Beijing Galaxy General Robot
Last reviewed
Sources
35 citations
Review status
Source-backed
Revision
v6 · 4,695 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
35 citations
Review status
Source-backed
Revision
v6 · 4,695 words
Add missing citations, update stale details, or suggest a clearer explanation.
Beijing Galaxy General Robot Co., Ltd. (Chinese: 北京银河通用机器人股份有限公司), commonly known as Galbot or Galaxy General, is a Chinese robotics and embodied AI company headquartered in Haidian District, Beijing. Founded in May 2023 by Wang He (王鹤), a Peking University professor, and co-founder Yao Tengzhou, Galbot develops general-purpose wheeled humanoid robots and the end-to-end vision-language-action models that control them.[1][2] By March 2026 the company had raised more than $800 million across multiple rounds at a valuation of roughly $3 billion, making it one of the most highly valued unlisted humanoid robot startups in China.[3][4]
Galbot's flagship product is the Galbot G1, a wheeled semi-humanoid mobile manipulator deployed commercially in unmanned pharmacies and convenience stores across China through a partnership with Meituan, and on automotive and battery production lines with Bosch and CATL. The company gained nationwide recognition after winning the gold medal at the 2025 World Humanoid Robot Games and appearing at the 2026 CCTV Spring Festival Gala.[5]
| Beijing Galaxy General Robot | |
|---|---|
| Company information | |
| Chinese name | 银河通用机器人 (Yinhe Tongyong Jiqiren) |
| Trade name | Galbot |
| Founded | May 2023 |
| Founder(s) | Wang He (王鹤), Yao Tengzhou |
| Headquarters | Haidian District, Beijing, China |
| Industry | Robotics, Embodied AI |
| Products | Galbot G1, Galbot G1 Premium, Galbot S1 |
| Total funding | Over $800 million (as of March 2026) |
| Valuation | ~$3 billion (as of December 2025) |
| Website | galbot.com |
Galbot (Beijing Galaxy General Robot) is a Chinese embodied-AI company that builds general-purpose robots and the foundation models that operate them. The trade name combines "Galaxy" and "robot," mirroring the Chinese name 银河通用机器人 (Yinhe Tongyong Jiqiren), literally "Galaxy General Robot," which reflects the company's stated goal of building general-purpose robots powered by embodied artificial intelligence.[2]
Rather than chasing visually dramatic walking robots, Galbot prioritizes wheeled mobile manipulators that emphasize stability, long battery life, and practical commercial deployment. Its robots are powered by a suite of proprietary vision-language-action models (GraspVLA, GroceryVLA, NavFoM, TrackVLA) trained largely in simulation and transferred to physical hardware. The company positions itself around doing repetitive, low-glamour work reliably. As founder Wang He put it in a July 2025 interview, "I strongly advise against hyping up embodied artificial general intelligence (AGI). We're likely five to ten years away from robots that can do everything."[6]
Beijing Galaxy General Robot was established in May 2023 by Wang He and Yao Tengzhou. From its inception, the company pursued a strategy centered on combining academic research in dexterous manipulation and computer vision with rapid commercialization.[2]
The founding team drew heavily from elite Chinese academic institutions and global technology companies. Core members came from Huawei's "Genius Youth" program, Baidu, Microsoft, and ABB Group, giving Galbot strengths in self-developed core components (motors and reducers), end-to-end large language model development, and mass-production engineering.[2]
In June 2024, Galbot completed an angel round of approximately 700 million CNY (roughly $96 million), led by Meituan at a round valuation of about 3.5 billion CNY (~$485 million). Other investors included Qiming Venture Partners, Lanchi Ventures, IDG Capital, and a Hong Kong government fund through the Hong Kong Investment Corporation (HKIC). At the time, this was reported as the largest angel round ever raised by a company in the embodied intelligence sector in China.[7][8]
The partnership with Meituan was both financial and strategic. As one of China's largest online services platforms, Meituan provided Galbot with real-world deployment scenarios in pharmacies and convenience stores, accelerating the transition from laboratory prototypes to commercial products. A further strategic round of about 500 million CNY followed in November 2024.
