Galbot's most recent disclosed valuation is approximately $3B, set in December 2025 when the company announced a funding round of over $300M. In March 2026, Galbot closed an additional round of RMB 2.5 billion, roughly $350M, bringing total lifetime funding to approximately $1.15B.

The March 2026 round included the China Integrated Circuit Industry Investment Fund, Sinopec, CITIC-affiliated entities, SAIC Motor's financial arm, Bank of China, Kunpeng Capital, E-Town International Investment and Development, and Wuxi Venture Capital Group, alongside existing shareholders.

Before the December 2025 round, Galbot had raised an angel round totaling RMB 700M and a strategic round of RMB 500M in 2024, followed by a RMB 1.1B round led by CATL in June 2025 that pushed the company past a $1B valuation. Investors across rounds have included institutions from China, Singapore, and the Middle East.

Galbot builds general-purpose embodied AI robots for physical labor in commercial and industrial environments. Rather than a single-purpose machine that repeats one fixed motion, the product is a mobile, dual-arm robot that can be retasked across workflows such as shelf replenishment, parts handling, pharmacy delivery, and inventory management.

The G1 is a wheeled robot standing roughly 1,730mm tall with a torso that lifts and extends to reach from near floor level up to about 2.4 meters. It has two arms with a combined payload of 10kg, a 360-degree omnidirectional wheel base for quiet in-place rotation, and a sensor suite including binocular head cameras, wrist depth cameras, 3D LiDAR, six-axis force sensors, microphone arrays, and a touchscreen. Onboard compute is an NVIDIA AGX Orin running Galbot's embodied AI system, Embosar, which fuses visual, auditory, and spatial signals to interpret scenes, recognize objects, plan actions, and execute manipulation tasks without fixed coordinates.

In a typical deployment, an operator maps the site using Galbot's tools, defines the workflow, such as picking an SKU from a shelf, delivering it to a station, or replenishing a row, and loads a product catalog or item library. The robot then navigates autonomously, identifies the correct item using vision and force feedback, grasps or suction-picks it, and completes the transfer. Staff step in for exceptions, and the robot returns to charge on its own.

The S1 is the heavier-duty system for industrial and logistics tasks requiring up to 30kg per dual-arm operation. It runs for eight hours on hot-swappable batteries for zero-downtime shift operations and reaches speeds of 1.5 meters per second. Compared with the G1's shelf-picking and service use cases, the S1 is aimed at intralogistics tasks such as line-side material handling and repetitive factory transfer.

Galbot also sells a software layer alongside the robots. Its developer platform has three tiers: a foundational SDK with simulation assets and low-level hardware access, pre-packaged scenario kits for workflows like greeting, sorting, and inspection, and end-to-end solutions for retail, industrial handling, and education. Customers can use it as a turnkey deployment or access raw APIs for custom development.

Galbot sells primarily B2B into enterprise manufacturing, retail, healthcare, and logistics, with a growing channel through systems integrators and solution partners that deploy on top of the platform. The core transaction is a hardware system sale combined with an annual software and support subscription, structured per deployed robot.

The developer platform makes the recurring layer explicit: enterprise-tier support runs at the equivalent of roughly $27,000 per robot per year and includes dedicated technical advisors, onsite deployment assistance, proactive health checks, and SLA-backed remote debugging. The per-machine annual fee creates a compounding revenue base as the deployed fleet grows, rather than a one-time capex event.

Near-term economics are service-heavy because deployments require workflow configuration, simulation validation, safety review, and ongoing field operations support. Galbot's packaged scenario kits are the mechanism for compressing that customization cost over time: once a retail replenishment or pharmacy delivery workflow is productized into a reusable kit, the marginal cost of the next deployment in that vertical falls. Margin improvement depends on standardizing the middle layer of the stack.

The self-reinforcing dynamic is straightforward. Each deployment generates real-world task data across manipulation, navigation, and human interaction. That data feeds Galbot's embodied AI models, including its specialized GraspVLA, TrackVLA, and GroceryVLA architectures, which improve task reliability. Higher reliability reduces integration cost, shortens sales cycles, and makes the ROI case easier for new customers.

The retail capsule format is a distinct monetization experiment. Galbot operates fully autonomous convenience store capsules across 20-plus cities, some exceeding 500 orders per day. Whether Galbot captures revenue through robot sales to operators, revenue-sharing arrangements, or direct store operations is not publicly disclosed, but the format indicates that the company is willing to own the operating layer, not just sell hardware into it.

Galbot competes in a market that is consolidating around scale, data, and industrialization. Its wheeled dual-arm architecture is better suited to structured commercial environments than some general-purpose humanoid designs, but competitors are advancing across adjacent form factors, price points, and software stacks.

AGIBOT is the most direct threat in China. It claims to have rolled out its 10,000th robot in March 2026, sells a lineup spanning humanoid, wheeled, and multi-form robots, and has launched a zero-code deployment layer called Genie Studio Agent that lowers adoption friction for non-technical customers. Relative to Galbot, AGIBOT's advantage is breadth across hardware and deployment tooling, while Galbot is more narrowly focused on high-value mobile manipulation use cases instead of covering multiple form factors at once.

UBTECH is the strongest incumbent-style Chinese rival in industrial humanoids, with Walker S robots deployed in automotive and electronics manufacturing and a multi-robot coordination platform called BrainNet. Unitree applies pressure from a different angle, with in-house actuators, open-source tooling, and entry-level humanoid pricing around $13,500. That price point sets a reference for the category and draws developers and smaller commercial buyers that might otherwise test Galbot.

