Data centre backlash goes national as protests hit 42 US states

This is more relevant as a permitting/regulatory signal than as a demand signal. In the next 1-3 months, the market usually overweights headline opposition, but the economic effect is mostly slower site conversion, higher legal/land costs, and more concessions to host communities rather than outright cancellation of AI capacity. That tends to compress returns for greenfield developers while pushing hyperscalers toward fewer, larger, better-connected sites.

The second-order winner is the infrastructure stack that makes constrained builds work: transmission, switchgear, grid services, and retrofit-heavy colocation. Names with existing footprint and power access should gain relative advantage versus developers depending on cheap land and local political goodwill. Regional lenders with CRE or construction exposure to data-center-adjacent projects are a quieter loser; if entitlements stretch, draws get delayed and loan spreads can look fine until the pipeline suddenly stalls.

The contrarian point is that the consensus may be mistaking visible local activism for an AI capex demand shock. The true bottlenecks are interconnection queues, transformer lead times, and utility delivery, so protest pressure mainly changes geography and timing, not the end-demand curve. A real reversal would require state-level moratoria, election-driven zoning changes, or evidence that hyperscalers are deferring capex rather than rerouting it; absent that, the selloff risk is more in speculative build-to-suit names than in the broader AI supply chain.