Humanoid robots need brains, and brains need GPUs — lots of them. On September 3, 2026 (UTC), Figure AI announced a multi-year strategic partnership with British AI cloud provider Nscale to deploy up to 100,000 NVIDIA Vera Rubin GPUs for training its robot AI models. The initial commitment is $3.5 billion, with the potential to expand beyond $6 billion.
The deal
Under the agreement, Nscale will build and operate a dedicated GPU cluster for Figure based on NVIDIA's Vera Rubin platform. The first batch of compute is expected to come online in the second half of 2027 (UTC) at a facility in Barstow, Texas. Nscale is also making a strategic investment in Figure as part of the partnership.
Figure has been candid about its bottleneck: the company says its current development phase is constrained by the data and compute supply needed to train Helix, its foundation model for humanoid robots. Figure plans to spend over $1 billion on data and compute in the next 12 months alone — and this Nscale deal is a big chunk of that.
Why Nscale?
Nscale, founded in 2024, has become one of the hottest names in AI infrastructure. The company just raised $3.5 billion in pre-IPO financing led by Third Point and NVIDIA, and its total contract backlog has roughly doubled to $103 billion in the past month — including a $45 billion deal with Anthropic that Microsoft and Google both walked away from.
NVIDIA has backed Nscale since its Pre-Series C SAFE round in October 2025 (UTC) and followed on in the March 2026 (UTC) Series C. For Figure, partnering with Nscale means access to next-generation Vera Rubin chips without building its own data centers — the same playbook Anthropic used.
The robot compute race
Figure isn't the only humanoid robot company throwing money at GPUs. The entire industry is waking up to the fact that training models that can perceive, reason, and act in the physical world is enormously compute-intensive. Figure's Helix model processes vision, language, and motor control simultaneously, requiring GPU clusters that rival those used for large language models.
Founded in 2022 by Brett Adcock, Figure has raised $675 million from investors including NVIDIA, Microsoft, and OpenAI. Its Figure 02 robot is being tested at BMW's Spartanburg plant in South Carolina, and the company has a partnership with OpenAI to build custom AI models for its robots.
The message from this deal is clear: the humanoid robot race won't be won by the best hardware alone. It'll be won by whoever can train the smartest AI — and that means whoever can secure the most GPUs. Figure just locked in a lot of them.
Sources: Figure, Nscale, 财经涂鸦, 格隆汇, 机器人前瞻, Sina Finance
No comments yet