AI Data Center Construction Continues, but Community Regulation and Power Costs Pose Risks
Democratic candidates argued that public land and power grids should not be made readily available to companies for data center construction. Their concerns included the possibility that large-scale power consumption could push up electricity bills, while local residents bear the burden of transmission infrastructure, water use, noise and heat, despite limited job creation and benefits for the local economy. Calls were also made for companies to shoulder more of the costs of land, power and transmission infrastructure and return benefits to communities.
Such an approach could delay permits for new data centers, but observers said it would be difficult to halt projects that had already received approval. Large cloud companies have already advanced much of the land rezoning and permitting process. However, tighter regulation in areas such as Texas and Ohio, where future data center construction plans are concentrated, could cause delays and additional costs across the industry.
Demand for AI semiconductors was also reflected in supply shortages in China and efforts to obtain chips through indirect channels. Smuggling methods reportedly involved separating HBM from lower-cost processors for use in China, and reports said the memory could be used in Huawei’s next-generation Ascend 960 DT. Nvidia was reported to have told new customers that at least 51% of the ownership of a data center operator must be held by parties outside China, in an effort to prevent Chinese companies from making purchases through intermediaries.