Nscale, Figure Sign $3.5B Humanoid Robot Compute Deal
Nscale and Figure announced a multi-year partnership that could deploy up to 100,000 Nvidia GPUs, starting with a $3.5 billion compute commitment The agreement may scale beyond $6 billion, with Nvidia Vera Rubin GPUs targeted for deployment in Barstow, Texas in H2 2027 Nscale will serve as Figure's preferred compute provider and become a shareholder via a strategic investment (amount undisclosed) The partnership centers on advancing "physical AI" for Figure's next-generation humanoid robots powe
Analysis
TL;DR
- Nscale and Figure announced a multi-year partnership that could deploy up to 100,000 Nvidia GPUs, starting with a $3.5 billion compute commitment
- The agreement may scale beyond $6 billion, with Nvidia Vera Rubin GPUs targeted for deployment in Barstow, Texas in H2 2027
- Nscale will serve as Figure's preferred compute provider and become a shareholder via a strategic investment (amount undisclosed)
- The partnership centers on advancing "physical AI" for Figure's next-generation humanoid robots powered by its Helix AI model
- Nvidia frames the deal as activating a "robotics flywheel" spanning training on Vera Rubin, simulation validation via Isaac Sim, and deployment on Nvidia GPUs in robots
Why It Matters
This partnership signals a major convergence between large-scale AI compute infrastructure and physical robotics, marking one of the largest dedicated compute commitments for humanoid robot development to date. For AI practitioners, it underscores that the next frontier—physical AI—demands compute scales comparable to frontier LLM training, reshaping how robotics companies approach infrastructure strategy.
Technical Details
- Compute Scale: Up to 100,000 Nvidia GPUs, with an initial $3.5 billion commitment potentially growing beyond $6 billion in total compute spend
- Hardware Platform: Nvidia Vera Rubin GPUs, scheduled for deployment beginning in the second half of 2027 at a facility in Barstow, Texas
- AI Model: Figure's Helix model, which scales in capability with increased data and compute, analogous to learned systems in traditional AI
- Full-Stack Integration: Nscale provides the power, computing platform, and orchestration layer integrated with Nvidia AI infrastructure, supporting Figure's models from training through deployment
- Robotics Flywheel Pipeline: Training on Nvidia Vera Rubin via Nscale's AI cloud → validation in Nvidia Isaac Sim → deployment on Nvidia GPUs embedded in Figure's humanoid robots
Industry Insight
- The $3.5B–$6B+ compute commitment for humanoid robotics demonstrates that physical AI is entering an era of infrastructure-scale investment previously reserved for large language models, suggesting robotics companies will need similar cloud partnerships to compete
- The vertical integration between compute provider (Nscale), chip designer (Nvidia), and robot manufacturer (Figure) creates a replicable blueprint for the robotics industry, where hardware, simulation, and training are tightly coupled
- Nscale's dual role as compute provider and shareholder—plus its exploration of using humanoid robots in its own supply chain—illustrates a emerging trend where AI infrastructure companies are becoming both enablers and early adopters of the robotics they power
Disclaimer: The above content is generated by AI and is for reference only.