Galbot unveiled its first commercial product, the Galbot G1, in June 2024. The G1 is a wheeled semi-humanoid robot that combines a dual-arm upper body with a 360-degree omnidirectional wheeled chassis, prioritizing stability, long battery life, and practical deployment over bipedal locomotion. By late 2024, the G1 began application testing for 24-hour unmanned operations in Meituan-partnered pharmacies across Beijing.[9]
In August 2025, Galbot unveiled the Galbot G1 Premium (also marketed as the G1 Pro) at the World Robotics Conference (WRC) in Beijing. The G1 Premium adopted the NVIDIA Jetson AGX Thor computing platform, making Galbot one of the first companies in the world to integrate the Jetson Thor processor into a commercially deployed humanoid robot.[10]
In January 2026, Galbot introduced the Galbot S1, a heavy-duty wheeled humanoid built for industrial work. Despite its humanoid upper body, the S1 uses a wheeled mobile base rather than walking legs. It features a 50 kg continuous dual-arm payload and autonomous battery-swap capability for around-the-clock operation.[11]
Galbot's fundraising trajectory accelerated through 2025 and into 2026. The company completed its Series A in June 2025 (approximately 1.1 billion CNY, or $153 million, led by CATL), its Series B in December 2025 (over $300 million at a $3 billion valuation, led by China Mobile's Chain Development Fund), and an additional round of 2.5 billion CNY (~$362 million) in March 2026. Investors in the 2026 round included Sinopec, the China Integrated Circuit Industry Investment Fund (the "Big Fund"), CITIC-affiliated entities, SAIC Motor's financial arm, and Bank of China. This marked the first investment by a national-level strategic fund in a Chinese embodied intelligence company.[3][12][13]
In preparation for a public listing, the entity converted from a limited liability company to a joint-stock company (北京银河通用机器人股份有限公司) on November 27, 2025.[1] As of late 2025, Bloomberg reported that Galbot had selected Citic Securities, Huatai Securities, and UBS to work on a potential initial public offering on the Hong Kong Stock Exchange, with a target valuation of $3 billion to $4 billion. Galbot has publicly characterized the listing reports as premature, and no formal Hong Kong Exchange filing had appeared as of mid-2026.[14][13]
Wang He (王鹤), born in 1992, serves as Galbot's founder and Chief Technology Officer. His educational background spans three of the most prestigious institutions in China and the United States:
| Milestone | Details |
|---|---|
| High school | Beijing No. 11 School (admitted to Tsinghua University through the physics competition) |
| Bachelor's degree | Electronic Engineering, Tsinghua University (2010-2014) |
| Ph.D. | Stanford University (2021), advised by Professor Leonidas J. Guibas (member of three U.S. National Academies) |
| Academic position | Tenure-track Assistant Professor and Ph.D. supervisor, Center on Frontiers of Computing Studies (CFCS), Peking University |
| Research role | Research scientist, Beijing Institute for General Artificial Intelligence (BIGAI) |
| Lab director | PKU-Galbot Joint Lab of Embodied AI; BAAI Center of Embodied AI |
Wang He's research focuses on embodied perception, dexterous manipulation, and 6D pose estimation. His pioneering work on category-level 6D pose estimation (NOCS) received the 2022 World Artificial Intelligence Conference Youth Outstanding Paper Award. He has published over 50 papers in top-tier venues including CVPR, ICCV, NeurIPS, ICLR, ICRA, and TRO.[15][16]
In 2025, Wang He was selected for Fortune China's "40 Under 40 Business Leaders" list.[17] On February 9, 2026, Chinese President Xi Jinping inspected the Galbot G1 robot during a pre-Spring Festival visit to Beijing, where Wang He was present to brief the president.[18]
Yao Tengzhou co-founded Galbot alongside Wang He. He holds a master's degree from Beihang University's (Beijing University of Aeronautics and Astronautics) Institute of Robotics, where he was mentored by Professor Wang Tianmiao, a prominent robotics scholar who serves as honorary director of the Beihang Robot Institute and president of the Zhongguancun Intelligent Friends Institute. Before co-founding Galbot, Yao Tengzhou worked at ABB Group's Shanghai Robotics R&D Center, accumulating years of experience in the development of industrial and service robots.[2]
Galbot's funding trajectory has been one of the most aggressive in the global humanoid robot industry. The following table summarizes the major publicly disclosed rounds.
