Nvidia to Open Source Advanced Physics Engine "Newton" in July

In Brief:

Nvidia will open-source Newton, the world’s most advanced physics engine, this July.

Developed with DeepMind and Disney Research, Newton will power virtual robot training with GPU acceleration and real-time performance.

On May 19, Nvidia CEO Jensen Huang announced that the company will open-source Newton, a cutting-edge physics engine designed to simulate real-world physical laws with high precision. Developed in collaboration with DeepMind and Disney Research, Newton is expected to become a foundational tool for training AI and robotics in virtual environments.

"Creating robots in the physical world is impractical—we must train them in a virtual one," Huang stated, emphasizing the need for accurate and scalable simulation.

Newton supports full GPU acceleration, high differentiability, and ultra-real-time operation, allowing AI systems to learn effectively through experience. These capabilities make it ideal for reinforcement learning and robotics research.

The engine will be integrated into Nvidia’s ISAAC simulator, enabling highly realistic simulations where digital robots can be trained before real-world deployment. By open-sourcing Newton, Nvidia aims to democratize access to next-generation simulation tools across academia, AI labs, and industrial R&D.

This move signals Nvidia’s broader push into robotics, AI training infrastructure, and digital twin technology—positioning the company at the intersection of AI development and physical system simulation.

‍