Meta is evaluating robots to perform tasks including plugging in cables, resetting servers and other operational duties within its data centers. This initiative, previously unreported, aims to manage its growing data center footprint with fewer human workers, controlling labor costs during increasing AI infrastructure investment.
The company is experimenting with robots and hardware from several vendors, including Watney Robotics, Kinova and ABB. A test involves a Kinova Gen3 robotic arm, which Meta is assessing for power cycling or cutting electricity to servers. Another robot is being tested to swap networking cables.
A Meta data center worker estimates robot deployment could replace up to 80 percent of some workers' tasks. The worker expressed concern, saying those performing physical tasks previously felt secure but now face potential displacement.
Meta declined to comment on the specific robot testing. However, company spokesperson Francis Brennan said Meta is making significant investments in training and hiring workers to build and operate its data centers. Brennan added the United States is experiencing its largest infrastructure boom since World War II, leading to a shortage of skilled workers.
Eric Xu, Meta senior manager for robotics, outlined the company's long-term goals for integrating robots into data centers at a conference last year. Xu said robots could improve incident response, monitor environmental conditions and perform preventative maintenance. He emphasized the need for these efforts, saying more collaboration and research are required.
Historically, robots were unsuited for the precise work required in data centers. They were expensive, and earlier industry trials sometimes resulted in damaged servers during basic tasks. Recent advancements have made hardware more affordable, and the AI models powering these robots are now more capable.
Several startups are developing commercial robots designed for server assembly, part replacement and wire swapping. These automated systems offer more consistent, lower-cost output than human labor. These solutions are valuable in less populous regions where data centers are common but qualified workers are scarce.
