On August 19, 2026, the World Robot Conference (WRC) opened in Yizhuang, Beijing. With over 300 enterprises, more than 3,000 exhibits, over 300 newly launched products, and 17,000 participating robots and teams, it is the largest in the history of the event and also the most down-to-earth one so far, because ordinary people can not only watch the show but also actually afford to buy the robots. Walking into the venue, one will find that embodied AI has entered the 10,000 RMB era. The robot dog from Vbot is priced at 9,988 RMB, and the "Bridge" from Zeroth Robotics is at the 10,000 RMB level. What brings robots closer to the wallets of ordinary people is precisely the countless grassroots players in the industry. Imagine if technological and academic resources were held by universities, while capital and application scenarios were controlled by tech giants. The embodied AI world would then become an oligarchy, leaving small enterprises and teams with no chance to stand out. Not only would the imagination and creativity of the entire industry be locked in the laboratories of a few giants, but the distance from ordinary people would also be hard to bridge. Because the scenario coverage of leading companies has boundaries, they must standardize and scale up to amortize R&D costs. However, real-world demands are fragmented and personalized. If a small bar wants a bartending robot, or a script-kill store wants an NPC robot, who will satisfy these needs that big companies look down upon and highly paid algorithm scientists find uneconomical to tackle? That is the grassroots world.
The Grassroots World Let's look at a set of data first. In April 2026, at the Humanoid Robot Half-Marathon in Yizhuang, Beijing, a small robot 75 centimeters tall with an antenna on its head ran onto the track with short shuffling steps. It collapsed from exhaustion at the 10-kilometer mark, arousing immense sympathy from netizens, who joked that it was an "underage child laborer" and a "toddler contestant." Everyone remembered this cute creature. But almost no one knows: who built it? This is precisely the cruel side of the embodied AI field; the public only sees the tip of the pyramid. Looking solely at the humanoid robot track, the "2025-2026 Humanoid Robot Industry Development Research Report" released by the Humanoid Robot Scenario Application Alliance showed that there were over 215 financing events for humanoid robots in 2025, whereas in 2021, this number was only 16, marking a more than tenfold increase over four years. With so many companies, apart from leading enterprises like Unitree Robotics, Agibot, Fourier Intelligence, and LimX Dynamics, do the remaining over 100 enterprises have no value in being remembered and seen? Of course not. If the academic faction is the prestigious school grown from the academic system, and the industrial capital faction is the powerful master nurtured by the wealthy families, then the countless grassroots and independent practitioners constitute the majority of the embodied AI world. Facing the prestigious and wealthy, the grassroots faction has its own unique survival tricks.
The First Trick: Acting Cute The adorable "Little Snub-nose" at the sports meet, from High Torque Robotics, is a typical small and micro startup. The founder, Zhang Junhui, comes from a programmer background and once founded an advertising company. In 2020, he crossed over into the robotics field, learning about motors and motion control from scratch. In June 2022, High Torque Robotics was established, relying entirely on its own funds and operating revenue for development. With limited funding and unable to afford humanoid robots that are 1.7 meters tall and cost hundreds of thousands of RMB each, High Torque Robotics took the route of acting cute, turning robots into collectible robot kits. These little guys cannot become spectacles, but they are more suitable for schools and families with limited budgets, providing educational, companionship, and emotional value. Because the technical system for cute-type robots is relatively simple, large models can call cloud-based foundations, and the development costs for the robot's small models and motion models are easy to bear, this track is also quite crowded. This year's World Robot Conference featured a large number of similar competing products. Among them, the "Big Head" robot dog, which has been mass-produced and delivered by Vbot, is priced at 9,988 RMB, with 6,540 units pre-sold and 23,700 kilometers of user accompanying mileage, making it a highlight in the pet robot track. The company also plans to invest its accumulated data and funds into humanoid robots in the future. And similar little humanoid robots include the "Bridge" from Zeroth Robotics, a company established only a year and a half ago. It can be said that the competition in the companionship track is quite fierce. Besides many similar small-factory competitors, this business is also within the firing range of medium and large companies. In fact, this trend has already emerged in the market. For example, Songyan Dynamics, which has an academic background, is also focused on making robots smaller and cheaper. In the same half-marathon, Songyan Dynamics' N2, running fully autonomously without a pacer, won the second place. Standing 1.18 meters tall, it is considered the most independent robot, with a starting price of only 39,900 RMB. Compared to the purely emotional "Little Snub-nose," the N2 is obviously smarter, and it also received thousands of orders after the race, becoming another humanoid robot enterprise with sales exceeding a thousand units after Unitree. Of course, whether they are small and micro enterprises or medium-sized companies, all realize that no matter how cute the bionic route is, or how good they are at acting cute, they cannot cross the threshold of being able to do actual work.
