The AI is beginning to be trusted with high-day functions, including running automated factories and guiding military drones through hostile airspace. But when it comes to managing data centers giving power to this AI revolution, human operators are far more alert.
according to a New survey By over 600 data center operators worldwide Uptime instituteA data center inspection and rating firm, only 14 percent say they will rely on the AI system to change the equipment configuration, even if it is trained in the years of historical data. In the same survey, 1 out of 3 operators say they will rely on the AI system to control data center equipment.
Their doubt can be justified: Despite putting tens of billions of billions of US dollars in the AI system, 95 percent of the organizations have lacked a clear return to investment so far, according to the recent MIT report Liberal AI use. Advanced industries, which include factory and data centers, are ranked near the lower part of the regions converted by AI, if at all.
Operator Trust in AI Systems
Prior to AI-powered push to expand data centers, data center operators themselves are known as a relatively changes-clash, which have been disappointed with the fat technologies of the past, Rose venchenkA research colleague at the Uptime Institute. Operators often have electrical engineering or technical mechanical backgrounds, in which training in running important features; Others work on the IT or network system side and are also considered an operator.
The Operator Trust in AI declined every year for three years after the release of Openai’s chat in 2022. When asked by Uptime if they rely on a trained AI system to run data center operations, 24 percent of the respondents stated that in 2022 NO said NO said in 2024 NO said that all have said that all are cheerful.
But now, the operators have entered some data center operations in a variety of AI systems “carefully testing and verification” “said the Uptime Research Analyst Max smolux In a public webinar of the latest survey results. To catch the changing feelings, Uptime asked operators in 2025 which applications AI could serve as a reliable decision-maker, who consider previous training. More than 70 percent of operators say they will rely on AI to analyze sensor data or predict maintenance tasks for equipment, showing survey.
“The data center operator is very happy to do some things using AI, and they will never trust AI anytime on AI,” Smolak said at the webinar.
AI unpredictory in data centers
There is less confidence in AI for significant control of equipment, one reason is the unexpectedness of technology. Data centers “good, old -fashioned” are run on engineering, such as if program is/then logic, says Robert WrightChief data center officer Illkeari Data CenterA data center startup company with two centers in Colombia and Iceland. “We say that we cannot walk on luck, we have to run on certainty.”
Data center There is a complex range of systems that feed each other. Only a few seconds can pass before horrifying failures, resulting in damaged chips, wasted money, angry customer, or Malignant fireIn the high-day environment of data centers, anonymous poster on R/Datastenter Reddit Forum IEEE spectrum Question Usually AI may bring that the risk failed to see a reason to justify the risk.
Disbelief can also mask the insecurity of an underlying job. Workers in many industries are worried that AI will take their jobs. But the 2025 uptime survey found that only one of the five operators sees AI as a way to reduce the average staffing level.
“Operators believe that today’s AI is not going to change the necessary employees to run their facilities,” Smolx said to the utteime webinar. “It can come for office staff, but data center jobs seem safe to AI for now.”
But it still makes sense for initial career operators feel As this technique is coming for their jobs, says electrical engineer Jackson FahroniWho has worked in data centers for eight years. Only a six -month -old person on the job can see the AI system as being said, “Here, train their replacement,” they say. In fact, they do not think AI will replace themselves or others inside the data centers. Nevertheless, the AI machine carries more “inauspicious” appearance in the workplace than the learning tools, which long has been part of an operator’s toolkit and to help operators while taking decisions.
It may be that AI is cherry at the top of an industry-wide tendency to reduce the number of operators within data centers, says Chris McLeanA data center design and construction advisor.
While 60 engineers may have run a data center in the past, now only six are needed, Macaline says. Those require less than six, as well as, more and more important maintenance is being outsourced for experts outside the data center. “Now you offset all your risks with low -cost human and high cost AI,” McLeen said. “And I think it’s scary for operators.”
He said, there are more data centers jobs than eligible applicants, as reported earlier SpectrumAccording to the 2025 survey of the uptime, two-thirds operator struggles with staff retention or recruitment, similar to reactions from the last two years of surveys.
Ai algorithm skilled for data centers
Nevertheless, there are useful algorithms on decades of machine learning research that can make data center operation more efficient. The most installed AI system for data centers is the future stating maintenance, called the right of the illumination. If the reading of a particular HVAC unit is growing faster than those of other units, for example, the system can estimate when that unit needs to be served.
Other AI systems focus on customizing chiller plants, which are, in fact, refrigerator systems that keep the data center cool by circulating cold water and air. Chillars are mostly part of the energy consumed by data centers. Data about weather patterns, load on grids, and the decline of equipment over time, all feed all in a single AI system on hardware to adapt to all energy consumption, say, says, says, says. Michael BergerWhich runs research and development in Australia -based Energy Software Company Protection it,
But Burzer is in a hurry to keep in mind that its AI optimization does not control software devices. It moves at the top of the original control loop and refines parameters to use low energy when achieving similar results, they say. Instead of Berger AI, this system prefers to call machine learning because how specific it is for the needs of the data center.
Other people completely embrace AI, both names and technology, such as Joe MinarikChief operational officer data BankA Dallas-based data center company with 73 data centers in the US and the United Kingdom. He credits his rapid attitude towards his 17 years working for Amazon web services, where the software is King. Currently, Databank uses AI to write software, and plans to roll out the AI system for automatic ticket generation and monitoring, as well as network configuration monitoring and adjustment by the end of the year. He said that AI for large tasks, such as cooling, is temporarily determined at the end of 2026, is subject to the time taken to train AI on sufficient data, he said.
AI has done hallucinations: Minarik has seen that it gives wrong information and sends his team down the wrong path. “We do, we see that it happens today. But we also see that once we give it more time, it is better and better,” they say.
Minoric states that the key to understand the system for AI is “tremendous amounts of data points”. It is not contrary to training a human data center engineer about every possible landscape that can occur within the hall of a data center.
Hypersscalers and enterprise data centers, whose single customer is the company owner of the data center, is deploying AI at a faster pace compared to commercial companies such as Databank. The Minarik AI is hearing the system that runs the entire network for the in-house data centers.
When Databank rolls AI for more important data center operations, it will be placed on a tight strap, calling minoric. The operators will still perform the final execution.
While the AI will undoubtedly change how the data centers go, the Minaric sees operators as the main part of that new future. Data centers are physical places with on-site activity. “AI can’t walk there and change a spark plug,” he says, or listen to a odd ram from a server rack. Although Minarik says that one day some of these issues may be sensors, they will still require physical human techniques to fix the equipment that are running data centers.
“If you want a safe job that can save you from AI,” Meenaric, “visit the data centers.”
From your site articles
Related articles around web
