Hefei: The Next Generation Leap in Industrial Robots

When it comes to stock trading, look at the Jin Qilin analyst research reports—authoritative, professional, timely, and comprehensive—helping you uncover promising investment themes and opportunities!

（Source: The Paper）

To improve productivity, every generation has its solutions. Tool-driven industries—industrial robotics—created alongside the upgrade of manufacturing, are now seeing their own upgrade.

Over the past 70 years, industrial robots, at their essence, have been “machines that carry out instructions.” Today, industrial embodied-intelligence robots are gradually learning capabilities for autonomous perception and decision-making.

In Hefei, this generational leap is unfolding.

Now, more than 200 robot-related companies have gathered in Hefei, and in 2025 Hefei’s industrial robot output will grow by 1.1 times. From the “brain” that delves into core algorithms, to the “small brain” that ensures stable operation, then to various nimble “joints” and “limbs,” and even whole-robot integration—Hefei has formed system-level supporting capabilities.

Finding the limit scenarios

In the past, the tough problems that automation and intelligentization couldn’t easily crack were often hidden in non-standard environments with dynamic disturbances and demanding precision requirements.

Newer companies directly designed embodied-intelligence agents in response to specific industrial scenarios. The logic is not to build a general-purpose robot, but to pick the hardest, most valuable scenario—grow a native intelligent agent for that scenario. Capabilities accumulated on that high ground can then spill over into other fields.

You can see a typical example of this path in U&I Zhihui. Its starting point was the School of Mechanical Engineering at Xi’an Jiaotong University. Founder Zhang Chaohui, as a PhD in robotics, conducted related research around smart factories with Professor Mei Xuesong during his time at the university. At that time, Zhang Chaohui had already led the development of multiple robot prototypes. After starting the business, the headquarters moved to Longgang, Shenzhen, where it established R&D and production centers. In 2024, the headquarters and the future listed entity moved to Feidong in Hefei.

This “mentor-supported + student entrepreneurship” model is not uncommon in the hard-tech sector, but U&I Zhihui’s special feature is that technology transfer is extremely fast. Within two years of founding, U&I Zhihui successfully completed deliveries for two major customer scenario deployments—Michelin tire inspection and customs inspections for prohibited goods—leveraging that momentum to begin breakthroughs and deeper development in industries such as power, semiconductors, and 3C.

“In the early stage of entrepreneurship, we felt that opportunities were everywhere—like ‘finding a nail with a hammer,’ but our analysis of the market wasn’t deep enough, so we wanted to try everything.” U&I Zhihui’s market director Guan Jian told The Paper reporter. After two or three years of practice and a deeper understanding of the market, they realized the team’s strengths were better suited to “high-difficulty scenarios.”

U&I Zhihui’s “one brain, multiple modalities” embodied-intelligence robot. Photo provided by interviewee

U&I Zhihui focuses on the extremely demanding scenario of “the semiconductor workshop,” where cleanliness requirements are high, space is tight, and docking precision requirements are extremely high. By integrating “a mobile base + a robotic arm + vision,” it designed from scratch a narrow-bodied, high-precision mobile manipulation robot. Today, in the semiconductor and energy-chemicals sectors, it has already reached a leading position in global market share. It has partnered with two of the top three global wafer foundries, enabling “zero loss” for wafer handling robots. In the energy sector, it performs unmanned inspections in scenarios such as thermal power plants and offshore oil platforms, replacing hazardous manual operations such as high-voltage cabinet handling. High temperature, high voltage, radiation, cramped spaces, and hazardous tasks—these are exactly the places where humans either don’t want to go or would face risk, and these are also where embodied-intelligence robots deliver value.

In industrial scenarios, embodied-intelligence robots are no longer merely “pilot projects,” but have entered the core production workflow as “productivity tools.” U&I Zhihui’s industrial embodied-intelligence robot solution has been sold in more than 30 countries and regions, including China, Japan, South Korea, Germany, and others.

