The rapid build-out of US data centers to power AI is facing significant roadblocks, with at least 47 large projects being delayed, withdrawn, or blocked in 2026, according to regulatory intelligence firm RegLayer. The pushback stems from a combination of insufficient power grid capacity, regulatory hurdles, and increasing local community opposition to the massive energy and water consumption of these facilities.
Adding to the construction challenges, a recent Bernstein report casts doubt on the viability of the current development boom, estimating that only about 35% of the planned US data center pipeline is credible. This confluence of physical limitations and financial skepticism suggests the unprecedented expansion of AI infrastructure is running into the unforgiving economics of the physical world, potentially slowing the pace of AI development.