed within hours. In the past, when switching production lines, traditional automated equipment often had to be idle for several days.

At the same time, each industrial embodied intelligent robot no longer produces in a “dumb” way according to the program. It can think like a human. For example, if a “robot colleague” next to it stops due to an unexpected situation, the task can be temporarily taken over by other robots to ensure the continuous production of the entire production line.

“Our goal is that this all - robot production group can do everything. It is a super production team that has both human intelligence and flexibility, and can be as efficient and stable as a machine.” Zhang Zhiqi, the chairman and CEO of Weiyi Intelligence, said.

Weiyi Intelligence has reached a cooperation agreement with Qiaxing Group, one of the world's top three beauty packaging enterprises. It is expected that this “super production team” will be implemented before the end of this year. In the future, high - end skin - care products such as L'Oréal and Estée Lauder bought by Chinese consumers may be packaged by the “Creation TRON” robot production line.

China's industrial manufacturing has entered the stage of intelligence and flexibility. AI vision technology is becoming the core driving force for the application of industrial robots.

On a new - energy vehicle production line, the integrally die - cast car body is being inspected and ground by robotic arms. Compared with manual grinding in the past, the processing time for a single piece has been reduced from 70 minutes to 18 minutes, and the qualified product rate has increased from 82% to 98%.

Recently, the “China AI Vision Industrial Robot Application Market Share 2024” report released by the international consulting firm IDC shows that Weiyi Intelligence ranks first in the domestic market share of AI vision industrial robot applications.

“The integration of AI vision and embodied intelligence has become the core driving force for the development of industrial robots.” Li Junlan, a research manager in the field of robotics and the Internet of Things at IDC China, said that industrial robots are evolving from “execution tools” to “embodied intelligence”. Chinese embodied intelligence enterprises such as Weiyi Intelligence are providing Chinese solutions for the digital and intelligent transformation of the global manufacturing industry.

Trend 2: “Cost - benefit Calculation” is the Key to Robot Implementation

Although the robot density in China has reached the top three in the world, and as a major manufacturing country, the industrial sector has become the preferred scenario for the implementation of robots in China. The installed capacity of Chinese industrial robots accounts for more than 50% of the global total, and its production volume, like that of automobiles, mobile phones, micro - computers, and color TVs, ranks first in the world. The “Government Work Report” in 2025 mentioned “embodied intelligence” and “intelligent robots” for the first time.

However, the development of industrial robots in China still lags behind the automation needs of the manufacturing industry. The IFR report shows that by 2030, the number of robots per 10,000 employees globally will be 300. This means that in 2030, the global manufacturing industry will still rely on about 388 million manual laborers, accounting for 97% of the total labor force, and will still be deeply trapped in the labor - intensive dilemma.

“How long will it take for the robot to recoup its cost?” At the Industry Fair, this is one of the most frequently asked questions by manufacturing bosses to robot enterprises. In the current highly competitive manufacturing industry, enterprises pay more attention to the return on investment.

“Customers don't care what technology you use behind the scenes. They only care if you can help them solve problems and make the cost - benefit calculation work.” Zhang Zhiqi gave an example. A typical industrial customer will conduct a strict ROI (return on investment) calculation: if the annual cost of a worker is 100,000 yuan, then the cost of the equipment to replace him must be recouped within 1.5 years, that is, the selling price cannot exceed 150,000 yuan.

“Many factories can afford to buy robots, but they can't afford to use them, and they're not easy to use.” Zhang Zhiqi said that in mixed workstations such as PCB board programming, fully automated solutions cannot cover hundreds of specifications of workpieces, nor can they handle the ever - changing material placement positions. Now, industrial embodied intelligent robots can achieve efficient and flexible production. Without having to retrain the robots, they can switch models and production lines, help workers independently introduce new products, and truly achieve cost reduction and efficiency improvement for enterprises, thus creating a new paradigm for industrial productivity.

In addition to cost, reliability is an “iron law” in the industrial sector. Industrial production has extremely low tolerance for misoperations. Any mistake may lead to the shutdown of the entire production line. Therefore, software and hardware must be optimized in a coordinated manner. The problem cannot be solved simply by piling up sensors or increasing computing power. Instead, the underlying algorithms should be improved to make industrial digitization safe and reliable.

Trend 3: Putting the Large Model in a “Cage”

In the past few years, the development path of industrial digitization was to release robots from the “cage”. Collaborative robots got the opportunity to collaborate with humans due to their safety advantages. However, this year, large AI models have gradually entered the industrial field, and new safety problems have begun to emerge.

