Nscale has agreed to supply Figure with at least $3.5 billion in compute resources, including access to up to 100,000 Nvidia GPUs, while acquiring a stake in Figure. This move integrates compute infrastructure, platform deployment, and robotic AI development in a tightly coupled ecosystem.
- Nscale supplies $3.5B+ compute with 100,000 Nvidia GPUs from 2027
- Equity investment aligns cloud provider with robotics AI customer
- Plans to co-locate robots in data centers for scaled physical automation
Infrastructure signal
Nscale’s commitment covers a substantial scale of GPU compute power, centered on Nvidia’s Vera Rubin platform, with the initial scope set at $3.5 billion and ambitions to exceed $6 billion. The physical infrastructure will be housed primarily at Nscale’s Texas facility, bringing significant cloud GPU capacity dedicated to robotics AI training and inference.
By embedding compute resources with direct equity in Figure, Nscale is creating a vertically integrated deployment model that blurs lines between cloud provider, hardware user, and developer platform. This closed-loop relationship is further reinforced by shared Nvidia hardware at every stage, from training through to real-world humanoid deployment.
Developer impact
Developers working on Figure’s Helix robotics models will gain exclusive access to a tailored compute environment optimized for large-scale AI training workloads. This dedicated setup promises more consistent performance, streamlined access to GPU resources, and reduced overhead from multi-tenant cloud environments.
The partnership also accelerates the feedback cycle by enabling seamless integration of model validation tools such as Nvidia Isaac Sim and actual hardware deployment. This integration supports more rapid iteration, enhanced observability of system performance, and improved end-to-end developer workflows across simulation, training, and deployment phases.
What teams should watch
Operations teams should monitor the planned exploration of deploying humanoid robots within Nscale’s data centers—a novel approach where robots physically automate repetitive tasks on-site. This could reshape facility staffing and influence future site design, integrating robotics directly into cloud infrastructure workflows.
Product and platform teams need to be aware of the strategic shift towards tighter coupling between compute provision and physical robotics output. This may lead to new API integrations, pipeline adjustments, and closer cross-team collaboration to support emerging workflows that span virtual AI training and physical robot operation.