Figure AI
The company that proved humanoids work in real factories. 30,000 cars, 11 months, 99% accuracy — then they made it sleeker.
Key Specs (Figure 03)
The latest generation, designed for production-scale deployment.
Latest Developments
Figure proved the business case. Now they're scaling.
Figure retired its F.02 fleet Terminator-style — a short film, "F.02 Decommission", shows the robot walking itself into a vat of molten steel in Finland, with six melts in 20-minute windows so proprietary hardware never leaves company control. Arnold Schwarzenegger lent his blessing, per Figure's own post and The Verge-era coverage from Interesting Engineering.
Figure introduced Helix 2.5, its most advanced neural network yet — pretrained from scratch entirely on the Index dataset. In blind evaluations across 30 unseen homes, the Index-pretrained policy succeeded on 56% of zero-shot trials vs 9% for a policy trained from scratch, per Figure's announcement — evidence that crowdsourced human video, not architecture, drives generalisation.
Figure signed a multi-year deal with AI cloud firm Nscale committing $3.5B in Nvidia GPU compute for training its physical AI models, according to Forbes and eWeek — more than $1.6B beyond its total funding to date. The company also launched Index, a gig-style app paying humans to generate robot training data.
Per Figure's own announcement, the Nscale partnership targets up to 100,000 NVIDIA Vera Rubin GPUs, initial deployment in Barstow, Texas from H2 2027, with the commitment intending to scale beyond $6B — and Nscale taking a strategic equity stake in Figure. Index, public since late August, has passed 16M videos from 264K downloads, with $15M paid to creators.
Figure 02 robots helped produce 30,000+ cars at BMW Plant Spartanburg over an 11-month deployment. 99% task accuracy. The industry's first proven ROI case.
Figure 03 prototype revealed — sleeker form factor, improved dexterity, designed for broader commercial deployment across multiple industries.
BotQ high-volume manufacturing facility announced. Figure's own factory for humanoid production, designed to scale rapidly.
Backed by OpenAI (vision + language models), Microsoft, NVIDIA, Jeff Bezos, Amazon. $39B valuation as of 2025.
Timeline
Focus Areas
Manufacturing
BMW partnership proved it. Assembly line tasks, parts handling, quality inspection. Structured environments are the sweet spot.
Logistics
Warehousing and distribution. Moving goods, palletizing, order fulfillment. Natural extension of manufacturing capability.
Assembly
Complex multi-step assembly tasks that require precision and adaptability. Where Figure's AI advantage matters most.
QA inspection
Vision-based quality control. Figure's OpenAI-powered vision models can spot defects humans miss.
Why It Matters
Figure AI did what no other humanoid company has done: prove ROI in a real factory, at scale, with real cars rolling off the line. The BMW deployment is the reference case for the entire industry. Every procurement manager considering humanoids will point to Figure and ask "when can you match this?"
Their OpenAI partnership gives them the best AI brains in the business. The BotQ factory signals they're serious about manufacturing. And the $39B valuation reflects investor belief that Figure might be the first humanoid company to actually ship at scale.
The others talk about factory deployment. Figure already did it. The race is now about who scales faster.