Meta is experimenting with robots that can take on physical tasks inside its data centers, according to several current and former workers familiar with the projects. The work includes testing machines that can plug in cables, reset servers, move heavy equipment, inspect failures, and support inventory tracking.
The experiments come as Meta’s data center footprint expands and spending on AI infrastructure rises. The company says it still needs skilled workers, but some employees see the robotics push as a sign that parts of the job could change quickly.
What Meta Is Testing
According to workers familiar with the projects, Meta is using robots and related hardware from several vendors, including Watney Robotics, Kinova, and ABB. Kinova and ABB declined to comment, while Watney didn’t respond to requests for comment.
One experiment involves a Kinova Gen3 robotic arm that Meta is evaluating for power cycling, which means cutting off electricity to servers. Another test focuses on a different robot that can swap networking cables.
One Meta data center worker estimated that, if the cable-swapping robot succeeds, it could replace up to 80 percent of some people’s workloads. That estimate does not mean the technology is already fully deployed, but it shows why the tests are being watched closely by workers who perform hands-on maintenance.
Meta has also quietly adopted a simpler robot in some facilities. It resembles a finger or a pointed stick and can press the power button on a Mac Mini or another device when a human remotely prompts it to restart the machine.
Why Data Centers Are Hard for Robots
Data centers may look orderly, but the physical work inside them can be delicate. Servers, cables, power buttons, indicator lights, and racks all require careful handling. The source article notes that, until recently, robots were often poorly suited to this environment.
Earlier industry trials sometimes went badly. A person briefed on experimentation said some robots crushed servers while attempting simple tasks. Cost was another barrier.
That picture is changing. Hardware has become cheaper, and the AI models that power robots are now more capable than they were a few years ago. Several startups are working on commercial robots that could assemble servers, swap parts, and replace wires.
Meta’s own stated robotics ambitions also go beyond one-off button pushing. At a conference last year, Eric Xu, senior manager for robotics at Meta, said the company’s long-term goals include putting robots in data centers. He argued that they could speed up incident response, monitor environments, and handle preventative maintenance.
The Labor Question
Meta declined to comment on the specific testing described in the source article. Company spokesperson Francis Brennan said Meta is investing heavily in training and hiring workers to build and operate its data centers. He also said there is a major shortage of skilled workers and that the company needs more workers, not fewer.
At the same time, workers described concern inside the company’s data center operations. In Altoona, Iowa, where Meta operates what is now its largest data center campus, a person familiar with worker group chats said the conversations have become demoralized over the prospect of robots taking jobs.
The Altoona campus employs hundreds of workers. Meta opened the facility in 2014 and has several years left on a property tax break that allows it to save tens of millions of dollars annually. The city has viewed the arrangement as a worthwhile tradeoff in part because of the Meta jobs associated with the site.
Some workers also worry that the quality of remaining roles could decline even if automation moves slowly. Two workers said Meta is developing AI software and other technologies that they believe could let the company hire lower-skilled and lower-paid workers while centralizing some roles in areas such as Denver, Colorado.
Meta points to training as evidence that it is still investing in people. This year, the company launched a long-term program to train thousands of people annually from any background in electrical, mechanical, and plumbing work at no cost to them. Meta is also guaranteeing employment in states such as Louisiana, Ohio, Indiana, and Texas for people who complete the training, and it has partnered with building trade unions to fund apprenticeship programs.
Robots Are Already Doing Some Jobs
Meta has publicly discussed some robotics work before. In a 2023 YouTube video, the company said it was using a self-driving tugger robot to move heavy server racks inside data centers. It also described a wheeled, barcode-reading robot developed in-house to track inventory.
People familiar with the projects said Meta’s tugger and inventory robots are now operating in a number of its data centers, including in Iowa and Virginia. They said the robots generally work well, although they still have limits.
The inventory robot is also being used for failure inspections. But it cannot distinguish green indicator lights from red ones because its camera detects only grayscale. That means humans still need to perform some checks when equipment breaks down.
The same robot can struggle with corners and with cables that sit in the path of its wheels. Because there are not many of them, the robots also need to be moved between buildings. That still requires a human to open doors and maneuver the machine with a remote control.
Why This Matters Beyond Meta
Meta is not alone in exploring robots for data center work. The source article says Microsoft and Google unveiled separate investments in robots for their data center operations the next year after Meta’s 2023 YouTube video. Amazon last year also discussed using robots to help recycle parts at its facilities.
Ashley Llorens, a managing director for Microsoft’s research accelerator, told WIRED that robotics for economically valuable tasks, such as replacing a faulty server rack, is something Microsoft is looking into.
Supporters of data center robots argue that machines could deliver more consistent output at lower cost than human labor, especially in less populous regions where data centers are common but qualified workers are scarce. They also argue that robots could take on repetitive or dangerous work, leaving technicians to focus on more complex tasks.
The longer-term vision is more speculative but still part of the discussion. If fewer people need to be physically present, data centers could eventually be easier to run in places that are difficult for humans, including underwater, in space, or in other inhospitable environments.
For communities weighing data center tax breaks and possible new rules on the industry, the labor implications may become harder to ignore. Altoona Mayor Dean O'Connor told WIRED he has not given much thought to robots taking jobs at Meta’s data center, while framing automation as inevitable. His view was simple: communities will deal with it as it comes.