Meta tests robots to replace data techs
Archive item — written before sources were shown.
Meta is piloting robotic arms and simple actuators from Kinova, ABB, and Watney Robotics to handle server power-cycling and cable swaps inside its data centers.
Meta is testing robots inside its data centers to handle tasks currently done by human technicians, including power-cycling servers and swapping networking cables, according to current and former workers describing previously unreported projects. The company is working with hardware from several vendors, among them Kinova, ABB, and Watney Robotics.
In one test, Meta is evaluating whether a Kinova Gen3 robotic arm can power-cycle or cut electricity to servers. A separate robot is being tested for cable-swapping. One worker estimated that, if the cable-swapping project succeeds, it could replace up to 80% of some technicians’ current workload. At some sites, Meta has also quietly deployed a simpler device, essentially a mechanical finger that presses a physical power button on a Mac Mini or similar device when remotely triggered. Meta declined to comment on the specific testing described, though a spokesperson confirmed the company is exploring automation in its facilities. Kinova and ABB declined to comment; Watney did not respond to requests for comment.
The stakes for operators, not just Meta
Data-center labor is a small line item next to the billions Meta is spending on AI compute itself, which is exactly why it is worth noting: even at that scale, the company is testing whether physical maintenance work can be automated out. That mirrors the logic behind Caterpillar bringing mining-automation lessons into general AI deployment and it is a preview of where the automation pressure AI infrastructure creates eventually points inward, at the facilities running the models, not just at white-collar work the models perform. It also sits alongside Nvidia’s $3.5 billion bet on keeping custom chips inside its own infrastructure standard, another sign that the physical layer under AI, not just the models on top of it, is where a lot of the real competitive maneuvering is happening right now. If you operate or lease data-center capacity, the maintenance-labor cost you’re pricing in today may not hold for long.
- 01Inside Meta's push to put robots to work in data centersarstechnica.com · reporting (WIRED syndication)
