Nscale has committed US$3.5 billion of AI compute to humanoid robotics company Figure under a multi-year partnership that could scale beyond US$6 billion and deploy up to 100,000 NVIDIA GPUs.
The agreement will make Nscale Figure’s preferred compute provider, supporting development of its Helix AI models and humanoid robots. Initial GPUs are targeted for deployment from the second half of 2027 at Nscale infrastructure in Barstow, Texas.
The companies said the potential deployment would use NVIDIA’s Vera Rubin platform, providing the computing capacity required to train increasingly sophisticated Physical AI systems. Nscale will also become a Figure shareholder through a strategic investment.
The partnership combines Nscale’s power, compute platform and orchestration capabilities with NVIDIA AI infrastructure. Figure expects the additional capacity to accelerate training and deployment of Helix, its AI model designed to enable humanoid robots to understand and interact with physical environments.
Josh Payne, CEO and founder, Nscale, said physical intelligence represented AI’s “next frontier”, following rapid growth in inference and Agentic AI.
Jensen Huang, founder and CEO, NVIDIA, said the collaboration created a “physical AI flywheel”, spanning model training on Vera Rubin, validation through NVIDIA Isaac Sim and deployment on NVIDIA GPUs inside Figure robots.
Figure founder and CEO Brett Adcock said data and compute remained key constraints on the company’s ambition to bring humanoid robots into homes.
The partners will also explore using Figure humanoids within Nscale’s supply chain, potentially creating another practical deployment environment for the technology as operations expand across global technology markets.
