Data center operators racing to keep pace with AI workloads are running into a problem that has little to do with chips: their maintenance schedules were built for a slower, simpler industry.
That’s the finding of a new white paper from IDC, sponsored by Schneider Electric, which argues that fixed-interval upkeep, the industry standard for decades, can’t keep up with facilities where rack densities have jumped from roughly 15 kilowatts in a standard hall to as much as 600 kilowatts in AI-heavy compute zones.
The paper, titled “The Self-Aware Datacenter,” makes the case for condition-based maintenance, or CBM, an approach that services equipment based on real-time signals rather than a fixed calendar.
“Your maintenance schedule doesn’t know when something is failing — your equipment does,” said Luis Fernandes, senior research manager at IDC and the paper’s author.
The shift comes as operators increasingly grow through acquisition rather than new construction, inheriting equipment from multiple vendors with little or no service history. “When operators acquire existing facilities rather than build from scratch, they introduce unknown equipment configurations from multiple vendors, with no operational history, requiring immediate integration with asset performance management systems,” Fernandes said.
Labor is compounding the problem. IDC’s paper cites a technician shortfall so severe that in the U.S., roughly seven open positions exist for every qualified candidate, a gap spanning electrical and mechanical engineers as well as commissioning specialists certified to work on high-voltage systems.
Surveys across the sector have repeatedly flagged skilled-labour shortages as a top operational risk, even as demand for new capacity shows no sign of slowing.
Schneider Electric’s answer is EcoCare, a services line that pairs AI-enabled monitoring with remote human oversight to track equipment against operating thresholds and flag deviations before they turn into failures.
“By combining remote monitoring capabilities with AI-assisted orchestration, you can gain insights regarding the health of your assets and systems, and get an early identification of abnormal behaviour that might precipitate a failure,” said Jerome Soltani, global head of services at Schneider Electric. “This ensures that downtime is minimised, but also that equipment that is working within specification is not disturbed or needlessly addressed.”
That last point cuts against a long-standing industry habit: opening up equipment on a fixed schedule regardless of whether it needs attention, a practice that can introduce its own risks.
IDC’s research found early adopters of AI-driven CBM logging fewer manual interventions, lower operating costs and less unplanned downtime, alongside longer equipment lifespans. Schneider Electric cites steeper figures for its own EcoCare customers — up to a 75 percent reduction in unplanned downtime and a 20 per cent cut in operating expenses, though those numbers come from the company rather than independent verification.
The broader argument in IDC’s paper is that predictive maintenance, done at scale across a fragmented, multivendor fleet, amounts to something closer to institutional memory for a data center, a system that learns how each asset behaves over time rather than treating every unit the same.
“Condition-Based Maintenance is an optimised operating model for AI-era infrastructure that reduces manual interventions, lowers OpEx, and extends asset lifecycle,” Fernandes said. “By scaling predictive analytics to correlate behaviour across every vendor, asset, and failure trajectory, CBM enables operators to build machine-driven, human-validated system intelligence.”
The white paper, “The Self-Aware Datacenter: How Condition-Based Maintenance Turns Fragmented, Multi-Vendor Datacenters into Predictable Infrastructure and System Intelligence” (IDC #META54611526), was published in June 2026 and is available through Schneider Electric
Join BusinessDay whatsapp Channel, to stay up to date
Open In Whatsapp
