메타가 자체 데이터센터에서 케이블 연결, 서버 재부팅, 전원 차단 등의 작업을 수행하는 로봇을 테스트 중인 것으로 확인됐습니다. AI 인프라 투자가 급증하는 가운데 인건비를 억제하고 확장 중인 데이터센터를 소수의 인력으로 운영하는 것이 목표입니다. 일부 로봇은 특정 직무의 최대 80% 작업을 대체할 수 있는 것으로 평가되고 있습니다.
번역된 본문
내부 관계자들에 따르면 메타는 자사 데이터센터 내에서 케이블을 연결하고, 서버를 재부팅하며, 기타 작업을 수행하는 로봇을 테스트하고 있다. 지금까지 알려지지 않았던 이 프로젝트가 성공하면 메타는 급속히 확장 중인 데이터센터를 더 적은 인력으로 운영할 수 있게 되어, AI 인프라 지출이 급증하는 상황에서 인건비를 통제할 수 있게 된다.
익명을 요구한 관계자들에 따르면 메타는 Watney Robotics, Kinova, ABB 등 여러 벤더의 로봇 및 관련 하드웨어를 사용하고 있다. Kinova와 ABB는 논평을 거절했고, Watney는 요청에 응답하지 않았다.
한 실험에서 메타는 Kinova Gen3 로봇 팔을 서버 전원 차단(파워 사이클링)에 사용할 수 있는지 평가하고 있다. 또한 다른 로봇으로 네트워크 케이블을 교체하는 테스트도 진행 중이다. 한 메타 데이터센터 근무자는 이 로봇이 성공할 경우 일부 인력의 업무량 최대 80%를 대체할 수 있을 것으로 추정했다. 그는 "물리적 작업을 하는 우리는 한동안 안전하다고 생각했지만 이제 아니다. 안타깝게도 로봇이 우리 모두를 노리고 있다"고 말했다.
일부 시설에서는 메타가 비교적 단순한 또 다른 로봇을 조용히 도입했다. 손가락, 혹은 뾰족한 막대처럼 생긴 이 로봇은 사람이 원격으로 지시하면 맥 미니나 다른 기기의 전원 버튼을 눌러 재시작할 수 있다.
메타는 이 테스트에 대한 논평을 거절했다. 회사 대변인 프랜시스 브래넌은 성명에서 메타가 데이터센터 건설과 운영 인력을 채용하고 교육하는 데 대규모 투자를 하고 있다고 밝혔다. 그는 "미국은 2차 대전 이후 최대의 인프라 붐을 겪고 있으며 이를 채울 숙련 인력이 크게 부족하다. 우리는 더 적은 게 아니라 더 많은 인력이 필요하다"고 말했다.
하지만 작년 한 컨퍼런스에서 메타의 로봇 담당 수석 매니저인 에릭 슈는 회사의 "장기 목표"는 데이터센터에 로봇을 투입하는 것이라며, 로봇이 사고 대응 속도를 높이고 환경을 모니터링하며 예방 정비를 처리할 수 있다고 주장했다. 그는 "더 많은 협업과 연구가 필요하지만, 시작하지 않으면 이룰 기회조차 없을 것"이라고 말했다.
최근까지 로봇은 데이터센터에서 이루어지는 섬세한 작업에 적합하지 않았다. 로봇은 비쌌고, 업계의 초기 시험에서는 간단한 작업을 시도하다 서버를 부수는 일도 있었다고 실험 내용을 전해들은 한 관계자는 말한다. 하지만 하드웨어 가격이 내렸고, 로봇을 구동하는 AI 모델도 몇 년 전보다 훨씬 능력이 향상됐다. 현재 여러 스타트업이 서버 조립, 부품 교체, 배선 교체가 가능한 상업용 로봇 개발에 나서고 있다.
로봇은 인력 대비 낮은 비용으로 더 일관된 성과를 낼 수 있으며, 특히 데이터센터는 많지만 자격을 갖춘 인력이 부족한 인구 밀도가 낮은 지역에서 그렇다. 지지자들은 친노동자적 메시지를 내세우기도 한다. AI 에이전트가 지식 노동자에게 약속하듯, 로봇도 데이터센터 기술자의 가장 반복적이고 위험한 작업을 맡아주어 더 복잡한 과제에 집중할 수 있게 해준다는 것이다.
장기적으로는 인력을 배제할 경우 데이터센터를 수중, 우주, 생명체에게 적대적인 기타 환경에서 운영하는 것이 더 현실적으로 될 수 있다. 폴 골딩은 "점점 더 많은 고객이 '휴머노이드 로봇으로 데이터센터를 운영하고 싶다'고 말한다. 더 높은 온도에서, 어둠 속에서 등 모든 환경에서 운영할 수 있기 때문"이라고 말했다.
