Figure AI Launches Massive Robot Dataset, Backs It With $1 Billion Investment
Quick Read
- Figure AI has launched “Project Go-Big,” aiming to build the world’s largest humanoid pretraining dataset.
- The effort uses first-person video from over 100,000 residential units through a partnership with Brookfield.
- The push backed by more than $1 billion from Figure’s Series C funding round, which valued the company at $39 billion.
- Early tests show robots learning to navigate cluttered homes just from watching human video.
- The goal is to make humanoid robots useful in everyday homes, not just factories.
Figure AI, the humanoid robotics company chasing a future where robots help around the house, has rolled out what it calls the world’s largest Figure AI robot dataset. Known internally as Project Go-Big, the initiative gathers first-person video footage from more than 100,000 residential units, all in partnership with investment giant Brookfield.
The idea is simple: instead of teaching robots one task at a time in a lab, Figure wants its machines to learn the way humans do, by watching how people actually move, grab objects, and get around real homes. That data feeds into Helix, Figure’s AI model that controls its humanoid robots’ bodies, including their hands.
This isn’t a Side Project. It’s Backed by Serious Money
Figure closed a Series C funding round of more than $1 billion in 2025, pushing the company’s valuation to $39 billion, and a large share of that capital is going toward building out this data pipeline, along with GPU infrastructure and expanded robot production.
Early results are already turning heads. Figure says robots trained purely on human video have managed to navigate cluttered, unfamiliar home environments after being given simple natural-language instructions, without needing task-specific programming first. That’s a meaningful step for a company whose robots have mostly proven themselves on factory floors, including a stint on BMW’s assembly line.
Brookfield’s involvement adds real scale. The firm manages hundreds of millions of square feet of property worldwide, giving Figure access to a huge and varied pool of real-world spaces to learn from, something far cheaper and faster than building simulated environments from scratch.