The Second Trick: Being a Specialist Instead of a Generalist Small and micro startups have relatively weak endowments in capital, talent, and data, and cannot rely on brute force like large enterprises. Therefore, learning a unique skill to solve a specific problem can also help them gain a foothold in the industry. This is like Duan Yu, who mastered the "Lingbo Weibu" (Lingbo Microstep) to the extreme and could still roam the martial arts world. In the embodied AI world, there is such a company that has practiced "lightness skill" to the extreme, and that is DroidUp. At the first Humanoid Robot Half-Marathon in Yizhuang, Beijing in 2025, DroidUp's "Xingzhe II" was the only robot in the field to finish the race without battery swaps. While other robots had to stop for battery swaps, it could run the whole way relying on a unique skill: tendon bionic drive technology. Unlike others that install motors directly inside the joints, DroidUp's full-stack bionic tendon cable drive simulates the human muscle and skeletal system, using tendons and distributed motors for coordinated drive, concentrating the motors at the hips. This makes the joint flexibility close to that of real humans and achieves mechanism lightweighting; with a height of 1.7m, it weighs only about 38kg. The advantage of this is light legs and low energy consumption. With this core barrier, "Xingzhe II" is 1.7 meters tall and weighs 38 kilograms, while similar full-size robots generally weigh over 60 kilograms. If a robot is too heavy, not only will its battery life shorten rapidly, but its safety will also be hard to guarantee. Once it falls, it takes some effort to lift it up, and it is even more dangerous if it falls on and injures someone. In July WAIC 2025, "Xingzhe III" was released, with further weight reduction in the leg tendons, making it over 20% lighter than its peers and reducing energy consumption by 50%. It also added waist degrees of freedom, so even if it falls, it won't hurt people. The safety brought by technical barriers has become a differentiated advantage, making it very suitable for deployment in high-traffic commercial venues or home healthcare scenarios for the elderly. There are many similar specialists. For example, since making the main body of a robot is too expensive, a group of startups simply focus on making the robot's organs, and they still live quite well. For example, SynapX, the most aggressive new face at this year's World Robot Conference, focuses on bionic cable-driven dexterous hands. Its grasping compliance is close to that of human hands, enabling it to collect data while working, solving the efficiency bottleneck of teleoperation data collection. This new face, established only on January 4, 2026, has no backing from big companies, but its debut was stunning; its booth drew large crowds of visitors. Taking a demand that is so small that giants don't want to tackle it, yet truly exists, to the extreme is one of the survival rules of the grassroots world.
The Third Trick: Being a Senior Worker Currently, leading embodied AI companies all face a paradox: they must continuously invest heavily in R&D to possibly maintain their competitiveness in the first tier. However, humanoid robots lack landing scenarios, leading to continuous losses for the companies, and no one knows when they can break even. But grassroots enterprises often do not have such troubles. After all, small enterprises never intended to pursue lofty goals like AGI (Artificial General Intelligence) robots from the very beginning. As long as they can do a good job and make money in specific scenarios, they can live quite comfortably. The total market size of these specific single scenarios is too small; for large companies, the input-output ratio of doing it is not cost-effective, while it is more than enough to support a small company with a team of dozens of people. ZeroPower Robotics in Hefei focuses on cleaning. Established less than a year ago, its products are already out working. They have been deployed in over 20 locations, including Hefei Xinqiao Airport and the OLE supermarket at MixC Shenzhen, mainly doing space organization and cleaning, capable of stable 24-hour operation. Compared to those star robot companies queuing up for IPO and not knowing which year they will turn a profit, these small enterprises that went out to work and earn money early might have avoided 20 years of detours.
