Americans are rallying against data centers. Surprisingly few are actually getting built

Key Points:

• Despite widespread public opposition and political resistance, AI data center construction faces significant delays and challenges, with only about half of the planned AI computing capacity expected to come online by target dates through 2028.
• Key obstacles include shortages of building materials, critical AI chips (notably from Taiwan’s TSMC), power supply constraints, labor shortages, and regulatory hurdles such as permit approvals and local moratoriums in several states.
• The U.S. currently has over 5,400 data centers, with nearly 4,000 more announced but only 802 under construction; however, experts estimate that only about one-third of planned AI data center capacity will actually be built due to feasibility issues.
• Despite these challenges, investment in data center construction remains robust, with spending rising 7% in June to $68.3 billion and contributing to inflation pressures, though the complexity and scale of projects mean significant risks of overinvestment and project failures.
• The anticipated AI data center boom represents a massive infrastructure expansion, potentially involving $10 trillion in investment over the next decade, but the industry must overcome supply chain bottlenecks, labor shortages, and regulatory barriers to realize this growth.