Covariant

Capacity describes five Covariant robotic putwalls in its New Jersey operation, adding four after its initial robot, to sort mixed-SKU totes and handle beauty-product launch demand.

Otto Group announced a long-term Covariant partnership to begin deploying over 100 robots, starting at Haldensleben and Altenkunstadt, for variable-product fulfilment automation.

Covariant Brain has been deployed in warehouse picking, induction, sortation, putwall, kitting and depalletization use cases, including scaled Radial deployments processing about 1.2 million units per month across 12 robots.

Covariant's March 2024 release describes RFM-1 as an 8-billion-parameter any-to-any sequence model tokenizing language, images, video, actions and physical measurements into a shared space, with generated future-action videos used for scenario selection.

Covariant launched RFM-1 as a commercial Robotics Foundation Model and built Covariant Brain as a universal manipulation intelligence trained on large-scale real-world robotic picking data.

Covariant launched RFM-1 as a commercial Robotics Foundation Model and built Covariant Brain as a universal manipulation intelligence trained on large-scale real-world robotic picking data.

Amazon's 2024 announcement reports a non-exclusive license to Covariant's robotic foundation models and hiring of three founders plus around one-quarter of its employees, while stating Covariant would continue serving dozens of customers. This documents a material organizational transition as well as commercial continuity.

Covariant historically deployed robotic systems globally across multiple warehouse use cases and had raised $222M by 2023, but current 2026 first-party evidence of organizational expansion is limited.

Amazon specifically identified Covariant founders and a group of research scientists and engineers joining its Fulfillment Technologies & Robotics team to develop and implement the licensed technology, providing a concrete historical example of robotics expertise moving between employers.

Covariant work spans robot learning, manipulation, perception, foundation models, warehouse systems and production deployment; the skills remain broadly portable across physical AI and robotics.

Covariant work spans robot learning, manipulation, perception, foundation models, warehouse systems and production deployment; the skills remain broadly portable across physical AI and robotics.