Meta Is Testing Data-Center Robots—but Human Technicians Are Still Essential

Meta’s push to automate its AI infrastructure has moved beyond software. The company is testing robots inside data centers for routine physical work, including cable connections and server resets, while separately pursuing a much broader plan to make its workforce “AI native.”

The business case is clear: Meta expects to spend at least $130 billion on AI infrastructure and chips in 2026, with estimates reported as high as $145 billion. 2 9 But the early robotics work also shows why automation is unlikely to eliminate the need for skilled data-center specialists soon.

What Meta’s data-center robots are being tested to do

According to reporting based on current and former workers familiar with the projects, Meta has tested robots and related hardware from companies including Watney Robotics, Kinova, and ABB at facilities such as Altoona, Iowa, and New Albany, Ohio.

The reported tasks include:

• Plugging in and swapping network cables
• Power-cycling servers
• Cutting power to equipment
• Pressing physical restart buttons
• Handling other repetitive hardware actions performed by data-center technicians 9

These are narrowly defined jobs with predictable physical steps. That makes them more realistic early targets for robotics than the complex installation, diagnosis, and repair work required across a large AI facility.

One worker estimated that a successful cable-swapping robot could eventually handle as much as 80% of some technicians’ workloads. That figure is an estimate of potential coverage, not a demonstrated result showing that robots are already performing 80% of the work. 9

The robots’ capabilities remain limited

The reported experiments do not describe fully autonomous data centers. The machines are slower than human technicians, require supervision and charging, and use relatively limited grayscale-camera systems. They also cannot perform intricate installation work such as laying the dense cabling needed for Nvidia GB300 systems.

That distinction matters. A robot that can repeat a standardized cable or restart procedure may reduce the amount of routine work technicians perform. It does not necessarily replace the people who decide what to do when equipment behaves unexpectedly, coordinate a complicated deployment, or work safely around high-powered electrical and computing systems.

Meta’s longer-term robotics ambitions are broader than simple button pressing. Its robotics leadership has described possible uses including faster incident response, environmental monitoring, and preventative maintenance in data centers. Those goals suggest an incremental model: robots handle more standardized actions while people remain responsible for supervision, exceptions, and complex work.

Meta says it still needs skilled data-center workers

The robotics tests sit alongside Meta’s stated investment in training and hiring people to build and operate its data centers. The company has framed the rapid expansion of AI infrastructure as creating a shortage of skilled labor and has said it needs more workers with the relevant expertise, not simply fewer workers. 15

That position may appear to conflict with automation, but the two strategies can coexist. Hiring specialists addresses the immediate labor bottleneck, while robotics could gradually increase the amount of infrastructure each specialist can oversee.

In that scenario, robots would function less like a wholesale replacement for technicians and more like a way to extend their capacity—especially for repetitive work that consumes time but does not require much judgment.

How the robot tests connect to Project OT

The physical-automation effort reflects the same broader question behind Project OT, Meta’s proposed “Organization Transformation.” Internal planning explored an “AI native” company in which AI agents would take over work performed by thousands of employees, with smaller human teams supervising automated workflows. Some teams were evaluated for potential reductions of as much as 60%. 1 6

The plan did not proceed in its original form. Meta ultimately carried out roughly 8,000 layoffs, while a planned second company-wide wave scheduled for November was canceled shortly before the first cuts began. 7

Reporting attributed the retreat to a combination of employee opposition and disappointing evidence that AI systems were not producing the expected productivity gains. 1 3 Meta described Project OT as scenario planning rather than a finalized plan affecting every team. 2

That history provides an important reference point for the robotics program. The question is not whether Meta wants to automate; it clearly does. The harder question is whether the technology can deliver reliable productivity gains in real operating environments, where unusual failures and safety requirements matter as much as routine speed.

Why Meta has a strong incentive to automate data-center operations

Meta’s infrastructure spending makes even modest efficiency improvements strategically valuable. The company reported $31.1 billion in quarterly capital expenditures in the second quarter of 2026, driven by servers, data centers, and network infrastructure. 14 It also said it expected to invest at least $130 billion in AI infrastructure and chips during the year, while other reporting put the full-year spending range at $130 billion–$145 billion. 2 9

As the physical footprint expands, labor becomes one part of a much larger operating challenge. Robots could potentially:

• Perform routine actions consistently
• Reduce the amount of time specialists spend on repetitive tasks
• Help respond to incidents more quickly
• Allow a limited number of technicians to monitor more equipment
• Reduce some long-term operating costs if the systems become reliable and economical

But capital intensity also raises the standard for success. A robot that is expensive, slow, difficult to supervise, or unable to cope with exceptions may add complexity rather than remove it. The cost of a mistake can also be high when the equipment supports large-scale AI workloads.

The likely near-term outcome is a hybrid workforce

The available evidence supports a narrower conclusion than “robots are replacing data-center technicians.” Meta is testing whether robots can automate standardized physical actions, while its broader AI strategy seeks to redesign knowledge work around smaller teams and automated agents.

For now, those ambitions are constrained by practical performance. The robots reportedly need human oversight and cannot handle the most complex cabling work. Project OT, meanwhile, was scaled back after its expected productivity gains failed to materialize at the required level. 1 3

The most plausible near-term model is therefore hybrid: robots handle repetitive tasks, AI software coordinates or assists with workflows, and skilled technicians manage complex installations, safety, diagnosis, and exceptions. Meta’s spending creates a powerful reason to keep pursuing that model, but its own workforce experiment shows why automation plans must be validated in operations—not just on a planning document.