In a typical deployment, a wheeled robot with two arms reaches into a shelf, pulls a product, and drops it into a pickup bin. Outside, a delivery rider scans a code and leaves. The store never closes, and the clerk never gets tired.
Key takeaways
- Galaxy General (银河通用), founded May 2023 in Beijing, builds the Galbot G1, a wheeled, dual-arm, foldable humanoid (人形机器人).
- On June 23, 2025 it closed an ¥1.1 billion (≈ US$155M / HK$1.2B) round led by CATL (宁德时代), called by Xinhua the largest single embodied-AI round in China to date.
- Its smart-retail setup has Galbot working 24/7 in a roughly 50 sqm unmanned store handling 5,000 product types across 6,000 shelves and 10,000-plus items; a new store can be set up in about a day.
- Galbot is already running in nearly ten Beijing stores and seven Beijing unmanned pharmacies, picking medicine and handing off to delivery riders.
- The company has also formed a joint venture with Bosch (博世) capital for industrial scenarios.
A robot that stocks the shelves
Galaxy General (北京银河通用机器人有限公司) was founded in May 2023 and is led by founder and CTO Wang He (王鹤), a Tsinghua undergraduate and Stanford PhD who is also an assistant professor at Peking University. Its first product, the Galbot G1, launched in June 2024: a wheeled chassis with two arms and a foldable torso, built to reach from low shelves to over two meters high.
Rather than walk on two legs, Galbot rolls. In a smart-retail deployment, it spends its shift picking, restocking, and packing in an unmanned space. According to the company and state technology-bureau coverage, a single Galbot can manage around 5,000 product types across 6,000 shelves and more than 10,000 items, and a new store can be brought online in roughly a day.
Why a wheel, not two feet
The wheeled, foldable design is a deliberate bet. Flat, indoor retail and pharmacy environments do not need bipedal walking; they need reach, stability, and endurance. Galbot's approach trades the romance of a walking humanoid for the practicality of a machine that can work a full shift without tiring.
Under the hood, Galaxy General trains its models on synthetic data. It released GraspVLA, an end-to-end embodied grasping model, and later GroceryVLA and TrackVLA for retail and navigation. According to financial and technology press, Galbot can grasp a wide range of objects at about 95 percent success using only synthetic-data pretraining — a claim about data strategy more than about showmanship.
The money behind the machine
On June 23, 2025, Galaxy General announced an ¥1.1 billion (≈ US$155M / HK$1.2B) new funding round led by CATL (宁德时代) via its strategic investment arm and Puquan Capital, with participation from the SDIC's Guokai Innovation, the Beijing Robotics Industry Fund, and GGV Capital. Xinhua described it as the largest single financing in China's embodied-intelligence field to date. Including earlier rounds, the company said its cumulative funding exceeded ¥2.4 billion (≈ US$335M / HK$2.6B).
CATL's involvement is strategic, not just financial: the battery giant is signaling that embodied robots belong in industrial and logistics workflows, not only in labs. Days earlier, on June 17, Galaxy General said it formed a joint venture with Bosch (博世) capital to pursue industrial scenarios.
From retail to factory
The near-term proof point is scale of deployment. Galaxy General says Galbot is already running in nearly ten Beijing stores under normalized operation, and in seven Beijing unmanned pharmacies where it picks medicine around the clock and hands off to delivery riders. The company has spoken of expanding to a hundred outlets across Beijing, Shanghai, and Shenzhen.
According to Li An, chief scientist at BrainNet (脑机网), the Galbot bet is that the first profitable humanoids will be the ones doing boring, repeatable indoor jobs — not the ones that look most like us. Li An's comments reflect his own views and do not represent NeuroAI's editorial position or an endorsement by BrainNet (脑机网).
Honest limitations
- The 5,000 / 6,000 / 10,000-plus figures and the "deploy in one day" claim come from Galaxy General and government event coverage; they describe design capacity, not independently audited per-store throughput.
- The "largest single round" label is as reported by Xinhua and financial press; rankings can shift as the field moves quickly.
- Unit economics are unclear. One report put a full "Galaxy Pod" site cost in the millions of yuan, and a per-unit Galbot price near ¥700,000 was quoted by an employee — single-source, so treat as indicative only.
- Pharmacy and store counts ("nearly ten," "seven") are company-stated and change fast; verify before treating them as current.
- Retail is a forgiving, structured environment; the harder industrial claims tied to the Bosch joint venture are early.
What readers can do now
- If you run retail, pilot a Galbot-style setup in one 24/7, SKU-dense site such as a pharmacy or convenience store, and measure labor substitution against amortized hardware cost.
- If you invest, track Galaxy General's store count and per-store order volume quarter over quarter — scale, not the funding headline, is the real test.
- If you build, note the wheeled-and-foldable choice: matching morphology to the environment beats forcing bipedal humanoids into flat, indoor work.
