Inside Meta's push to put robots to work in its data centers

[QuietObserver34]

Meta is quietly testing robots inside several of its data centers to handle physical maintenance tasks that have historically required human technicians, according to current and former workers who spoke with reporters. The robots, sourced from vendors including Watney Robotics, Kinova and ABB, are being trialed for jobs like swapping networking cables, power cycling frozen servers, reseating hardware and inspecting equipment at facilities in Iowa and Ohio. One worker estimated that if the cable swapping robot specifically proves successful, it could eventually replace up to 80 percent of some technicians' actual daily workload. The project hasn't been previously reported and Meta hasn't made any public announcement about it

The technology still has genuine limits though, with the robots reportedly struggling around corners and dense cabling, and their cameras apparently unable to distinguish between green and red status indicators on equipment. Battery recharging downtime and the need to manually move some robots between buildings add further practical friction. Meta says it continues hiring and training workers given an industry wide skilled labor shortage, even while running these automation pilots in parallel

Workers describe genuine anxiety about what this means long term, with concerns that automating the easier physical tasks first could eventually push remaining work toward lower paid staff following AI generated instructions rather than eliminating jobs outright immediately. Curious what people think this actually signals for data center employment more broadly as AI infrastructure spending keeps accelerating

Share your take below on whether this feels like inevitable progress or a more troubling pattern. Genuinely curious how people weigh the efficiency argument against the job security concerns here

Inside Meta's push to put robots to work in data centersarstechnica.com

[Sofia_61]

The two stage pattern described here, automate the easy tasks first then push the remainder to lower paid workers following AI instructions, is exactly the deskilling pattern economists have described in manufacturing automation for decades. Nothing genuinely new about the mechanism, just a new industry experiencing it. Feels almost predictable in hindsight

[EdgeNode Joel]

Struggling to tell red from green status lights is such a small sounding limitation but honestly speaks to how far these systems still are from full autonomy in messy real world environments. Data centers sound clean and controlled from the outside but clearly aren't simple enough yet for current robotics. Real world deployment always exposes these small gaps

[DarkKnight66]

Meta simultaneously hiring more skilled workers while running these automation pilots feels less contradictory than it sounds honestly, since the actual timeline for full robot reliability is clearly still years out based on the described limitations. Companies hedge both directions during exactly this kind of transitional period. Doesn't mean the long term trajectory isn't still toward fewer human jobs eventually though