Understanding “generalization”

As embodied intelligence has entered the public eye, understanding of “generalization” has also become deeper. For the many complex scenarios in the industrial field, “generalization” does not mean one robot can do everything; rather, within a specific sub-scenario, it is the capability to solve most variations of needs with one platform.

Leju Robotics’ practice is also an interpretation of this idea. In 2016, Leju was founded in Shenzhen. To accelerate the rollout of embodied robots and achieve industrialization of the industrial sector, it settled in Hefei Yaohai in 2024.

The process of industrial exploration wasn’t very smooth. When Leju first got involved in industrial projects, Leju Robotics’ assistant vice president Ren Guangjie and the team visited more than 70 factories, recording the standardized process of hundreds of scenarios. In the end, the approach converged: choose a few broadly applicable skills and scenarios, and try to validate them first.

In the past, factory automation addressed “big logistics.” Vertical warehouses, conveyor belts, and stackers handled “from the warehouse to the factory gate.” Traditional AGVs handled “from the factory gate to the line-side warehouse.” Because the space is narrow, incoming materials are non-standard, and precision requirements are high, the segment from line-side warehouses to machines, from machine to machine, and from finished goods off-line to the inspection stations has a relatively low automation coverage. This “last mile,” in the past, often relied on people—using experience and physical effort to get it done. What embodied intelligence needs to take over is precisely this most difficult-to-automate work segment.

In 2024, for Leju’s humanoid robot project at FAW Hongqi’s Changchun automobile plant, the team made many small specialized model sets, digging into every specific scenario. Ren Guangjie and the team “stayed in” the factory—observing workers’ logic, recording and learning—so the robots could mimic every detail and figure out every move. The team conducted continuous testing and debugging, iterating multiple vision models and navigation plans, with core algorithm optimization carried out dozens of times.

“Only after verifying this technical route can you know it’s not feasible.” Ren Guangjie said that in exploring humanoid robots entering the industrial field, there was no reference target to look at—so they had to explore on their own. Which hardware is more suitable and which technical architecture is the most practical must be refined repeatedly across different scenarios. “We basically stayed in the factory, did a lot of ‘dirty and exhausting work,’ negating ourselves again and again, overturning ourselves, and iterating ourselves. In the end, it could be deployed because there had been a large amount of trial and error within it. This helped them back out product iteration from customer needs.”

“Kuafu” carries out the task of applying sealant and loading parts at FAW Hongqi’s factory. Photo provided by interviewee

Industrial robots must meet production requirements, which place demands on operating time, yield rate, and stability. They also need to solve the stability of the “brain” algorithms, the “small brain” control, and the stability of the onboard hardware. Today, Leju’s “Kuafu” humanoid robot has been deployed for testing on FAW Hongqi’s vehicle production lines, and can complete operations such as material handling and precise grasping. In laboratory real-world tests, continuous operation time with high stability has reached more than 9 hours.

Continuously evolving

The foundation of Hefei’s robotics industry is inseparable from the close collaboration of long-established companies like HeLi, which has spent nearly 70 years deepening its efforts.

Anhui Heli Co., Ltd.’s core business is forklifts and material handling. And today, one of the most mature commercial scenarios in the robotics industry is smart logistics. As it has progressed to where it is now, Heli has witnessed the entire journey of China’s manufacturing—from “nothing but empty foundation” to becoming the world’s number one in terms of scale. Over nearly 70 years of development, it has kept the ability to evolve—from mining machinery to forklifts, from internal-combustion to electric, from mechanical transmission to intelligent decision-making.

Mining machinery is the foundation of Heli’s beginnings. In 1958, a new plant for Hefei Mining Machinery Factory was established. In 1960, it was renamed to Hefei Hoisting and Transport Machinery Factory, starting to extend its reach into the material handling sector. In 1983, it decided to make forklifts the leading product and spun off other mining machinery equipment. The foresight of this decision lay in anticipating the huge demand for material handling equipment during China’s industrialization process. After that, for four decades, forklifts became Heli’s core main business, establishing its leading position in China’s industrial vehicle sector.

