That is according to Kelsey Nowicki, technical consulting manager for data centres at Ecolab.

The dissolved calcium in the municipal water reacted with phosphate-based corrosion inhibitors in the glycol coolant, forming calcium phosphate scale throughout the system. Within around two weeks, the scale had begun plugging cold plates, restricting flow and cutting cooling performance.

The entire loop had to be drained, cleaned, flushed and refilled with fresh glycol and deionised water. Details of the site and operator were not disclosed. Nowicki said the case illustrates a wider industry problem: D2C loops are still often treated as conventional facility plumbing rather than controlled process systems.

“It was a costly reminder that D2C cooling loops must be treated as controlled process systems, not conventional facility plumbing,” she said.

Narrow margins, faster failures

Nowicki said TCS loops carry more risk than traditional chilled water systems for two structural reasons. Frequent hardware and blade change-outs create repeated opportunities to introduce contaminants, dissolved oxygen or microorganisms. And many TCS loops run at temperatures more favourable to microbial growth than the lower temperatures typical of chilled water systems.

Combined with narrow flow passages and higher heat-transfer surface area per unit volume, she said, small amounts of fouling or biological growth become operationally significant faster. Problems that might take months to surface in a traditional chilled water loop can become noticeable within weeks in a TCS environment, according to Nowicki, though she cautioned there is no universal timeframe as it depends on operating conditions and coolant chemistry.

The pre-commissioning gap

The most common failure point, Nowicki said, is inadequate cleaning and preparation before startup: poor chemical cleaning, insufficient flushing velocity, or failing to fully drain the system before the final fill. Residual water can dilute propylene glycol concentration below target range from day one.

Caught during commissioning, the fix is typically additional cleaning and flushing, adding a few days to the schedule. Caught after startup, remediation can mean draining and refilling the loop, replacing coolant and cleaning fouled components while equipment is offline, at significantly higher cost.

Why periodic sampling falls short

Nowicki argued that AI-density D2C environments have little margin for error. Every additional percentage of propylene glycol reduces heat-transfer efficiency and increases pumping energy, while concentrations that are too low raise the risk of biological growth. That narrow operating window, she said, is what makes periodic sampling inadequate, since it leaves operators blind to changes between tests.

Continuous monitoring, by contrast, “typically requires a significantly smaller investment compared to the potential costs of a cooling-related event,” Nowicki said.

Addressing the conflict of interest

Ecolab sells water treatment and monitoring services into the same market it is warning about. Asked to address scepticism that the risk is being overstated to sell contracts, Nowicki said TCS loops are not truly closed systems in practice, given that routine maintenance and blade change-outs create repeated opportunities for contamination.

“These risks are not theoretical,” she said. “They are based on real-world outcomes where contamination, fouling, improper coolant chemistry, or biological growth have resulted in performance degradation, remediation efforts, and costly downtime.”

RELATED STORIES

Data centres could use 2.7% of Texas’s water by 2030: Where they land matters

How to pass the test with data centre interconnects