Comment Loader Save Story Save this story Comment Loader Save Story Save this story Meta is testing robots that can plug in cables, reset servers, and handle other tasks inside its data centers , according to several current and former workers familiar with the projects. The ongoing effort, which has not been previously reported, may eventually allow Meta to operate its rapidly expanding data center footprint with fewer humans, keeping labor costs in check as its spending on AI infrastructure soars . Meta is using robots and related hardware from several different vendors, including Watney Robotics, Kinova, and ABB, according to the same workers, who asked to remain anonymous because they weren’t authorized to speak to the media. Kinova and ABB declined to comment. Watney didn’t respond to requests for comment. In one experiment, Meta is evaluating whether a Kinova Gen3 robotic arm could be used for power cycling, or cutting off electricity to servers. The company is also testing a different robot to swap networking cables. One Meta data center worker estimates that if it’s successful, the bot could replace up to 80 percent of some people’s workloads. “We thought those of us performing the physical tasks were safe for a while, but not anymore,” says the worker. “It’s coming for us all, unfortunately.” At some facilities, Meta quietly adopted another, relatively simple robot. It looks like a finger—or perhaps a pointy stick—that can press the power button on a Mac Mini or other device to restart it when remotely prompted by a human. Meta declined to comment on the testing described in this story. Company spokesperson Francis Brennan said in a statement that Meta is investing heavily in training and hiring workers to build and operate its data centers. “America is in the middle of its biggest infrastructure boom since World War II and there’s a major shortage of skilled workers to fill the roles; we need more workers, not fewer,” he said. But at a conference last year , Eric Xu, senior manager for robotics at Meta, said the company’s “long-term goals” are to put robots in data centers, arguing they could speed up incident response, monitor the environment, and handle preventative maintenance. “We believe more collaboration and research will be needed, but we have to start otherwise we don't have a chance to do this,” he said. Until recently, robots largely weren’t well suited for the delicate work happening in data centers. They were expensive, and in earlier trials in the industry, sometimes crushed servers while attempting simple tasks, a person briefed on the experimentation says. But the hardware has gotten cheaper, and the AI models powering robots are now far more capable than they were a few years ago. Several startups are currently working to build commercial robots capable of assembling servers, swapping parts, and replacing wires. Got a Tip? Do you currently work at a data center or have previously and want to talk about what's happening in the industry? We'd like to hear from you. Using a nonwork phone or computer, contact the reporter securely on Signal at peard33.24 Robots could offer more consistent output at a lower cost than human labor, especially in less populous regions where data centers are plentiful but qualified workers are scarce. Proponents also rally behind a pro-worker message: Like AI agents promise to do for knowledge workers, they argue robots could take on data center technicians’ most repetitive and dangerous tasks while freeing them to focus on more complex challenges. Longer term, getting workers out of the picture could make it more feasible to run data centers underwater, in space, or in other environments potentially inhospitable to living, breathing species. “More and more of our customers are telling us ‘we want to run our data centers using humanoid [robots]’ because you can run them at higher temperatures, in the dark, and all these things,” says Paul Golding, who oversees so-called physical intelligence at the chipmaker Analog Devices. The growing use of robots has received little attention from the activists driving a powerful nationwide movement against data centers. Tech companies often tout jobs for humans as an upside to the developments that can be used to justify tax breaks for them. In Altoona, Iowa, Meta operates what is now its largest data center campus , which employs hundreds of workers. The company opened the facility in 2014 and has several years left of a property tax break that allows it to save tens of millions of dollars annually , which the city has viewed as a worthwhile tradeoff for having its cachet upped by having Meta jobs. In Altoona, Meta worker group chats have taken on a demoralized tone, “saying we’re all going to be gone in a few years anyway because of the robots,” a person familiar with the discussions says. Even if adoption of robots proves slower than anticipated, some worry the quality of jobs available is trending downward. Two of the workers say that Meta is developing AI software and other technologies that they believe will allow it to hire lower-skilled—and thus, lower-paid—workers and centralize some roles in areas such as Denver, Colorado that are more financially advantageous to the company. “They no longer want people who can think independently, come up with creative solutions, or perform complex troubleshooting,” one Meta data center worker says. “They want ‘smart hands,’ people who are just capable enough to follow AI instructions and not mess anything up.” Meta contends its need for skilled workers remains robust. 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 even guaranteeing employment in states such as Louisiana, Ohio, Indiana, and Texas for those who complete the training. It also partnered with building trade unions to fund apprenticeship programs. But as communities rethink data center tax breaks and weigh putting new rules on the industry, the prospect of robots replacing at least some jobs may need to become a bigger concern. Altoona Mayor Dean O'Connor tells WIRED he hasn’t given much thought to robots taking jobs at Meta’s data center, though he frames automation as inevitable. “We will deal with it as it comes,” he says. Tripping Cables In a 2023 YouTube video, Meta touted that it was using a self-driving tugger robot to transport heavy server racks within its data centers as well as a wheeled, barcode-reading robot developed in-house to track inventory. The next year, Microsoft and Google unveiled separate investments in robots for their data center operations, and Amazon last year discussed using robots to help recycle parts at its facilities. Pursuing robotics for “economically valuable tasks,” like “replacing a faulty server rack” is “absolutely something we're looking into at Microsoft,” Ashley Llorens, a managing director for the company’s research accelerator, tells WIRED. Meta’s tugger and inventory robots are now operating in a number of the company’s data centers, including in Iowa and Virginia, according to the people familiar with the projects. They say the robots generally work well, though there are limitations. The inventory bot is now also being used for failure inspections, but it can’t differentiate between green and red indicator lights on equipment because its camera only detects grayscale, meaning human checks are still required to assess some breakdowns. The device also struggles with corners and traversing the cables that are often in the paths of its wheels. Because there aren't that many of them, the robots need to be moved between buildings, which requires a door-opening human who maneuvers the robot with a remote control. Two sources also describe the inventory bot as too slow. One says Meta previously explored using drones as a faster alternative, but ultimat