Citi: data center backlash hits speculative projects, not pipeline

The political heat around AI infrastructure is not new, but Citi's read is that it has so far changed the mix of projects rather than the total volume. Local governments have introduced moratoriums, and at least 15 state legislatures have proposed tighter regulatory restrictions, yet demand for AI infrastructure continues to support development spending. The split Citi draws is between speculative and early-stage projects — increasingly delayed or cancelled during local approvals — and late-stage developments with secured sites and grid connections that keep moving ahead.

The body also ties the backlash to a broader U.S.–China AI competition. The performance gap between models from the two countries has narrowed to ultra-low single digits on standard benchmarks, though U.S. developers retain the lead in frontier intelligence. Chinese developers remain constrained by access to advanced hardware under U.S. export controls, pushing them toward algorithmic efficiency and domestic chips such as Huawei's Ascend platform. That progress has not translated into stronger investor sentiment toward Chinese technology stocks, as price competition pushed inference costs lower and compressed margins, leaving investors focused on whether AI investment can produce earnings rather than benchmark gains.

For the companies and communities involved, the outcome likely hinges on whether permitting pressure stays confined to speculative builds or starts to reach projects that have already secured sites. Citi also flags a potentially greater risk — governments restricting models deemed too dangerous, which a pause in model training could turn into abrupt excess computing capacity — though Citi does not expect such restrictions in the near term.