The initial agreement represents a $3.5 billion compute commitment from Figure, with plans to scale beyond $6 billion. The first GPUs are expected to come online in the second half of 2027 at Nscale’s Barstow, Texas facility.
The infrastructure will be built around NVIDIA’s Vera Rubin platform and used for both training and deployment of Figure’s models. Nscale will become Figure’s preferred compute provider and will also make a strategic investment in the robotics company, becoming a shareholder.
Figure CEO Brett Adcock said the company sees data and compute as two of the main constraints on scaling humanoid intelligence. Its Helix models improve as they are trained on larger amounts of physical-world data, increasing the amount of compute required as Figure expands its training infrastructure.
The agreement follows Figure’s launch of Index in August, a crowdsourced data collection network built to gather human video for robot training. Figure says Index has already collected more than 16 million videos from users across more than 100 countries. The company said at the time that it planned to spend more than $1 billion on data and compute over the following 12 months.
The Nscale agreement pushes that infrastructure plan much further. At full scale, the partnership could give Figure access to as many as 100,000 next-generation GPUs as it works toward training Helix for increasingly general tasks across industrial and home environments.
The companies will also explore using Figure’s humanoid robots within Nscale’s own supply chain and infrastructure operations.
