Meta Is Testing Robots for Server Resets, Cable Swaps, and Data Center Work

Introduction

Meta is moving robots beyond basic logistics and into the hands-on work of maintaining data centers. Current and former workers familiar with the projects say the company is testing machines that can plug in cables, reset servers, cut power, and inspect equipment. Meta declined to comment on the specific experiments, while saying publicly that it continues to invest heavily in training and hiring workers to build and operate its facilities.

The push comes as Meta expands its data center footprint and spends heavily on AI infrastructure. Automation could be especially attractive at sites where qualified technicians are difficult to recruit. It could also deliver more consistent performance for repetitive tasks. At the same time, workers are questioning whether the same systems will eventually reduce headcount or narrow the skills expected from people who remain.

Key points

• Meta is evaluating a Kinova Gen3 robotic arm for power-cycling or disconnecting servers, alongside another robot designed to swap network cables.
• A simpler device can press the power button on a Mac Mini or similar equipment after receiving a remote instruction from a human.
• The company already uses an autonomous tugger to move heavy server racks and a wheeled, barcode-reading robot for inventory and some failure inspections.
• Current systems are far from autonomous. One inventory robot cannot reliably distinguish green from red indicator lights, struggles with corners and floor cables, and is considered too slow by some workers. Moving it between buildings can also require a person to open doors and operate remote controls.
• Meta’s stated long-term goals include faster incident response, environmental monitoring, and preventive maintenance. Microsoft, Google, and Amazon are exploring related applications.

From physical assistance to labor redesign

Robots are not simply a replacement for every technician. Their immediate value is more likely to come from handling repetitive, hazardous, or physically demanding work. That could leave employees with more time for complex troubleshooting. Yet some Meta workers fear that AI-guided tools will allow the company to hire lower-skilled employees whose role is mainly to follow machine-generated instructions.

The distinction matters to communities that offer tax incentives for data centers. Meta’s campus in Altoona, Iowa, has been associated with hundreds of jobs and substantial local tax benefits. If more maintenance work becomes automated, a facility may continue to consume significant electricity, land, and public resources while producing fewer jobs or less technically demanding ones.

Why it matters

Full autonomy remains difficult. Data centers contain dense cable layouts, varied hardware, narrow passages, and unexpected failures. These conditions favor technicians who can interpret context and improvise. Still, cheaper hardware and more capable AI models are making robotic deployments more plausible than they were during earlier industry trials, when machines reportedly damaged equipment while attempting simple tasks.

The longer-term ambition is broader than cost reduction. Robots can operate in darkness or at higher temperatures, and could eventually make facilities in underwater or space environments more practical. The central question is therefore not whether robots will replace everyone, but which tasks will disappear, which skills will become more valuable, and who will benefit from the productivity gains. Meta’s free training programs in electrical, mechanical, and plumbing work show that human labor remains important in the near term. They do not, however, settle how quickly the job structure will change as physical automation improves.

Source: Ars Technica AI