In recent years, Heli has undergone technological accumulation—from mechanical transmission to electric transmission—and an architecture upgrade—from single-unit control to system-level intelligence.

In the 1980s and 1990s, Japanese forklift technology was leading globally. Heli introduced advanced design concepts, manufacturing processes, and quality control systems through cooperation with Japanese companies. Then, with the explosion of lithium battery technology for new energy vehicles, Heli transitioned from traditional forklifts to the development of electric forklifts. Unmanned operation is Heli’s most valued segment right now. The VCU (vehicle control unit) is the “brain” of an electric forklift, responsible for power distribution, state monitoring, and safety control. True intelligence requires layering perception, decision-making, and planning capabilities on top of this “brain,” enabling the forklift not only to run according to instructions, but also to “see for itself, think for itself, and drive for itself.” To achieve this, Heli dispatched technical teams overseas for training, then after returning, developed core control systems.

Heli’s 55-ton lithium electric forklift and Heli i-series intelligent industrial vehicle products. Photo provided by interviewee

As a controlling subsidiary of Anhui Heli, Anhui Heli Yufeng Intelligent Technology Co., Ltd. has become the core carrier for Heli’s transition from a forklift manufacturer to an intelligent logistics integrator. “Intelligent logistics equipment is the ‘next stage’ for industrial vehicles.” Heli Yufeng’s R&D head Wang Lei told The Paper reporter. From the development trends of the AGV (automated guided vehicle) industry, the current domestic layout is relatively complete, and it will further expand overseas. “As a leader, we need to solve problems others can’t solve. Advancing composite navigation is not a single solution; it is the integration of mechanical, vision, and algorithms, with stronger autonomous decision-making capability.”

Old trees with new branches

As the leader in China’s industrial vehicle industry, Anhui Heli’s value goes far beyond “high market share.” What it has accumulated on this land is an entire set of methodologies, talent systems, and industrial culture about “how to build industrial equipment.” For decades, Heli has consistently cultivated professional talent for roles including mechanical design, hydraulic systems, electrical control technology, and production management.

A group of technical backbones and management personnel who came out of Heli later founded various companies related to logistics equipment and industrial automation in Hefei and surrounding areas. Among them, JingSong Intelligent is one of the most successful representatives.

At that time, although Heli remained a market leader in traditional forklift domains, in emerging areas like AGVs and intelligent warehousing it was still in the exploration stage. JingSong Intelligent’s founder Yao Zhijian chose an entrepreneurial path of “doing new things in one’s old line of work”—intertwining Heli’s “logistics equipment gene” with the new “intelligent technology gene.”

In 2007, Yao Zhijian established the independent brand JingSong Intelligent, starting from automated vertical warehouses.

“In the first few years, we revitalized one order by relying on another. In the heavy-asset sector, you need to invest before you can get returns. In the early stage, the investment was all paid out of the founders’ own pockets.” JingSong Intelligent’s deputy general manager Wang Dan told The Paper reporter. When she joined JingSong Intelligent in 2013, because the company had secured several major orders, things improved somewhat. After obtaining land allocation indicators that year, it consolidated the previously fragmented segments, and the company began to develop for the better.

“The sunniest days” happened to be the best window period for transformation. The business model of vertical warehouses determines that it is a “long-cycle, low-frequency” operation. After a vertical warehouse system is built, its service life is often more than 10 years, and customers typically won’t have new purchasing needs for a considerable time. This means it is hard for the vertical warehouse business to form a virtuous “repeat purchase” cycle. At that time, widely used magnetic-guided and QR-code concealed AGVs had high maintenance costs. This not only increased customers’ operating costs, but also reduced the overall reliability of the system.

