This is some really great signal for robotics foundational model business.

Last month Skild AI released a model capable of "in-context learning," i.e. allowing robots to master multistep tasks, including sequences lasting more than 10 minutes, by watching a video.

And, they have reached a $100M ARR, just 10 months after commercial deployments began.

Manufacturers with frequent product changes have the clearest economic reason to pay for S1’s flexibility, which Skild says lets robots take instructions from a video demonstration without retraining. a revised assembly sequence can avoid another round of data collection and model updates.

Its general-purpose AI “brain” now:
- Powers hundreds of robots
- Serves 60+ companies, up from 8 earlier this year
- Operates across warehouses, factories, hospitals, homes and cafes
- Supports humanoids, arms, hands and quadrupeds