Researchers have demonstrated that AI coding agents can perform complex physical tasks autonomously, including installing GPUs, connecting cables, and configuring hardware. The breakthrough suggests agentic AI is moving from software into hardware deployment.
The experiments involved giving AI agents high-level instructions (e.g., "set up this GPU cluster") and allowing them to decompose tasks, execute commands, and adapt when they encounter obstacles. The agents succeeded in real-world hardware installation without constant human prompting.
This is a significant capability expansion. If AI agents can manage hardware deployment, they can handle infrastructure setup, server configuration, and datacenter optimization—jobs that currently require skilled technicians. The implications stretch from cost reduction to autonomy concerns: systems that can physically manipulate infrastructure need careful governance.