Kevin O’Leary says data centres use less water than golf courses. The numbers are more complicated.

The market implication is not that water suddenly stops mattering; it’s that water is probably not the binding constraint in the AI buildout. If public opposition gets redirected away from water usage, the discount rate on hyperscale permitting should improve modestly, which is supportive for data-center landlords and operators whose biggest risk has been local veto power rather than economics.

The real second-order effect is a shift in the political fight from water to power and land-use. That favors names with exposure to electrical infrastructure, substation buildouts, and interconnect spend, while the utility-side winners are the ones that can monetize grid upgrades without facing a capex cap. By contrast, ESG-sensitive investors may be forced to reframe their bear case on AI from "resource intensity" to "grid congestion," which is a harder argument to win over 1-3 months.

Contrarian view: consensus may be overestimating how much this changes the regulatory path. Water is a visible headline issue, but the harder bottleneck remains transmission and community acceptance, so the move is more sentiment than fundamental—unless a state starts mandating water-impact disclosure or moratoriums. The clean falsifier is a fresh permitting delay tied to water or a cut to data-center capex guidance; absent that, this is a modest de-risking event, not a thesis changer.