Nscale signs a multi-year partnership with Figure to deploy up to 100,000 Nvidia GPUs, anchored by a $3.5 billion compute commitment to support Figure’s next generation of humanoid robots.

The collaboration emphasizes a vertically integrated model where Nscale handles power, compute, and orchestration, leveraging NVIDIA infrastructure to optimize training and deployment.

The goal is to advance physical AI by expanding data and compute for Figure’s Helix model, with validation and testing framed around Nvidia Vera Rubin GPUs and Nvidia Isaac Sim.

Index, Figure’s data platform, collects diverse household and workplace activities by processing and filtering large volumes of user-contributed video to train broader robotic capabilities.

Industry momentum is strong, with humanoid robotics funding totaling $8.6 billion in 2026 so far, more than double the amount for all of 2025.

Figure aims to convert extensive training and data into productive, billable robot deployments, not just research demos.

Figure specializes in autonomous general-purpose humanoid robots for industrial and home use, headquartered in San Jose, California.

The press release distribution is handled by PR Newswire, including standard informational context.

Some observers question whether the circular model of investing in customers who buy the provider’s compute inflates demand, though proponents say it accelerates robotics development.

Figure has paid contributors around $15 million to date and plans to substantially increase data and compute spending to move humanoid robots toward reliable, general-purpose work.

Scaling physical intelligence requires vast compute and data to enable robots to perceive, understand, and safely interact with the real world.

Helix enables on-device processing to speed decision-making, while training such models demands massive data and processing power.