China’s Only AI Giant-Backed Embodied Intelligence Startup Hits $14B Valuation
X Square Robot, a Shenzhen-based embodied AI company, has closed a RMB 1 billion (~$137 million) A++ round, bringing its total funding to over RMB 3 billion (~$410M) across nine rounds in just two years.
The round was co-led by ByteDance, Sequoia China, and Shenzhen Capital Group, cementing X² Robot’s status as the only embodied intelligence startup in China backed by all three internet giants:
- ByteDance (content + traffic)
- Alibaba (cloud + AI infrastructure)
- Meituan (local services + real-world operational data)
This rare tripartite endorsement isn’t just financial—it’s strategic:
X² Robot now has direct access to China’s largest consumer touchpoints, cloud compute, and physical-world deployment channels.
At a post-money valuation of RMB 10 billion (~$1.37B), the company has grown 50x from its seed valuation—making it one of the fastest-scaling deep-tech startups in China.
The Core Thesis: One Model to Rule All Robots
Founded in December 2023 by Wang Qian, a former quant fund founder and USC robotics PhD, X² Robot rejected the industry’s dominant modular architecture (perception → planning → control) in favor of a single, end-to-end Vision-Language-Action (VLA) model.
“Traditional robots play a ‘telephone game’—errors compound at every handoff,” Wang explains.
“We want robots to see, think, and act in one neural flow—like humans.”
This “unified brain” approach aims to eliminate task-specific tuning, enabling true generalization across environments—from factories to homes.
Critically, X² Robot rejects synthetic data for complex manipulation tasks.
“For contact-rich interactions—like hand dexterity—real-world data is non-negotiable,” Wang states.
The company has built proprietary data collection rigs to gather high-fidelity physical interaction data, prioritizing quality over quantity.
Commercial Strategy: From B2B to Home—But Not via Factory PR
Unlike peers chasing automotive assembly lines, X² Robot argues that industrial “pilot projects” are often PR stunts that avoid the hardest AI challenges.
Instead, the company targets complex, long-horizon service tasks:
- Phase 1: B2B service scenarios (hospitality, retail, logistics)
- Phase 2: Consumer home robots (3–4 year horizon)
Wang believes service and domestic environments—with their unstructured chaos, social nuance, and multi-step reasoning—are the true testbeds for general-purpose intelligence.
To support this, X² Robot is vertically integrated:
- In-house development of hardware bodies
- Full-stack AI model training
- Real-world data flywheel from deployments
This “soft-hard一体化” model ensures tight coupling between brain and body—essential for reliable physical reasoning.
Global Context: East vs. West in the “Brain Race”
The global race for the robot “OS” has split into two philosophies:
| Region | Approach | Key Players | Strength | Risk |
|---|---|---|---|---|
| China (East) | Commercial-first | X² Robot, Agibot | Real revenue, industrial validation, supply chain | Narrower technical scope |
| U.S. (West) | Platform-first | Physical Intelligence, Skild AI | Pure software, cross-hardware generality, massive simulation | Distant commercialization |
- Physical Intelligence (PI): Building a universal brain for any robot body; backed by OpenAI, Bezos, NVIDIA
- Skild AI: Training models on “1,000 years” of simulated experience; valued at $14B
- X² Robot: Betting that real-world data + end-to-end learning will outperform simulation-heavy approaches in messy, human-centric environments
As one investor put it:
“The West dreams of a robot OS. China is building the first app store—with paying customers.”
Why the Giants Are All In
The backing of ByteDance, Alibaba, and Meituan is unprecedented—and strategic:
- ByteDance: Provides massive video datasets for pre-training and user engagement channels for consumer adoption
- Alibaba Cloud: Offers scalable AI training infrastructure and integration with Tmall Genie ecosystem
- Meituan: Delivers real-world service scenarios—food delivery, hotel check-in, warehouse logistics—for live model refinement
This trifecta gives X² Robot what no competitor has:
A closed loop from data → model → deployment → feedback—in the world’s largest digital-physical economy.
Investment Takeaway: The End-to-End Wager
X² Robot represents a high-conviction bet on a single idea:
General-purpose physical intelligence requires unified neural architectures—not patched modules.
Risks are real:
- End-to-end models are computationally expensive and hard to debug
- Industrial buyers still prefer modular, explainable systems
- Consumer adoption remains 3–4 years out
But if X² Robot succeeds, it won’t just sell robots.
It will own the intelligence layer for China’s next-generation service automation—across hundreds of millions of daily interactions.
In a market flooded with “me-too” humanoids,
X² Robot is betting that the future belongs not to the best walker—
but to the first truly thinking machine.
And with $1.37B in fresh capital and the backing of China’s tech trinity,
it has the resources to prove it.