Figure is the clearest Western analogue on the AI-first humanoid axis. By April 2026, Figure's BotQ facility had delivered 350-plus Figure 03 units and improved production rate from one robot per day to one per hour, while its Helix 02 VLA stack extends to full-body autonomy. The competitive risk to Galbot is less geographic overlap than Figure becoming the category reference point for full-stack humanoid scaling among multinational customers.

Agility Robotics competes on deployed industrial workflows, with Digit at Toyota Motor Manufacturing Canada and Mercado Libre and bundled with Arc cloud management software. Apptronik's Apollo targets warehouses and manufacturing with hot-swappable batteries and a Jabil manufacturing partnership that gives it a credible path to industrialization. Both are useful benchmarks for turning humanoid robots into operating infrastructure rather than advanced hardware projects, the same transition Galbot is attempting through its Bosch-linked Bowintec joint venture.

1X, Sunday Robotics, and The Bot Company are not direct competitors to Galbot in industrial manufacturing today, but they represent a longer-term threat from a different direction. If home-first systems like 1X's NEO, priced at $20,000 upfront or $499 per month, generate large volumes of human-centric manipulation and interaction data from consumer deployments, those models could later move into commercial service and light industrial roles.

Sunday's zero-robot-data approach, training exclusively on human demonstrations captured via gloves distributed to 1,000-plus households, and The Bot Company's autonomy-first stack built around end-to-end neural networks both rest on the view that the richest training data comes from unstructured home environments rather than structured factories. If that thesis proves out, Galbot's enterprise-first data flywheel could be narrower than it appears today.

Galbot's expansion logic is to use its current industrial and commercial deployments as a proving ground for an embodied AI platform that could extend to workflows requiring mobile manipulation in human-shared environments. The near-term vectors are vertical depth and geographic reach, while the longer-term vector is expanding control of the stack.

The G1 and S1 already cover light-duty commercial manipulation and heavy-duty industrial handling, but the larger product opportunity is the software layer above the hardware. Galbot's packaged scenario kits for greeting, sorting, and inspection suggest a reusable application layer that can shorten deployment time and reduce customization cost per new vertical. If those kits mature, Galbot could move beyond robot OEM toward a software and workflow layer with recurring revenue on top of hardware placements.

Healthcare is a notable vertical expansion area. Galbot already has a pharmacy robot in service in Beijing and a joint research relationship with Xuanwu Hospital. The progression from pharmacy fulfillment and internal hospital delivery to ward assistance, eldercare support, and rehabilitation adjacency is gradual, but each step can use the same core platform. Healthcare also has 24/7 operating requirements and persistent labor intensity, which can make ROI easier to justify than in consumer settings.

Galbot's reference customers in automotive and electronics manufacturing, CATL, Bosch, Toyota, BAIC, SAIC, are globally active companies with plants across Southeast Asia, Europe, and Mexico. Following those customers into international facilities is a more credible near-term path than selling generalized service robots abroad without channel support. The Bowintec joint venture with Bosch's investment platform is the clearest mechanism for that expansion, combining Galbot's embodied AI with Bosch-linked manufacturing know-how and industrial distribution to target loading, assembly, and quality inspection in complex factory settings outside China.

Retail is a second expansion vector that appears more developed than it might seem. The Galbot Store capsule format, 100-plus locations across 20-plus cities, some exceeding 500 orders per day, is both a robot deployment and a test of a fully autonomous retail operating format. If the unit economics hold, Galbot could extend from selling to retailers into providing the operating layer for unmanned micro-retail, instant fulfillment, and chain pharmacy dispensing, capturing value from store operations as well as robot sales.

Galbot's appearances at CES 2026 and Hannover Messe 2026 indicate investment in international visibility at a point when enterprise buyers are beginning to shortlist embodied AI vendors. The most credible path is cross-border industrial expansion through existing multinational accounts rather than consumer internationalization.

Manufacturers facing labor scarcity and reshoring pressures in Southeast Asia, Mexico, and Central Europe are structurally similar to Galbot's current Chinese customers. A deployment record with globally recognized brands like Bosch and Toyota is likely to be the most useful reference in those sales processes.

Model narrowness: Galbot's three specialized VLA models, GraspVLA, TrackVLA, and GroceryVLA, are a near-term strength but also a structural constraint because an intelligence stack built around narrow, task-specific architectures is likely to face rising integration costs and slower expansion into new verticals as the category shifts toward more general-purpose embodied AI that can handle novel environments without per-workflow retraining.

Compute dependence: Galbot's robots run on NVIDIA AGX Orin and its training pipeline relies on NVIDIA's physical AI stack, so any disruption to NVIDIA supply, export controls affecting Chinese access to advanced compute, or a shift in the ecosystem toward a competing compute substrate could raise Galbot's hardware costs, slow model iteration, and erode the differentiation of its embodied AI platform when scaling production matters most.

Integration intensity: Because each new vertical deployment still requires substantial customer-specific workflow configuration, safety review, and ongoing field operations support, Galbot's growth in pilots and order backlog may not translate cleanly into scalable margins unless the packaged scenario kit layer matures fast enough to reduce the services burden per site before larger rivals with deeper global service footprints use their own deployment infrastructure to undercut Galbot on total cost of ownership.