Voters rejected AI data centers over resource concerns. Here's why measuring water usage is hard

Proposed 1GW data centers from OpenAI, Meta, SpaceX, Amazon, Google and others were a key issue in the midterms, amid concerns about power and water use. Liquid cooling can limit water use inside many modern AI data centers, but external cooling in hot climates can still require substantial power or water.

Analysis

The investable risk is not a sudden technology constraint; it is a higher and less predictable cost of getting large projects permitted and connected. As political attention translates into local review, developers may face longer siting timelines, stricter water conditions, or pressure to fund grid upgrades. That can shift project economics even where on-site liquid cooling limits water use: cooling demand still has an electricity cost, while renewable contracts do not by themselves guarantee power at the required time and location.

For Amazon, Alphabet, and Meta, the likely near-term exposure is execution friction and incremental infrastructure cost, not enough evidence here to revise consolidated earnings. A bigger second-order effect is that costs may fall on ratepayers or utilities; disputes over who pays for transmission and generation could become the permitting bottleneck. Electrical-equipment, cooling, and water-infrastructure suppliers could benefit if projects proceed with more infrastructure, but delays would defer orders too.

Time horizon: near-term sentiment risk is modest absent a concrete rule or project delay; the next 1–3 months hinge on local permitting and policy proposals, while 6–18 months could bring site selection toward areas with dependable power and less water stress. Contrarian point: treating liquid cooling as a complete resource solution misses the external power and siting constraints. Conversely, campaign attention alone is not proof of a broad regulatory clampdown. No directional hyperscaler trade is justified without project-level evidence.

Key Decisions for Investors

• Avoid a broad short of Amazon, Alphabet, or Meta on this signal alone; the article supplies no enacted restriction, named project delay, or quantified earnings exposure.
• Track local permitting decisions, utility interconnection queues, water-use conditions, and who is assigned grid-upgrade costs. Escalate to a negative view only if these produce specific schedule slippage or capex guidance changes.
• Keep electrical-infrastructure, cooling, and water-infrastructure suppliers on a conditional watchlist, not an unconditional long: order growth depends on projects proceeding, and delays can defer the same demand.
• Falsifiers: accelerated approvals or credible evidence that projects can secure firm power and cooling capacity without material added cost would weaken the constraint thesis; repeated permit denials or hyperscaler guidance citing power, water, or connection delays would strengthen it.