| Round | Date | Amount | Key investors |
|---|---|---|---|
| Angel round | June 2024 | Meituan, Qiming Venture Partners, Lanchi Ventures, IDG Capital, Hong Kong Investment Corporation (HKIC) | |
| Strategic round | November 2024 | ~500M CNY | Lanchi-backed and other institutional investors |
| Series A | June 2025 | CATL (lead), Boyuan Capital (Bosch), multiple institutional investors | |
| Series B | December 2025 | ~$300M | China Mobile Chain Development Fund (lead), Beijing Robotics Industry Fund, GGV Capital, CICC Capital, Chinese Academy of Sciences Fund |
| Additional round | March 2026 | Sinopec, China Integrated Circuit Industry Investment Fund, CITIC, SAIC Motor, Bank of China, CATL, CMG Media Fund |
The cumulative funding exceeds $800 million, with a post-money valuation of approximately $3 billion as of December 2025. The March 2026 round made Galbot one of the highest-valued unlisted firms in China's humanoid robotics sector.[3][12][13]
Galbot employs a multi-tier product strategy, offering robots tailored to different market segments. All of its current robots use wheeled mobile bases rather than bipedal legs.
| Product | Type | Introduced | Target market | Key specification | Approximate price |
|---|---|---|---|---|---|
| Galbot G1 | Wheeled semi-humanoid | 2024 | Retail, pharmacy, household | 5 kg/arm payload, 10-hour battery | ~699,900 CNY |
| Galbot G1 Premium | Wheeled semi-humanoid | 2025 | Light industrial, advanced retail | NVIDIA Jetson Thor compute | Not disclosed |
| Galbot S1 | Heavy-duty wheeled humanoid | 2026 | Heavy manufacturing, logistics | 50 kg dual-arm payload | Not disclosed |
The Galbot G1 is Galbot's first and most widely deployed product. Standing 173 cm tall with a 190 cm arm span, the G1 uses a 360-degree omnidirectional wheeled chassis rather than bipedal legs. This design provides up to 10 hours of battery life and substantially reduces manufacturing cost compared with a bipedal leg system. The G1 achieves a 95 to 97 percent grasp success rate across diverse materials and object types, and up to 99.5 percent in structured pharmacy and retail settings.[19]
The G1 is commercially available on JD.com at a listed price of 699,900 CNY. Galbot has stated plans to drive per-unit cost down toward approximately 50,000 CNY (~$7,130) as production scales up.
The Galbot G1 Premium (also marketed as the G1 Pro) retains the G1's wheeled form factor but upgrades the computing platform to NVIDIA Jetson AGX Thor (Blackwell architecture). The Jetson AGX Thor module delivers up to 2,070 FP4 TFLOPS of AI compute, a large increase over the Jetson Orin used in the base G1. Galbot describes the robot as capable of fully autonomous operation without teleoperation.[10] Announcing the upgrade, Wang He said, "Our G1 Premium, now running on NVIDIA Jetson Thor, delivers significant gains in speed and real-time reasoning, enabling our proprietary VLA models to achieve enhanced real-world performance."[10]
The Galbot S1 is Galbot's heavy-duty wheeled humanoid, unveiled in January 2026 for industrial applications. Its 50 kg continuous dual-arm payload far exceeds the 20 to 25 kg range typical of most humanoid robots. The S1 features a dual-battery quick-swap design that allows it to autonomously replace its own batteries, enabling 24/7 operation. It has been deployed on CATL battery manufacturing production lines and operates using pure visual perception for 360-degree obstacle avoidance without QR codes or external markers.[11]
Galbot's robots are powered by a family of proprietary end-to-end embodied AI models. The company describes GraspVLA as among the world's first end-to-end foundation models for robotic grasping.
