Who Controls Physical AI? Orchestration Power in Robotics and Antitrust Law
Abstract
The integration of advanced artificial intelligence into robotics and industrial systems—commonly described as “physical AI”—raises a fundamental question for antitrust law: who controls the systems through which artificial intelligence is translated into real-world economic activity? Robotics manufacturers have traditionally exercised substantial control over industrial automation systems, but emerging forms of orchestration increasingly allow other actors to influence how physical-AI systems are developed, deployed, coordinated, and improved.
This Article argues that antitrust analysis should focus not only on ownership of robots, foundation models, or AI infrastructure, but also on orchestration power: control over the interfaces and coordination layers through which artificial intelligence is translated into physical action. It identifies four principal sources of orchestration power: factory-operation interfaces linking robotics manufacturers and industrial users; agent-based orchestration systems developed by foundation-model firms; open-weight robotics foundation models that provide a common technological foundation for robotics development; and infrastructure-centered robotics-development environments, illustrated by NVIDIA’s integration of computing infrastructure, simulation tools, robot-learning environments, robotics foundation models, and deployment systems.
These forms of orchestration may generate dependency and bottlenecks even where no single firm controls every layer of the technology. The Article therefore develops an antitrust framework for assessing when such bottlenecks become relevant to market power, exclusionary conduct, and remedies. It argues that existing doctrines—including monopolization, refusal to deal, discriminatory access conditions, exclusive arrangements, and tying—can address many physical-AI concerns without treating orchestration power as an independent category of antitrust liability.
At the same time, the Article emphasizes that orchestration power does not itself justify antitrust intervention. Physical-AI systems are technologically complex, safety-sensitive, and rapidly evolving. Antitrust law should therefore distinguish exclusionary uses of strategically important bottlenecks from technical integration that promotes safety, reliability, innovation, and system performance. The goal is not to dictate the architecture of physical-AI systems, but to preserve competitive conditions under which innovation, experimentation, and technological development can continue.
Read the full piece at SSRN.