Meta is quietly piloting maintenance robots inside its data centers, running experiments that could shift much of the physical work of keeping its AI infrastructure online from human technicians to automated systems, as first reported by Ars Technica and other outlets. The company is testing multiple robot platforms that can plug and unplug cables, reboot misbehaving servers, and even reseat hardware components—tasks that have long been the domain of on-site ops teams.
Several reports say Meta is trialing robots and related hardware from Watney Robotics, Kinova, and industrial automation giant ABB across different facilities. Kinova’s Gen3 robotic arm is being evaluated to remotely cut power to servers and bring them back online, essentially automating what used to be a technician walking over to flip a switch or press a button. Watney Robotics, a startup that explicitly focuses on autonomous systems for data centers and supercomputing infrastructure, has reportedly supplied dual-arm robots that Meta has been testing since mid-2025 in an Altoona, Iowa data center building to handle portions of cable routing work under human supervision. ABB robots are said to be in trials at an Ohio facility, where they have been used to reseat server components, hinting at a broader push to industrial-grade automation on the server floor.
For rank-and-file workers, the implications are stark. One Meta data center employee estimated that a successful cable-swapping robot could take over as much as 80 percent of some technicians’ workloads, a figure echoed in multiple summaries of the tests. Employees who once assumed that hands-on, physical maintenance roles would be shielded from AI disruption now describe a very different mood, with many expecting robots to encroach on most of their daily tasks if these pilots scale up. Internal unease has spilled into public forums and social media as staffers warn that the same automation wave that transformed warehouses and logistics is arriving for white-and-gray-collar data center jobs.
The experiments go well beyond simple button-pressing. In addition to the Kinova Gen3 tests, Meta is evaluating a separate system that can automatically swap networking cables—precise, error-prone work that normally requires a human to trace lines and reseat connectors. One prototype, described as resembling a finger or stylus, has been used to remotely press power buttons on devices such as Mac Mini units, enabling remote restarts without dispatching a person to the rack. Other accounts describe robots tasked with moving server racks, replacing cables, and performing visual inspections of equipment, expanding automation from emergency fixes to routine preventative maintenance.
Meta’s hiring patterns suggest these trials are part of a longer-term strategy rather than a one-off science project. Job listings for roles like “Robotics Manipulation Engineer” and Infra Robotics positions describe responsibilities such as designing robotic solutions for server retrieval and insertion, component replacement, and integration with diagnostics systems, as well as leading site acceptance testing and post-release support for automated server-handling cells. In a 2024 public talk highlighted by industry observers, Meta robotics leader Eric Xu outlined a vision of robots improving incident response, environmental monitoring, and preventive maintenance in data centers, reinforcing the idea that these pilots are stepping stones toward a largely automated backbone for Meta’s AI ambitions. Analysts and aggregators note that the push comes as Meta’s AI infrastructure spending surges, making labor an obvious cost center to target for efficiency gains.
Meta is not alone in this shift. Commentators point out that Microsoft, Google, and Amazon are also exploring robots for delicate data center tasks, with industry chatter framing Meta’s efforts as part of an arms race to build hyperscale AI infrastructure with minimal human overhead. Advocates argue that robots could ultimately improve safety by handling heavy equipment, high-voltage power operations, and repetitive strain tasks, while also enabling faster, more consistent incident response at massive scale. For anyone who has sunk hours into sci-fi strategy games about automated megastructures, the sight of dual-armed bots quietly tending server racks in Iowa feels eerily on the nose—and for the people who currently do that work by hand, it is less a cool future scenario than a very real, very near-term career problem.