| Model | Function | Description |
|---|---|---|
| GraspVLA | Manipulation | A vision-language-action model (arXiv 2505.03233) trained on billions of simulated movements. Galbot describes it as a foundational model for real-time closed-loop grasp generation. It supports zero-shot learning, enabling the robot to grasp rigid and deformable objects under varying lighting and backgrounds without object-specific training. |
| GroceryVLA | Retail operations | Specialized model for dense retail environments. Handles SKU identification, shelf scanning, and inventory management across thousands of product types. Enables manipulation of fragile jars and soft packaging without reprogramming. |
| NavFoM | Navigation | Foundation model for autonomous navigation using pure vision-based perception. Robots navigate environments without QR codes, pre-mapped routes, or external markers. |
| TrackVLA | Object tracking | Handles real-time tracking of objects and people in dynamic environments (arXiv 2505.23189). Requires only a single natural language instruction to guide a robot through complex environments while navigating obstacles and interacting with humans. |
These models combine elements of large language models, computer vision, and robotic action planning into unified architectures. On the G1 Premium's Jetson Thor platform, multiple models run concurrently using Blackwell Multi-Instance GPU (MIG) technology, enabling simultaneous visual processing, manipulation planning, obstacle tracking, and voice interaction within real-time latency requirements.[10][20]
Galbot researchers developed the DexGraspNet dataset to address one of the core challenges in robotics: teaching robots to grasp a wide variety of objects with dexterous hands. The work was carried out using NVIDIA simulation tools, with version 2.0 built on NVIDIA Isaac Sim, a robotics simulation platform built on the Omniverse 3D graphics engine.
| Version | Objects | Grasps | Categories | Real-world success rate |
|---|---|---|---|---|
| DexGraspNet 1.0 | 5,355 | 1.32 million (ShadowHand) | 133+ | N/A |
| DexGraspNet 2.0 | Expanded (cluttered scenes) | Expanded | Expanded | 90.70% |
DexGraspNet 1.0 (arXiv 2210.02697) contains two orders of magnitude more grasps than the previous state-of-the-art Deep Differentiable Grasp dataset. The Galbot research team also proposed UniDexGrasp++, an object-independent approach for learning generalized dexterous grasping strategies from real point cloud observations and proprioceptive information, using a methodology called Geometry-Aware Iterative Generalist-Specialist Learning (GiGSL).[21]
Galbot employs a Sim2Real methodology that relies on NVIDIA Isaac Sim and Isaac Lab for building large-scale training datasets. The company's proprietary simulation dataset spans grasping, navigation, deformable object manipulation, and multi-step task planning at a scale of billions of robotic action data points, which Galbot uses to train VLA models that transfer to physical robots with high fidelity.[20] Wang He has described why this matters: "The lack of real-world robotic action data has long been a bottleneck for the robotics industry."[22]
The company also co-developed the OpenWBT_Isaac platform with Tsinghua University and the Shanghai Qi Zhi Institute. This simulation platform enables whole-body teleoperation of humanoid robots and serves as a tool for collecting high-quality human demonstration data for training new manipulation skills.[10]
In June 2026, Galbot released AstraBrain-WBC 0.5, a whole-body motion-control foundation model with about 80.4 million parameters that the company reported achieves a 92.58 percent zero-shot success rate on whole-body control tasks.[23]
Meituan has been Galbot's most important strategic partner since the company's founding. Beyond leading the angel funding round, Meituan provides real-world deployment scenarios for Galbot's robots in the retail and healthcare sectors.
