Figure AI

Robotics · AI Infrastructure

Launch

Figure AI launches Index data app, seals $3.5B Nscale deal for 100K GPUs

General-purpose robots can't learn from internet text, so collecting real-world human video has become as capital-intensive as chip supply.

• Figure AI came out of stealth with Index, a human-data collection app, on Aug 25, then on Sept 3 signed a multi-year deal with Nscale to deploy up to 100,000 NVIDIA GPUs, with Nscale taking an equity stake and becoming Figure's preferred compute provider.
• Figure builds Helix, a vision-language-action model that runs a humanoid robot's upper body at 200 commands per second; the two moves target its self-described twin bottlenecks of data and compute.
• Index crossed 264,000 downloads across 108 countries and 44,000 weekly active users who uploaded over 16 million videos, now processing 30 minutes of video every second and paying Creators $15 million so far.
• The Nscale deal carries an initial $3.5 billion compute commitment with intent to scale beyond $6 billion, running NVIDIA's Vera Rubin platform starting in the second half of 2027 in Barstow, Texas.
• Figure plans to spend over $1 billion on data and compute in the next 12 months to scale Index collection 100x, months after closing a $1 billion Series C at a $39 billion valuation.
• Robotic manipulation data can't be scraped from the internet like text, so Figure's twin bets on owning its data pipeline and locking down GPU supply show physical AI's real competition is now over infrastructure, not just models.