Meta tests robots in data centers but hits real-world limits

Meta is experimenting with robots to handle infrastructure work at its data centers, but the rollout faces significant hurdles. Two pilot projects—one using humanoid machines and another using wheeled platforms—reveal both the potential and the practical constraints of automating warehouse-scale computing facilities.

Testing Grounds in Altoona and Beyond#

The Altoona data center campus in Pennsylvania serves as Meta's innovation lab for infrastructure technologies; prototypes tested there often move into production elsewhere across the company's network. Since mid-2023, Meta has been evaluating a pair of dual-armed Watney robots for cabling tasks at one Altoona facility. While supervised by humans and slower than their biological counterparts, the machines are laying groundwork for broader deployment.

Meta is also running trials at its newer Prometheus campus in New Albany, Ohio, where ABB-manufactured four-wheel robots equipped with scissor-lift mechanisms and six-axis arms are being deployed to reseat components and eventually handle additional duties with less human oversight.

Current Reality: Promise Tempered by Practical Problems#

Performance gaps remain stark. Workers have flagged significant charging downtime, and the machines struggle with nuanced tasks—particularly the intricate wiring needed for Nvidia's advanced computing clusters. As one former Meta staffer noted, infrastructure has long been engineered around human capability; retrofitting it for robotic handling demands substantial redesign work.

The broader industry hasn't yet delivered a turnkey solution either. Helen Oleynikova, CEO of Exclaim Robotics, summed up the state of play: "There's been a lot of pilots and demos, but no provable working solution." Meta reportedly explored faster alternatives, such as drones for inventory management, before deciding against pursuing them.

What's Next#

Despite today's limitations, Meta is advancing its robotic workforce incrementally. The company acknowledges internally that robots must eventually handle complex tasks, but recognizes that redesign and time are prerequisites. The path forward hinges on making systems intelligent enough and infrastructure flexible enough to let machines work alongside—and eventually in place of—humans.