Why RMB 2 Billion is Flowing into Embodied AI Data and Simulation Infrastructure
Guanglun Intelligent has secured RMB 2 billion in cumulative funding as investors recognize that scalable simulation and data evaluation not just models are the strategic choke points for physical AI deployment.
The flow of capital in the robotics sector is shifting from hardware prototypes to the invisible architecture that makes those prototypes useful. Over a two week period in June 2026, Beijing based physical AI infrastructure company Guanglun Intelligent completed back to back funding rounds totaling RMB 1 billion. This brings the company's cumulative financing to RMB 2 billion and solidifies its position as the first unicorn dedicated exclusively to embodied data and simulation.
The investment syndicate reveals the strategic priority being placed on this layer of the technology stack. Backers include state guided funds such as the Zhongguancun Science City Fund, Sichuan Development Sci Tech Innovation Fund, and Shandong Development Sci Tech Innovation Venture Capital, alongside industrial capital from New Hope Group and Giant Network [1]. The composition of this investor base indicates that data and simulation are no longer viewed merely as software tools. They are being treated as critical national infrastructure for the physical AI transition.
The Strategic Bottleneck for Physical AI
The next bottleneck for physical AI is not chips or foundation models. The primary constraint is data, simulation, and evaluation infrastructure.
In the early stages of the humanoid robotics boom, capital flowed toward companies proving that robots could walk, balance, and execute basic tasks. As the industry attempts to move from demonstration to deployment, the fundamental problem has changed. A robot operating in a controlled laboratory environment requires only a fraction of the training data needed for a robot navigating the unpredictable variables of a factory floor or a logistics center.
Guanglun Intelligent CEO Yang Haibo notes that the scaling law for embodied data is now widely recognized among leading development teams [1]. Relying solely on physical robots to collect real world data is too slow, too expensive, and too dangerous to support the volume of training required for general purpose physical AI. The solution is synthetic data generation and high fidelity simulation, where millions of iterations can occur safely and rapidly before a physical robot ever attempts a task.
The Shift from Verification to Scale
The market demand for this infrastructure is expanding rapidly. In previous years, robotics teams required hundreds or thousands of hours of data to verify basic capabilities. In 2026, the demand from world model developers and robotics companies has surged to hundreds of thousands or millions of hours of data [1].
This volume requirement fundamentally changes the business model for data providers. High quality embodied data is becoming an infrastructure asset with high resale value. A single comprehensive dataset capturing a specific industrial environment or complex manipulation task can serve multiple clients across different stages of model development. Guanglun Intelligent reports that its first quarter new orders for 2026 reached RMB 550 million, reflecting this transition from experimental data collection to industrial scale data consumption [1].
The Real Constraint on Deployment
Simulation software and physics engines have historically been dominated by Western technology companies. As physical AI becomes a central pillar of industrial competitiveness, reliance on external simulation platforms presents a strategic vulnerability for Chinese robotics developers.
The ability to create a safe, repeatable, and scalable virtual environment is just as critical as the semiconductor hardware running the computations. Without a robust domestic simulation and evaluation ecosystem, the transition from prototype to widespread industrial deployment will stall.
The Bigger Signal
The massive capital injection into Guanglun Intelligent signals that the physical AI sector is maturing. Investors are moving up the value chain, recognizing that the companies supplying the "picks and shovels" for the AI robotics gold rush may capture the most durable value.
As the industry pushes toward commercialization, the focus is shifting from the physical capabilities of the robot to the virtual environments where the robot learns. The organizations that control the training data, the simulation physics, and the standardized evaluation metrics will ultimately dictate the pace at which physical AI integrates into the global economy.
Sources:
