Skild aims to dominate robot middleware
Project Prometheus
This reveals that Skild is trying to become the software layer that rides on top of existing robot makers, not another robot OEM. The important part is distribution. ABB and Universal Robots already sell into factories, and NVIDIA supplies the simulation and compute stack, so Skild can slot its model into robots customers already trust instead of spending years building hardware, channels, and industrial credibility from scratch.
-
Skild Brain is designed as a hardware agnostic control layer. A manufacturer maps a robot’s joints, sensors, and cameras into Skild’s interface, then the model can handle jobs like grasping, navigation, or pick and place across arms, humanoids, and mobile robots. That is why omni-bodied matters commercially, it lets one model sell across many robot categories.
-
The ABB and Universal Robots relationships matter because they connect Skild to installed industrial channels. ABB is adding NVIDIA Omniverse into RobotStudio for factory simulation, and Skild says its brain will be deployed through ABB and UR across industrial and collaborative robots, which puts it closer to live production environments and the data they generate.
-
Compared with Physical Intelligence, Skild is leaning harder into closed industrial distribution while Physical Intelligence has leaned into open source model release and developer adoption. Physical Intelligence published π0 weights and code, while Skild is pairing a proprietary model with OEM and infrastructure partners that can bring it directly into factories and assembly lines.
The next phase is a race to own the robot middleware layer inside real production workflows. If Skild can turn ABB, UR, and NVIDIA into repeatable deployments, it compounds data from many robot types and factories at once. That would make it harder for newer labs to compete on model quality alone, because the edge shifts from research demos to installed industrial learning loops.