The robotics foundation model company has also established more than 60 deployment partnerships, with its technology now being used across manufacturing, logistics, inspection, security, food preparation and other applications.
The commercial figures were disclosed alongside new details about S1, Skild’s latest robot foundation model. S1 is designed to learn previously unseen tasks from a single video demonstration without updating its model weights or undergoing task-specific post-training.
An operator can record a task and provide that video to the model as a prompt. S1 then interprets the objects, sequence and intended outcome before translating the demonstration into actions for the robot it is controlling.
Skild has demonstrated the system performing previously unseen tasks lasting as long as 10 minutes, including potting a plant, making pancakes, brewing pour-over coffee and completing multi-step kit assembly. In one plant-potting test, the company went from recording the demonstration to autonomous execution on physical hardware in 11 minutes.
In Skild’s testing on new multi-step tasks, S1 achieved approximately 66% success per step, compared with 9% for a comparable AI system. Skild estimates that one short video demonstration can provide roughly the same benefit as 380 hands-on training examples. These performance figures are company-reported and have not been independently benchmarked.
The technology is already moving into industrial environments. Skild, NVIDIA and Foxconn are deploying the Skild Brain on dual-arm manipulators for assembly of NVIDIA Blackwell systems. One demonstrated workflow involves installing a busbar and limit block, fastening 16 screws and adapting when the environment changes.
Skild and NVIDIA are also jointly developing GPU-accelerated simulation solvers for robot contact, gripping and manipulation that are planned to become available to developers through NVIDIA’s Newton physics engine.
