A Chinese robotics startup has raised nearly $150 million to develop a new approach to teaching robots how to understand and navigate the physical world, using first-person video recordings of humans rather than traditional robot demonstrations. Infiforce, backed by state-owned investment platforms and venture capital firms, plans to use the funding to expand its embodied AI (EAI) systems, scale its data infrastructure, and deploy more robots across industrial and commercial settings in China.

What Makes Infiforce's Approach Different?

Most robot training relies on either watching demonstrations from an outside camera angle or having humans remotely control robots to generate training data. Infiforce is taking a different path by collecting what it calls "Ego" data, or first-person recordings of people interacting with the physical world. This approach captures video, movement patterns, spatial relationships, and environmental feedback from a human perspective, then uses that information to train its AI models.

The company argues this method provides more training data at lower cost than collecting everything directly from robots. Rather than needing hundreds of robots running thousands of hours to generate training examples, Infiforce can use handheld recording devices, first-person cameras, and remote robot operation to build a larger dataset more efficiently. The goal is to create a feedback loop where robot deployments generate additional real-world data that improves future models.

How Is Infiforce Testing Its Technology?

The company has published several benchmark results from its research. Its AtomVLA model achieved a 97% success rate on the Libero robot-learning benchmark, while its HiMem-WAM model reached 97.7% on the same test. Infiforce's third-generation AIM world model scored 93.1% on RoboTwin 2.0, and its SAM3D method reached 99.1% on Libero. These results come from company-reported research rather than independent third-party validation.

The company is also testing whether the same underlying intelligence can transfer across different robot bodies. Infiforce's hardware lineup includes the AstroDroid wheeled humanoid, UltraDroid general-purpose robot, Little Atom bipedal robot, and specialized Force systems. This "one brain, multiple forms" strategy could allow a single AI system to control different robot types without retraining from scratch.

Where Are These Robots Being Deployed?

Infiforce reports that its robots and systems are being tested or commercialized across more than 30 Chinese cities and more than 100 different scenarios. The company is active in manufacturing, logistics, warehousing, and commercial services. It is also working with CRRC High-Tech on embodied intelligence for infrastructure applications and has participated in developing a proposed national specification for crowdsourced embodied-intelligence data collection and management.

• Manufacturing: Robots are being deployed in production facilities to handle tasks that require physical manipulation and environmental understanding.
• Logistics and Warehousing: Systems are being tested to automate sorting, moving, and organizing goods in distribution centers.
• Commercial Services: Robots are being used in retail, hospitality, and other service environments where they interact with customers and navigate complex spaces.
• Infrastructure Applications: In partnership with CRRC High-Tech, Infiforce is exploring how embodied AI can support infrastructure inspection and maintenance.

What Does Infiforce Plan to Do With This Funding?

The $150 million raised across Series A and Series A+ rounds will support three main areas. First, the company will continue developing its AtomBrain embodied-intelligence system and causal world models, which are designed to provide a common intelligence layer across different robot forms. Second, Infiforce will expand its DataGrid AI infrastructure, which combines data collection, processing, and hardware to support multiple data-gathering methods. Third, the company will increase deployments of several types of robots in manufacturing and other commercial environments.

"The arrival of the physical AI era will begin with a truly embodied brain that understands the world. And a truly embodied brain that understands the world will eventually emerge in the real world," stated Infiforce in its funding announcement.

Infiforce, company statement

Looking ahead, Infiforce plans to continue expanding deployments in manufacturing, inspection, logistics, energy, and commercial settings. The company is also developing world models with longer-term memory and continuous-learning capabilities, which could allow robots to improve their understanding of the physical world over time rather than remaining static after initial training.

How Does This Fit Into the Broader Physical AI Trend?

Infiforce's funding reflects growing investment in embodied AI, a field focused on creating robots and systems that can understand and interact with the physical world. Unlike large language models that process text, embodied AI systems must perceive their environment through sensors, make decisions about physical actions, and learn from the consequences of those actions. The company's emphasis on using first-person data and transferable intelligence across different robot bodies suggests a shift toward more practical, scalable approaches to robot training.

The funding also highlights China's strategic focus on robotics and AI. Infiforce's investors include state-owned investment platforms and Zhejiang University Science and Technology Innovation Group, indicating government support for the company's vision of scaling embodied AI across industrial and commercial applications. As the company expands deployments across more cities and scenarios, it will generate more real-world data that can be fed back into its models, creating a virtuous cycle of improvement and deployment.