JingSong Intelligent realized that true “unmanned operation” should not be “a different way to have people serve the equipment,” but rather “the equipment adapts to the environment by itself.” This pointed to another technical route: unguided unmanned forklifts without rails.

In the early days, unmanned forklifts were still a novelty, and most customers held a wait-and-see attitude. Those willing to try were often leading companies or customers with special requirements (cold storage, high-risk environments, etc.). As the demographic dividend faded, labor costs rose, and policies for smart manufacturing advanced, unmanned forklifts entered a phase of rapid popularization. JingSong’s orders began to grow significantly.

“The technology advanced faster than we expected. From 2013 to 2016, all kinds of information and demands flooded in. When we went to exhibitions, there were over a thousand pieces of information at one show—many customers’ needs—which represented a market trend.” In the past, the profits from automated vertical warehouses supported the AGV business. Today, about 80% of JingSong Intelligent’s business is AGV demand. Even if there are vertical warehouse needs, they are also accompanied by AGV-solution package operations. In the past, persuading customers meant convincing them; now customers more often want to verify the reliability, authenticity, and validity of data. “To reduce labor by 30%, whether customers think it’s 30% or 35%—they want it to be 40%. How do we make it happen? That’s the kind of scenario we have. ” Wang Dan said.

JingSong Intelligent positions its products at a high end and in heavy-duty load applications. “These are our strength products. We choose a differentiated competition track and do the hardest parts. We truly invest a lot in innovation, and our R&D expenses are higher than the industry’s average R&D investment. Slow work produces fine results.”

JingSong Intelligent’s multiple AGVs coordinating for collaborative work. Photo provided by interviewee

Based on customers’ needs, JingSong Intelligent has done many innovative case studies, and also created numerous domestic firsts and industry firsts, breaking international technology barriers. With the manufacturing case scenarios it has already completed—2,000 case scenarios—it is also working on the challenge of rapid deployment, building big data and big models so that unmanned forklifts become smarter and have their own judgment capabilities. For scenarios still requiring human judgment for picking and sorting, JingSong Intelligent is currently researching humanoid robots, to solve the “last mile” problem for customers.

A virtuous feedback loop

A basic rule of industrial robot industry evolution is that no single company or one technical route can independently satisfy all the needs of a smart factory. The diversity of industrial scenarios determines that technical paths must be diverse.

Newer embodied-intelligence companies like U&I Zhihui and Leju, from day one, define product architecture with an AI way of thinking, running perception, decision-making, and control within an end-to-end framework. This endowment means they are most suited to extreme scenarios—the clean environment of semiconductor workshops, narrow corridors, and high-precision docking; precision assembly for 3C manufacturing; high-risk inspection in energy and chemical industries. In these scenarios, where fault tolerance is low, technical thresholds are high, and value density is high, the advantages of “native design” can be fully leveraged.

For logistics equipment companies like Heli and JingSong, their core competitiveness lies in deep understanding of industrial scenarios and their capability for large-scale delivery, along with deep insights into the implicit rules of factory logistics. This endowment means their main battlefield is large-scale logistics in structured environments. These scenarios also demand high reliability, but the technical difficulty is relatively more constrained, putting more emphasis on cost control and delivery efficiency.

These two are not zero-sum competition; they are misaligned but complementary. At the same time, they chose a type of soil where technology can evolve quickly.

Hefei has several major industrial clusters—new energy vehicles, photovoltaics, new displays, home appliances, and others. These industries are precisely the largest application markets for industrial robots and embodied intelligence. When local companies do R&D, they can find seed users and testing grounds nearby. The maturity of hard-tech comes from growing “in real scenarios.” Hefei’s industrial clusters provide robot companies with high-density, diversified, and realistically collaborative scenarios. This kind of scenario resource is more precious than any policy subsidy.

From this perspective, the rise of Hefei’s robotics industry is the inevitable result of a “positive feedback loop” between industrial clusters and technological innovation.

A massive amount of information and precise analysis—available in the Sina Finance APP