Covariant is an AI robotics company that built AI-powered robotic picking systems for warehouse, logistics, and manufacturing environments. Founded in 2017 and headquartered in the United States by former OpenAI and academic researchers, the company became known for applying large-scale machine learning to enable robots to grasp and manipulate a wide variety of objects, including items they had not seen before.
Covariant's technical centerpiece was RFM-1, a robotics foundation model the company announced in 2024. RFM-1 was positioned as giving robots a more human-like ability to reason about and understand their environment, allowing them to follow instructions provided as text or images, answer questions, and request clarification. The model was trained on large volumes of real robot interaction data, and Covariant's systems were deployed in production logistics operations.
A defining event in the company's history occurred in 2024, when Amazon entered into an agreement that brought Covariant's three co-founders, Pieter Abbeel, Peter Chen, and Rocky Duan, and roughly a quarter of its staff to Amazon, along with a non-exclusive license to Covariant's models. Reporting characterized this as a reverse-acquihire structure in which Amazon obtained the core technology and key talent without a full acquisition.
Following that arrangement, leadership of the remaining company transitioned, with its chief operating officer moving into the chief executive role and other remaining leaders continuing to operate the business. The events materially changed Covariant's trajectory relative to its earlier independent path, and prospective users and partners should evaluate the company's current status and offerings in light of the 2024 Amazon deal.
Covariant historically competed with other warehouse automation and robotic manipulation providers. Its differentiation was a foundation-model approach to generalizable robotic manipulation rather than narrowly scripted automation.
Given the 2024 talent and licensing transition to Amazon, the company is most relevant today to those tracking the evolution of robotics foundation models, warehouse automation, and the consolidation of AI robotics talent into large technology platforms.