China’s most well-funded humanoid startup has flown relatively under the radar compared to its contemporaries.
Fellow Chinese firms like Unitree Robotics and Agibot tend to steal the spotlight with their Kung Fu-fighting humanoids, while Western media focus on perceived leaders in the humanoid race like Boston Dynamics, Figure AI, and Tesla.
Beijing’s Galbot just reached a new level of global visibility thanks to a viral visit to one of its futuristic storefronts during Trump’s summit with President Xi. If US-China relations seriously improve, Galbot’s human-free outlets could one day become as common American cities as McDonald’s is within the People’s Republic.
Barely three years old, the startup is already pursuing an IPO on the Hong Kong Stock Exchange (HKEX) after securing a series of record-breaking funding rounds. It would join the much more established humanoid players, Shenzhen’s UBTECH Robotics and Dobot on the exchange. With its unusually strong national and industrial capital base, Galbot has aggressive plans as it positions itself as the company that can make humanoids boring, useful, and everywhere.
Galbot’s Origin
Galbot was legally established in May 2023 as Beijing General Robot Co. Ltd. Its co-founders are Wang He and Yao Yao TengZhou. Wang studied at Tsinghua University and got a PhD from Standford in 2021 under the influential computer scientist Leonidas Guibas. He returned to China as an assistant professor at Peking University, directing its Embodied Perception and Interaction Lab (EPIC).
Wang serves as Galbot’s chief technology officer. Yao, who serves as president, has a master’s from Beihang University and has worked on commercial products for ABB in Shanghai. Galbot’s core team also includes talent recruited from giants like Huawei, Baidu, and Microsoft.
With around $2.5 million in initial registered capital, the team designed its first humanoid immediately and quickly moved to pilot deployments. Its launch came a few months before the Chinese government set a national goal to mass manufacture humanoids by 2025 and to lead the market by 2027.
Massive Funding
In Beijing, Galbot’s contemporaries include players like:
- RobotEra, which is scaling the deployment of its sword-slinging humanoids into China’s logistics sector,
- NOETIX Robotics, known for its budget pint-sized humanoids and its ultra-realistic androids,
- Booster Robotics, whose T1 robots pioneered autonomous humanoid soccer,
- Linkerbot, which says it’s scaling toward 10,000 dexterous robot hands per month,
- and the Beijing Humanoid Robot Innovation Center, also known as X-Humanoid, a government-backed lab that’s commercializing its robots known for their speed and agility.
They all have substantial financial backing and ambitious roadmaps, but Galbot likely has the largest war chest. The 3-year-old startup has already raised around $930 million to date. Its Series D round of $362 million, reported in March 2026, pushed its valuation to nearly $3 billion.
Its funding is on par with Western competitors like Google-backed Apptronik. Galbot’s money goes further in China, which has the world’s most robust humanoid supply chain, though Apptronik has a big advantage with access to DeepMind’s frontier physical AI models.
Galbot G1
Galbot’s flagship humanoid is called the G1, short for Generation 1, not to be confused with fellow Chinese humanoid company Unitree’s viral G1.
Galbot’s G1 is a wheeled dual-arm robot that stands about 173 cm (5 feet 8 inches) and weighs 80 kg (176 lbs). Its 7-degree-of-freedom (DoF) arms support a combined payload of 10 kg (22 lbs). With a vertical reach up to 2.4 meters (7 feet 10 inches), the G1 can reach higher than most humans standing on the ground. Its powertrain is driven by an NVIDIA Jetson AGX Orin computer that delivers 275 TOPS.
While Hangzhou-based Unitree has shipped way more units, Galbot’s deployments so far look much more practical than theatrical. The robots, priced around $100,000 each, have reportedly been installed in at least 10 pharmacies around Beijing.
According to Galbot, they run autonomously, handling thousands of different medications daily. They perceive their surroundings with stereo cameras in the head, depth cameras in the wrists, and more cameras and sensors in the torso. A 3D LiDAR sensor in its base enables real-time 3D mapping and navigation. Its ultrasonic sensors provide close-range obstacle detection, helping the G1 sense nearby people and objects that may be too close or awkwardly positioned for LiDAR alone.
The humanoid communicates through voice interaction and its touchscreen display. It integrates with grippers and modular dexterous robot hands designed for manipulation tasks. Galbot’s five-fingered end effectors are described as having 12 DoF each and tactile sensing for contact-force feedback.
Galbot S1 for Industry
In 2026, Galbot introduced its heavy-dusty humanoid for industry, the S1.
Its dual-arm payload of up to 50 kg (110 lbs) is 5x higher than the G1’s. According to Galbot, the industrial humanoid has quick-swap batteries that last 8 hours per charge for near-continuous operation.
S1 robots have reportedly been deployed at production lines for CATL, a Galbot investor and one of the world’s largest battery manufacturers.
Robot-Run Retail
Galbot’s robots have won contests at spectacles like Beijing’s World Humanoid Games and have entertained at public events like the 2026 Spring Festival Gala.
But what’s probably gotten the most attention is its futuristic retail stores like the one in Beijing that went viral during a visit from Fox News reporter Brett Baier. Galbot store kiosks are now operational is over 30 cities throughout China and serve thousands of customers daily.
Astrabrain AI
Galbot’s robots run on its self-developed AI system called Astrabrain.
The startup says it combines simulation training with real-world dasta to teach its humanoid to execute on human instructions while working long shifts in structured commercial environments. Galbot uses massive amounts of synthetic training data created in simulation. The data feeds into the system’s Grasp vision-language-action (VLA) model that helps robots recognize many different objects and figure out how to pick them up.
Frontier Research
Galbot is also investing heavily into frontier research for physical AI.
The startup worked with researchers from Beijing’s Tsinghua and Peking universities to develop a humanoid robot tennis player. The customized Unitree G1 pulled off a shocking rally against a human opponent using their new framework called LATENT, which is short for Learning Athletic Human Tennis Skills from Imperfect Motion Data.
According to the team, the system teaches robots athletic movements using incomplete examples of human motion. Instead of recording full tennis matches, the researchers fed the humanoid’s artificial intelligence fragments of motion data captured from tracking devices worn by amateur tennis players.
Though incomplete, the researchers said the data still contained valuable information. Using machine learning, the system corrected and recombined the pieces into complete actions that the robot could learn in simulation.
Engineers attached a tennis racket to the humanoid’s right arm using a 3D-printed adapter. They also added reflective markets to track its position and movements during rallies.
In real-world experiments, the robot returned incoming tennis balls with a more than 90 percent success rate for forehand shots and 78 percent for backhands. While the skills pale in comparison to a trained human player, the team said future systems could rapidly improve through robot-vs-robot training.
Looking Ahead
As Galbot aims to have more than 10,000 robots working in real retail stores and factories within three years, another fast-rising Beijing firm says it’s doubling its monthly output.
Across five factories in China, Linkerbot is already manufacturing around 5,000 dexterous robotic hands per month. The startup, reportedly seeking a $6 billion valuation, may seem like it came from out of nowhere but its rise says a lot about where physical automation could be headed.