On April 10th, Zhiyuan Kuto held its first media communication conference at the Shanghai Robot Valley.
Zhiyuan Robot recently made a significant adjustment to its business structure, spinning off its quadruped robot division to establish an independent subsidiary named Zhiyuan Kuto. Data indicates that Zhiyuan Kuto's revenue for the first quarter of this year has already surpassed its total revenue for the entire previous year. More notably, medium-sized quadruped robots have been sold to the point where "there is no inventory left in the warehouse." In preparation for a partner conference scheduled for April 17th, the company has even recalled some demonstration units.
However, facing a quadruped robot market that already includes major players like Unitree and Deep Robotics, Qiu Heng, Director, COO, and President of Marketing and Service at Zhiyuan Kuto, stated: "There is essentially no competition in making quadruped robots."
In Qiu Heng's view, most robotic dogs on the market "lack a brain." A true quadruped robot should possess capabilities such as adaptive dynamic balance, comprehensive environmental understanding across all scenarios, autonomous task planning, real-time anomaly handling, and natural language interaction. It should be able to perform complex tasks independently, rather than relying on pre-set commands or manual remote control.
"The difference between a dog and a human isn't two legs versus four legs; it's the difference in intelligence level," Qiu Heng said. Zhiyuan Kuto's goal is to create quadruped robot "persons," leveraging the full-stack AGI and embodied intelligence technology from the Zhiyuan Group to build all-terrain intelligent mobility and operation platforms.
"I believe there is essentially no competition in making quadruped robots because the products currently made by companies in the market are essentially brainless. Therefore, if we can create a quadruped robot that truly possesses human-like cognitive abilities, customers will certainly be willing to pay for it," Qiu Heng remarked candidly. This is the fundamental logic behind Zhiyuan Kuto's spin-off from the parent group.
The Zhiyuan Group originally had four product lines: Yuanzheng, Lingxi, Jingling (all humanoid), and Kuto (quadruped). Qiu Heng explained, "The quadruped business itself is developing very rapidly. This represents a huge future business; it shouldn't be kept in the 'shadow' of a giant. It is itself a giant and needs to be brought out for everyone to see, enabling it to grow at scale."
Qiu Heng revealed that Zhiyuan Kuto is the most important subsidiary of the Zhiyuan Group. Humanoid robots will be developed by Zhiyuan, while quadruped robots will be handled by Kuto.
It is understood that Zhiyuan's humanoid robots focus primarily on stationary operations with mobility as a secondary function, emphasizing precise upper-limb manipulation. Quadruped robots, conversely, prioritize mobility with arm-based operations as secondary, focusing on all-terrain traversal capability. Both share a highly reusable technological foundation, including underlying hardware, power systems, top-level embodied intelligence models, and access to vast amounts of real-world scenario data, all leveraging the group's shared technological assets.
However, Zhiyuan Kuto will also develop its own differentiated enhancements, such as VLN within its "brain." VLN refers to Vision Language Navigation. Since quadruped robots require navigation—a capability less critical for the initial humanoid models—it needs to be specifically strengthened.
Qiu Heng believes Zhiyuan is the domestic robotics company with the largest volume of data. The training data for perception, understanding, and planning models is almost identical for both humanoid and quadruped robots. "We analyzed it. Using Genie Studio (Zhiyuan's one-stop development platform) to train embodied operation models for quadruped robots can reduce the workload by 50% to 70% compared to not using it," he stated.
Zhiyuan Kuto has chosen a clear business-to-business commercial path. "We have no plans to enter the B2C market for the time being," Qiu Heng stated clearly. "Our goal is not to create a pet, but a productivity tool." He explained that home environments are highly diverse, with significant variations between households. The current priority is to succeed in sectors like power grid, coal mining, security, and logistics before considering residential applications.
The company has set clear business targets: achieving annual shipments of 300,000 units and revenue of 10 billion yuan by 2030, with overseas revenue accounting for over 35%. Qiu Heng disclosed that the revenue target for 2026 is 500 million yuan, with overseas orders already exceeding 40% in the first quarter.
In terms of product portfolio, Zhiyuan Kuto already covers small, medium, and large-sized quadruped robots. Small models are for education and entertainment, medium models for security patrols and industrial inspections, and large models for logistics handling—capable of carrying loads up to 50 kg with an endurance of 5.5 hours. The medium-sized quadruped robot is currently the primary focus for key applications: it can carry 25 kg, supporting equipment like cameras, loudspeakers, communication devices, or inspection robotic arms.
Furthermore, Qiu Heng previewed the next-generation flagship model, the D2 Max, positioned as the industry's first all-terrain Level 3 autonomous driving quadruped robot. It will feature enhanced terrain adaptation, precise foot placement, dynamic obstacle avoidance, autonomous navigation, and autonomous carrying capabilities, enabling stable execution of inspection, transport, loading, and unloading tasks in environments like warehouses, campuses, utility tunnels, and construction sites.
Regarding the much-discussed topic of generalization capability, Qiu Heng provided a clear roadmap. "The first step is achieving generalization between different tasks within a single industry, realizing industry-specific AGI. Only after that will we consider generalization for specific single tasks or multiple tasks across partially related industries."
He admitted that customers typically wouldn't use a power grid quadruped robot in an oil field. The most important task currently is to master one industry thoroughly. The capability for multi-task transfer within a single industry is expected to be achievable upon the launch of the D2 Max.
On the subject of price competition, Qiu Heng believes the market does not currently show intense competition, and it is not a core issue at this stage. "If a customer buys one 'dumb' dog, they won't want to buy a second. But if it's a very intelligent quadruped robot, after others buy one, they'll want to buy more. That is real business."
As for financing and IPO plans, Qiu Heng's response was straightforward: "We will not seek financing for the time being; we do not wish to dilute our equity. This market is very, very large and worth a great deal of money; we shouldn't sell out now. Will there be an IPO someday? We look forward to that possibility."
Comments