AWS warns US data-center restrictions could constrain AI infrastructure expansion

AWS CEO Matt Garman has called for the US to avoid broad restrictions on data-center construction, arguing that continued build-out is tied to the country’s AI competitiveness, economic position and national security.

In a blog post of more than 3,000 words, Garman said that more than 100 measures to pause data-center construction are under discussion across the country. Against that backdrop, Amazon said it intends to invest more than $1 billion over the next five years in communities where it operates data centers. The figure is a future commitment, rather than confirmed spending already deployed.

The pressure on project pipelines is illustrated by Calvert County, Maryland. AWS withdrew a plan in August for a data center of about 223,000 square metres, following local opposition, according to the source. The county commission subsequently imposed a six-month pause on approvals for new data-center projects. The withdrawn proposal was a planned facility, not an operational asset.

Data Center Watch estimates that at least 120 US data-center projects were delayed or impeded by local opposition in the first half of 2026, with an estimated aggregate value of $198 billion. The source does not disclose the methodology, project mix or status of the projects covered by that estimate.

For operators and infrastructure investors, the development risk is increasingly not limited to power availability, land and grid interconnection. Community acceptance and local permitting timelines could become decisive variables in site selection and campus delivery. Amazon’s proposed community spending indicates that host-community investment is becoming part of the industry’s response, although it may not prevent project withdrawals or local approval pauses.

The source also reports that Amazon invested $276 billion in data centers between 2011 and 2025 and planned $220 billion in capital expenditure in the current year, with a significant share intended for AI infrastructure. It does not specify the precise portion of that annual CAPEX allocated to data centers or AI infrastructure.