RoboScience Raises ‘Hundreds of Millions of Yuan’ for Embodied AI

BEIJING -- Embodied artificial intelligence startup RoboScience has secured 'hundreds of millions of yuan' in a Series A2 investment round to create AI models and scale up trials in business and industrial settings.

RoboScience disclosed the fundraising deal on Nov. 2, naming Amber Capital, Hongruida Investment and Shanghai Semiconductor Industry Investment as lead investors. Shuang Capital, 360 Strategic Investment and Jinding Capital also participated in the funding round.

The proceeds will finance the company’s AI model enhancemenet, concept tests in logistics, retail and manufacturing sectors, and additional hiring. RoboScience has not revealed the total sum raised or valuation.

Founded in December 2024, RoboScience builds AI models aimed at controlling various robots without the need to train each machine separately. The company’s Visics model uses a combination of a predictive system that determines how objects move and a system that transforms predictions into control instructions for robots.

Alongside that in August, RoboScience unveiled REX G1, which is a wheeled humanoid robot suitable for manipulation tasks in logistics and retail.

As per RoboScience’s claims, its video training data set grew from 1 million hours recorded at the start of 2026 to 4 million hours in October. The firm further mentions that the cost of collecting individual training data samples has fallen to 0.05 yuan.

The Visics platform will be accessible via cloud interfaces as RoboScience has partnered with Tencent Cloud. RoboScience says Visics is being tested in retail and logistics applications but provided no further details.