In a data centre the overriding fear is anything that touches uptime, so cooling water is run by an incumbent regime no one revisits, even as its cost and its risk rise.
Cooling, not the servers, is where the controllable cost sits. At a typical efficiency, around a third of total facility energy is overhead, dominated by cooling, and cooling failure is the second most common cause of impactful outages, more than half of which now cost over 100,000 dollars each.
Water has become the public problem. Most new UK capacity is planned for the most water-stressed parts of the country, and 84% of water-intensive proposed developments sit in areas already short of water or projected to be by 2040. Operators are increasingly asked to justify their water draw at the planning stage, in public.
Wet cooling towers sit in the highest Legionella risk category by design, under ACOP L8 and HSG274. Yet the water-treatment regime is usually an annual compliance contract no one wants to touch, because changing it feels like risking uptime, so the cost and the water keep climbing and no one is asked to own all three together.