US AI Data-Centre Power Crunch and Grid Costs

17 Oct 2026

The rapid expansion of US AI data centres is creating a power and infrastructure bottleneck that could constrain computing growth even when chips are available. Developers may face a substantial electricity shortfall through 2028 as demand rises faster than generation, transmission, substations, and grid connections can be built. The resulting policy fight centres on whether hyperscale operators or household ratepayers should pay for dedicated upgrades and stranded costs, with lawmakers considering standards for large electricity users. Local communities are also demanding greater transparency about projects negotiated with utilities and governments, including their effects on power, water, land, subsidies, and roads. What remains at stake is who captures the economic benefits of AI expansion and who bears its infrastructure and environmental costs.