Beginning in late 2024, Galbot and Meituan developed a 24-hour unmanned pharmacy concept. The Galbot G1 operates autonomously in approximately 50 to 70 square meter pharmacies, performing the full cycle of retail operations: scanning shelves across approximately 5,000 SKUs, autonomously restocking when items run low, locating and picking items for customer orders, and delivering products directly to customers or delivery couriers.[9][24]
In March 2026, China's first fully robot-operated pharmacy went into service in Beijing's Haidian District, with the Galbot robot handling all medication sorting, retrieval, and dispensing tasks without human assistance. As of early 2026, more than 10 pharmacies across Beijing were using Galbot units for 24-hour unmanned operations, with plans to scale to over 100 smart retail locations by the end of 2026.[24]
Galbot has also deployed its robots in unmanned convenience stores branded as "Galaxy Space Capsule" (银河太空舱) locations. Each capsule is a compact 9-square-meter cabin housing a single Galbot robot that provides fully autonomous service, from greeting customers and processing payments to grabbing items and delivering them. Despite the small footprint, each store offers over 300 items including beverages, snacks, cultural products, and daily-use medicines with dedicated refrigerated and frozen storage.[25]
Galbot has described the rollout as a "ten cities, hundred stores" (十城百店) program. As of early 2026, more than 100 Galaxy Space Capsule locations had launched across roughly ten cities, including Beijing, Shanghai, Suzhou, Hangzhou, and Chengdu.[25]
In June 2025, Galbot signed a three-way cooperation agreement with Bosch China (through its investment platform Boyuan Capital) to form the Boyin Innovation Alliance, a joint venture focused on deploying embodied AI robots in automotive and high-precision manufacturing. The joint venture also signed a memorandum of understanding with UAES (United Automotive Electronic Systems) to establish RoboFab, a joint laboratory dedicated to developing automotive-specific expertise in embodied AI.[26]
The partnership follows a "global design, local production" strategy, positioning Galbot to enter international markets including Europe, North America, and Southeast Asia through Bosch's extensive existing relationships with automotive OEMs. Dr. Ingo Ramesohl, Managing Partner of Bosch Ventures, stated that "embodied AI holds transformative potential to redefine manufacturing processes."[26] In May 2026, the affiliated joint venture (博银合创, Bowintec) raised a Pre-A round of nearly 300 million CNY.[27]
CATL, the world's largest battery manufacturer, has been both a major investor and deployment partner. CATL led Galbot's Series A round in June 2025 and continued investing in subsequent rounds. The Galbot S1 is deployed on CATL's core production lines for heavy material handling tasks, marking one of the first real-world deployments of a humanoid robot in battery manufacturing.[11] In June 2026, the two companies signed a broader global strategic cooperation agreement to expand S1 deployment and jointly develop after-market service standards for embodied robots.[28]
Galbot has established partnerships with several major automotive companies for manufacturing floor deployments. Confirmed partners include Toyota, Hyundai, BAIC Group, SAIC Group, and Zeekr. In April 2026, Galbot also partnered with automotive interiors supplier Yanfeng to deploy its embodied-AI software on auto-interior production lines.[29] As of March 2026, the company had secured cumulative orders totaling several thousand units from these and other industrial clients.[3]
Galbot won the gold medal in the Robot Skills Competition at the 2025 World Humanoid Robot Games, held in Beijing, China. The robot competed with 100 percent autonomy against 22 other teams, many of which relied on remote teleoperation.
In the Pharmaceutical Sorting Challenge, the Galbot robot autonomously scanned six tall shelves, identified nine specific medications from hundreds of items, and delivered them in 10 minutes and 22 seconds. The final score of 336 points exceeded the second-place competitor by 160 points. The victory demonstrated the practical applicability of Galbot's embodied AI models in real-world-like settings.[30]
Galbot was selected as the exclusive official robot platform for the 2nd International Olympiad in Artificial Intelligence (IOAI), held in Beijing in August 2025. This marked the first time in history that an international AI competition incorporated real-robot practice. More than 300 high school students from 63 countries participated in the "Future Factory" challenge, which integrated simulation-phase development with real-world deployment on Galbot robots.[31]
On January 25, 2026, China Media Group (CMG) officially named Galbot the "Designated Embodied Large Model Robot" for the 2026 Spring Festival Gala, China's most-watched television broadcast (comparable to the Super Bowl in the United States). Galbot appeared alongside Unitree Robotics, Noetix, and MagicLab in the gala on February 16, 2026.[32]
In the holiday short film "The Night I Remember Most," the Galbot robot demonstrated practical domestic skills including voice interaction, folding clothes, picking up objects, cracking walnuts, skewering sausages for grilling, and retrieving items from shelves. Following the broadcast, robot searches on JD.com surged 300 percent within two hours, customer service inquiries rose 460 percent, and order volume jumped 150 percent. New orders came from over 100 cities across China.[33]
On February 9, 2026, Chinese President Xi Jinping inspected the Galbot G1 robot during a pre-Spring Festival visit to Beijing. Founder Wang He participated in the meeting and expressed the company's readiness to serve as a vehicle for China's participation in global technology competition during the 15th Five-Year Plan period.[18]
Galbot operates in a rapidly growing global market for humanoid robots. In 2025, approximately 13,000 to 18,000 humanoid robots were shipped worldwide, a sharp increase over the previous year. Chinese companies accounted for the large majority of all humanoid robots sold globally, and China controls much of the humanoid robot supply chain.[34]
The following table compares Galbot with its primary competitors.