“The certainty of the industrial sector collides with the randomness of AI. The content generated or operations executed by AI cannot go beyond the bottom line. It's not about not using AI, but about taking protective measures.” Chen Bingbing, the chief engineer of Dawning Network, told a reporter from “IT Times”. Why is the industrial field zero - tolerant of large - model hallucinations? Because once the output result is incorrect, it will lead to serious production accidents. A real - life example is that a factory tried to control a large model to raise a sluice gate by 40 centimeters, but the large model made a mistake and raised the sluice gate by 400 centimeters within 3 seconds, which was extremely dangerous and even caused the sluice gate to collapse.

Therefore, the AI audit model came into being. It puts the large model in a “cage” to prevent the robot from “going crazy”.

At this year's Industry Fair, Dawning Network proposed the concept of “integration of computing, control, and security” for the first time in the industry. Supported by more than 20 years of accumulated core technologies, it connects the three links of computing, control, and security, moving from “providing single products” to “outputting systematic methods” to avoid repeated exploration by enterprises in technology integration.

Among them, “computing” refers to the domestic computing hardware platform independently developed by Sugon, which includes standard industrial control computers, industrial motherboards, and AI models, providing diverse business models; “control” is an integrated industrial development platform that combines the advantages of PLC (programmable logic controller) ladder - diagram programming and the development habits of high - level languages, and also has the ability of semi - physical simulation; “security” comprehensively ensures industrial security through industrial network data analysis and firewalls.

At the booth of the 2025 Industry Fair, the four major products of Dawning Network are no longer isolated tools. They are oriented towards key links such as industrial programming, network operation and maintenance, simulation computing, and edge intelligence, each playing a different role in “computing, control, and security”, and together forming a complete industrial digital ecosystem.

“Even today, the awareness of industrial control security in some enterprises is not much different from that 40 years ago, and they can't afford the consulting fees that often amount to millions of yuan.” Chen Bingbing told a reporter from “IT Times”.

Liu Li, the president of Dawning Network, explained the reason behind this. The digital foundation of a large number of industrial enterprises in China is weak. Due to insufficient funds and a shortage of IT talents, small and medium - sized enterprises haven't even fully deployed basic process software, let alone integrate “computing, control, and security” and use a systematic thinking for long - term development.

When computing, control, and security move from being fragmented to being integrated, industrial digitization will truly have systematic certainty and resilience, providing a safe and reliable new path for the long - term competitiveness of the Chinese manufacturing industry.

Trend 4: Accelerated Localization

In addition to cost and reliability, the localization rate is also strongly linked to security.

Thanks to the localization of core components and the significant progress in robot controller technology, the localization ratio of Chinese industrial robots is entering an accelerated phase. A report from Soochow Securities shows that the localization rate of Chinese industrial robots has exceeded 50%.

Behind this figure is the domestic substitution of core components. The core components of industrial embodied intelligent robots include controllers (MCU), reducers, and servo motors. However, the key to determining the progress of domestic substitution is usability.

“The robot marathon and boxing match exposed some problems. Humanoid robots often had problems such as stopping and falling, and it took a long time for manual diagnosis. The staff was constantly spraying coolant on the joints. Behind this are problems such as reliability and safety redundancy.” Li Xiaojian, the product marketing manager of the MCU division of GigaDevice, said that as an upstream robot manufacturer with a cumulative shipment of more than 2 billion MCUs, GigaDevice is gradually solving problems such as battery life, joint overheating, processing capacity, and motion coordination from the perspective of full - stack chips such as MCUs, storage, and analog chips.

Currently, GigaDevice has cooperated with 127 global robot enterprises. Its full - stack chips have entered the mass - production lines of several domestic robot manufacturers, helping them achieve the goal of large - scale shipments of 500 to 5,000 units per year, and has become an important force in promoting the localization of robots.

“What can robots see? What can they hear? What can they do?” In Li Xiaojian's view, since robots and humans are in the same environment, it is necessary to establish functional safety standards and improve laws and regulations on information security, which requires the collaboration of the entire industrial chain, both upstream and downstream.

To enable manufacturing enterprises to “afford to buy, afford to use, and use well” domestic robots, the key lies in safety, controllability, and autonomy. Localization is an inevitable path and a path that forces the upstream and downstream sectors to develop.

This article is from the WeChat public account “IT Times” (ID: vittimes), author: Sun Yan, editors: Wang Xin and Sun Yan. It is published by 36Kr with authorization.