The Fourth Trick: Open Source For a small robot company, to grow from small to large, besides transforming itself into a large product company like giant enterprises, it can also scale up the ecosystem. Through open source and openness, turning its products into public resources, absorbing more developers and small teams, and making itself the infrastructure of the industry, it can also possess commercial value. RoboParty is a typical representative. In 2025, founder Huang Yi established RoboParty in Zhangjiang, Shanghai, becoming the youngest humanoid robot CEO in the country at the time. In Huang Yi's view, the advantage of the new generation of entrepreneurs lies in having no path dependence, but they must find the right ecological niche in the entrepreneurial game with limited resources. RoboParty's ecological niche is open source, aiming to be the Android in the robotics field. The team found that many companies in the industry sell "black boxes." Users don't know how the hardware is designed after buying it, and when problems arise, they can't tell whether it's the hardware or the code's fault. The threshold for secondary development is extremely high. If there could be open-source hardware, turning the black box into a white box, developers who want to adjust the main body can modify the drawings themselves, replace the parts, and find suppliers on their own. Aiming for "low cost, high performance, and open-source sharing," RoboParty launched the world's first fully open-source bipedal humanoid robot: the ROBOTO Origin (RPO). This robot is 120 centimeters tall and has 23 degrees of freedom. It can be fully assembled by purchasing parts on Taobao (China’s e-commerce platform) and prototyping with JLCPCB (PCB manufacturing service), allowing developers and enthusiasts to build it by hand. Then, relying on open-source training and deployment code, basic functions like walking and running can be realized. Moreover, the full set of hardware costs less than 7,000 US dollars, and the selling price is only 15,000 RMB, while the prices of similar products are generally between 100,000 and 200,000 RMB. As of July 2026, the project has surpassed 2,000 stars on GitHub, and the global developer community has exceeded 5,000 members. It has also been selected as one of the top ten open-source projects by EAI OpenSource and won the Best Embodied AI Open Source Contribution Award of the Yuanqi Award. At the same time, RoboParty also developed the supporting Party OS R&D foundation, precipitating tedious processes like data collection, training frameworks, and policy deployment into an out-of-the-box system. Open source, however, does not mean working for free purely out of passion. Currently, the company has begun to run through the open-source commercial closed loop, with orders coming from global developers, universities, research institutions, and model companies that need real machines for demos. This model has also been recognized by the capital market. Established for a year and a half, it has completed six rounds of financing, raising nearly RMB 500 million. An investor commented: "The team chose the full-stack open-source path, demonstrating a strong geek spirit. In the long term, it is expected to use high-cost-performance, full-stack open-source humanoid robots as a carrier to build an industrial closed loop of whole machines + joint modules + developer ecosystem." It should be noted that the open-source business model relies on a sufficiently large ecosystem. For developers within the open-source community, performance remains the top priority. This requires robot manufacturers to continuously invest and maintain advanced capabilities to keep their attractiveness to the ecosystem. At the same time, large companies may also participate in reverse; for example, Unitree, Agibot, and Fourier Intelligence are all stepping up their open-source efforts, which may also divert some developers. Open source is becoming the standard for the entire industry, no longer the exclusive weapon of small companies. Precisely because of this, the value of this group of small players is even more worthy of being seen. They are the infrastructure that promotes the explosive innovation of the entire industry, and also the technological equalizers that make closed-source companies open.
The Era of the Grassroots It can be found that leading manufacturers are constantly creating visual spectacles of robots, cool fights, and magical dances. However, those robots whose names we cannot recall might be the embodied AI closest to ordinary people, holding the promise of truly helping us with housework and picking up children. And they often come from some unknown small team. Only a world gathered by the grassroots will not be monopolized, will not become rigid, and will not lose its imagination and vitality. This is the significance of this group of people being seen.