| Company | Country | Key products | Focus area | Approximate valuation | Notable partnerships |
|---|---|---|---|---|---|
| Galbot | China | G1, G1 Premium, S1 | Retail, pharmacy, heavy industry | ~$3B | Meituan, Bosch, CATL, Toyota |
| Unitree Robotics | China | H1, G1, H2 | Consumer, industrial | ~$3B (Series C) | Broad consumer market |
| Agibot | China | A2, A2 Max, Genie | Automotive manufacturing | Undisclosed | SAIC Motor, NIO |
| Figure AI | United States | Figure 02, Figure 03 | General-purpose, manufacturing | ~$39.5B | BMW, NVIDIA |
| Tesla | United States | Optimus Gen 3 | Consumer, factory | Part of Tesla | Tesla factories |
| UBTech | China | Walker S, Walker C | Education, service | Public (HKEX: 9880) | Multiple sectors |
| Fourier Intelligence | China | GR-2 | Rehabilitation, healthcare | Undisclosed | Healthcare institutions |
Galbot differentiates itself through several strategic choices. First, its wheeled-base product line prioritizes practical commercial deployment over visually dramatic bipedal locomotion, offering tangible advantages in stability, runtime, and cost for flat-floor environments. Second, the company's deep partnerships with Meituan, Bosch, CATL, and major automakers provide real commercial deployment channels that many competitors lack. Third, Galbot's full-stack approach pairs its own robots with proprietary VLA foundation models trained at simulation scale.[34]
However, Galbot faces intense competition. Unitree has achieved the largest shipment volumes in the consumer segment with aggressively priced robots. Agibot leads in automotive manufacturing deployments. Figure AI commands the highest valuation in the sector globally. And Tesla's Optimus program benefits from Tesla's massive manufacturing infrastructure and brand recognition.
Galbot's rapid growth has been supported by China's national policy emphasis on humanoid robotics. The Chinese government has identified humanoid robots as a strategic industry, and multiple government-affiliated funds have invested in Galbot's rounds, including the Beijing Robotics Industry Fund, the Chinese Academy of Sciences Fund, and the China Integrated Circuit Industry Investment Fund. In March 2026, China released its first national humanoid robot standard system to provide a framework for long-term industry growth.[35]
Beijing, Galbot's home base, has been particularly active in fostering the robotics industry. The city has established dedicated robotics industrial parks and innovation zones in Haidian District, where Galbot is headquartered, and the Beijing Robotics Industry Fund has been a recurring investor in Galbot's funding rounds.
Galbot's near-term plans include:
The company's rapid funding trajectory, early commercial deployments, and strong government support position it among the leading humanoid robotics companies in China in the race to bring general-purpose robots to mass-market adoption.
Galbot is a Chinese company that builds robots with two arms and a roll-around wheeled base instead of legs. Think of a robot shopkeeper: it can read the shelves in a tiny store or pharmacy, grab the thing you want, and hand it to you, all by itself, 24 hours a day. The robots learn most of their skills inside a video-game-like computer simulation before they ever touch a real object, which is how they get good at picking up thousands of different items. Galbot started in 2023, is run by a young Peking University professor named Wang He, and has become one of China's most valuable robot startups.