The money keeps flooding into Chinese robotics—sometimes faster than the companies can deploy it.
Galaxea Dynamics, a Beijing-based embodied-AI startup barely fifteen months removed from stealth mode, closed a 1 billion yuan Series B on February 11, bringing its valuation to roughly 10 billion yuan, or about $1.4 billion at current exchange rates. Jinding Capital, an industrial investor with deep manufacturing ties, led the round. BAIC Group Industrial Investment and Hone Capital joined in, while existing backers including Cathay Capital, Meituan's Dragonball venture arm, and Capital Today doubled down.
It's the latest data point in what's become a full-blown financing sprint across China's humanoid robotics ecosystem—one that has some industry watchers comparing the moment to the early autonomous-vehicle boom, with all the attendant promises and question marks.
Galaxea has now raised close to 3 billion yuan since its pre-A round in November 2024. That's nearly $430 million in just over a year, a pace that suggests either extraordinary conviction from investors or a fear of missing the next big platform shift. Probably both.
The Industrial Money Arrives
The composition of this latest round tells its own story. Jinding Capital isn't a traditional venture shop—it's the kind of investor that builds factories and manages supply chains. BAIC Group, the automotive conglomerate's investment arm, brings manufacturing heft. According to reports from Caixin and other Chinese business outlets, the cap table also includes Hillhouse Ventures, Qianhai Ark, and Loyal Valley Capital, though Galaxea hasn't released an official breakdown.
This marks a shift. Early-stage robotics rounds in China have traditionally been dominated by tech-focused VCs—Ant Group, Meituan, Tencent. Now the industrial players are showing up, checkbooks open, which usually means one of two things: either the technology has matured enough for serious deployment conversations, or there's pressure to nationalize critical supply chains. In China's current policy environment, it's likely both.
Galaxea's funding timeline is almost comically compressed. Pre-A in November 2024 brought in over 200 million yuan from GL Ventures and Ant Group, with miHoYo (the gaming giant behind Genshin Impact) taking a side bet. Then came a rapid-fire A-series: 300 million yuan in an Ant-led A1, another 300-plus million in A2/A3 rounds led by Cathay, and finally more than $100 million across A4 and A5 tranches in July 2025, with Meituan and Capital Today participating. By mid-2025, the company was valued around $700 million. The latest B more than doubles that figure in yuan terms—though exchange-rate fluctuations complicate clean comparisons.
Even in a market accustomed to aggressive growth-stage financing, that's a lot of capital absorbed in a short window. The obvious question: can the company deploy it faster than expectations inflate?
What Galaxea Actually Sells
Strip away the embodied-intelligence branding, and Galaxea is building two things: hardware robots and the software stack to animate them.
On the hardware side, the company's developer storefront lists wheeled dual-arm humanoids—the R1 Pro, R1, and R1 Lite—alongside a lightweight desktop robotic arm dubbed the A1 X. At CES 2026 in Las Vegas, it previewed the DEXO, a 17-degree-of-freedom dexterous hand that the company claims achieves 0.1-newton tactile resolution. (Whether that holds up in real-world grasping tasks remains to be seen; robotics labs have been chasing human-level tactile sensing for decades.)
Pricing starts at 199,000 yuan—roughly $28,000—for entry-level configurations and climbs to 459,900 yuan for fully loaded R1 units equipped with five-finger hands. There's also a bundled "VLA All-in-One Unit" shipping at $49,999, which pairs the R1 Lite with compute hardware and teleoperation gear. That's still expensive for most industrial buyers accustomed to traditional automation ROI math, but it's a fraction of what Boston Dynamics or ABB charge for comparable systems.
The software piece is where Galaxea makes its real platform bet. The company open-sourced its G0 vision-language-action (VLA) model in September 2025, then released an upgraded G0 Plus in January 2026. VLA models—neural networks that map visual inputs and natural-language commands directly to robot actions—are the current obsession in embodied AI, though the jury is still out on whether they'll scale beyond lab demos.
Galaxea's R1 series has shown up in academic projects at Stanford and MIT. Physical Intelligence, the well-funded San Francisco startup, cited the R1 Lite as a data-collection platform for its π 0.5 work. Chinese media report "thousands" of units ordered by universities and industrial customers globally, though the company hasn't disclosed verified shipment numbers or anything resembling a cohort analysis. In robotics, it's easy to rack up pilot orders. Converting those into recurring revenue is the hard part.
A Crowded Field Gets More Crowded
Galaxea's Series B lands in the middle of what can only be described as a unicorn wave in Chinese embodied AI.
Unitree, the Hangzhou-based maker of quadrupeds and humanoids, hit a $1.7 billion valuation in a June 2025 Series C backed by Geely, Ant, Tencent, and Alibaba—basically every major tech and industrial player in China. LimX Dynamics announced a $200 million Series B on February 2, just days before Galaxea's close. EngineAI raised roughly 1 billion yuan across pre-A and A1 rounds in mid-2025. These are big checks chasing a sector that, by most measures, is still pre-revenue at scale.
Several of these companies have restructured into joint-stock entities—Galaxea converted from a limited-liability company to a foreign-invested joint-stock corporation in late January 2026—a corporate maneuver often associated with pre-IPO preparation in China. Unitree is rumored to be eyeing a $7 billion STAR Market listing later this year, according to people familiar with the discussions.
Caixin, the respected Chinese business outlet, framed the moment as a cluster of billion-yuan rounds converging on robotics hardware, with IPO chatter intensifying. There's a certain momentum to it, the kind that makes investors nervous about sitting out the next big thing—even if unit economics remain murky and commercialization timelines keep sliding right.
The Scaling-Law Question
Galaxea CEO Gao Jiyang—a Tsinghua graduate who earned his PhD at USC, worked on prediction systems at Waymo, and later led AI infrastructure at Momenta, the Chinese autonomous-driving startup—has said little publicly about margins or unit economics. That's not unusual for early-stage founders, but it's also not particularly reassuring given the capital intensity of robotics hardware.
CFO Luo Tianqi told Chinese media in February that the company is focused on "scaling-law economics" in VLA training and building out a developer ecosystem around hardware, data, and models. That phrasing—"scaling-law economics"—is borrowed from large-language-model research, where more compute and more data reliably yield better performance. Whether that intuition transfers cleanly to embodied AI is an open question. Robot training is messier: the data is harder to collect, simulations don't always transfer to the real world, and physical deployment introduces failure modes that don't exist in software.
The B-round proceeds will fund continued R&D on the G0 model family and support industrial pilots in the first half of 2026, according to coverage in Gasgoo and other outlets. Where exactly those pilots will land—logistics, manufacturing, services—remains unspecified. Perhaps deliberately so.
What Comes Next
For now, Galaxea holds a slot in a crowded field of Chinese robotics unicorns chasing the same verticals with strikingly similar hardware philosophies: wheeled mobile manipulators, dual arms, dexterous hands, open software stacks. The bet on open models and modular platforms is strategically sound—it lowers barriers for developers and accelerates the data flywheel—but it's not unique. Unitree, LimX, and several other players are making nearly identical moves.
Differentiation will come down to execution. Who ships fastest. Who builds the stickiest developer ecosystem. Who proves out use cases beyond the research lab and convinces industrial buyers that humanoid robots can deliver ROI within a reasonable payback period.
The funding window, for now, remains wide open. Whether it stays that way—or whether consolidation arrives sooner than the founders expect—depends on how quickly these companies can translate research traction into recurring revenue. The clock is ticking, and the capital markets have a way of becoming impatient when the next earnings call arrives and the robots are still mostly in pilot mode.
