SoftBank and MGX participated, two investors said. OpenAI had originally targeted up to $30 billion and sought a $1.5 trillion valuation.
Physicl raises $10 million for robot-training simulations
Inovia Capital led the seed round. The startup is releasing Studio, a tool with 10,000 3D environments and 100,000 assets for generating synthetic training data.

Physicl has secured $10 million in seed financing, with Inovia Capital leading the round, to develop more realistic simulations for robot training and world models. The venture is led by founder and CEO Alex de Vigan.
The startup is releasing Studio, which provides 10,000 3D environments and 100,000 assets representing different objects. Physicl says users can build scenes that produce synthetic training data, with a particular focus on difficult cases during post-training. Those could include helping a robot recognize a cup amid clutter or in shadow.
De Vigan previously spent a decade building nfinite, a business that helped retailers and brands create digital 3D versions of products for online displays. As generative AI pushed that company toward image generation, he identified another use for its technology: producing environments with realistic geometry, materials and physics for physical AI.
World model developer SpAItial worked with Physicl as an early customer. SpAItial founder and CEO Matthias Niessner said its technology, which creates 3D simulations using 2D images, needed highly realistic synthetic data for training.
De Vigan wants Physicl to supply tools for continuous data generation rather than simply sell datasets. He envisions customers feeding images and lidar recordings into the platform to create 3D representations and multiple training variations.
