Axis Robotics Raises $12 Million in Seed Funding to Expand Physical AI Data Infrastructure
- Axis Robotics said it raised $12 million in seed funding in a round led by Hack VC.
- The company said it plans to use the proceeds to enhance data-generation technology for robot AI training and expand its global data network.
- Axis Robotics said it is building physical AI data infrastructure through development of a compounding data engine and the sequential release of Sim Dataset V2 and the DAgger Dataset.
Forecast Trend Report by Period
Axis Robotics, a physical AI data infrastructure company, has raised $12 million in seed funding.
The company announced on July 27 that it secured the funding in a seed round led by Hack VC. Nomad Capital, Pi Network, 10K Ventures and multiple angel investors also participated. Axis Robotics plans to use the proceeds to enhance data-generation technology for robot AI training and expand its global data network.
Unlike large language models, physical AI faces a constraint in securing large volumes of high-quality data that real-world robots can learn from. Data reflecting a range of work environments and robot types remains limited. As a result, the cost and time required to build datasets are key factors in determining performance.
Axis Robotics is developing a “compounding data engine” that integrates data generation, collection, training and improvement on a single platform. Through a browser-based remote data collection platform and a mobile data collection system, the company enables contributors around the world to generate robot training data. More than 100,000 contributors globally are currently participating, and the company collects more than 1,200 hours of simulation data and more than 20,000 hours of real-world data each month.
The company also disclosed technical results. Its simulation dataset, Sim Dataset V1, improved the overall success rate by 12.9 percentage points over the existing Pi0.5 baseline on LIBERO-Plus, a benchmark for evaluating robot performance.
Axis Robotics is also stepping up commercialization. The company is working with Booster Robotics, Manicore Tech, Pigeon Robotics, Dexmal, Lotus and Geely Auto to supply customized training data to robot manufacturers and physical AI companies. It plans to release Sim Dataset V2 in September and the DAgger Dataset in November.
Founder Chris Feng said competitiveness in physical AI depends less on the model than on how quickly data can be accumulated and improved. He added that the company will build the data infrastructure needed for general-purpose robot intelligence through its global data network and compounding data engine.
