The rapid expansion of data centers, crucial for the AI boom, presents a significant test for US industrial resolve. China and the US are leading the charge, accounting for roughly 70% of global data center energy consumption in 2024. Projections indicate global installed data center capacity will reach 220 gigawatts by 2030, a six-fold increase from 2020. This massive build-out is driven by the race among hyperscalers to capture demand growth and secure quick returns on rapidly depreciating assets, with companies like OpenAI, Anthropic, Google, and Meta heavily investing in AI model training and operation. The energy consumption of a single data center can be equivalent to that of 100,000 households, with the largest under construction expected to be 20 times bigger.
This insatiable demand for power is problematic, as the distribution of data facilities often does not align with energy supply. Data centers accounted for 1.5% of global electricity consumption in 2024, growing at 15% annually over the prior five years. The International Energy Agency forecasts this consumption could more than double from 415 terawatt hours in 2024 to 945 TWh by 2030, while McKinsey projects an even higher demand of 1,400 TWh by 2030, representing 4% of the world's consumption. The US acknowledges that its AI ambitions are hampered by an aging and inadequate electrical grid, with the White House's AI Action Plan emphasizing the need for grid stabilization and optimization.
Local resistance is also a significant hurdle, with conflicts arising over public water use and wildlife habitats. While some areas like Reno, Nevada, have attracted data centers with fast-tracked permits, other regions are seeing stalled projects and legislative efforts to slow construction. For example, Maine lawmakers passed a bill to ban data centers drawing over 20MW until 2027, though it was vetoed. Georgia, Virginia, and Oklahoma are also considering temporary bans, reflecting a broader public reluctance to have data centers in their neighborhoods. Approximately 70% of respondents in a recent Heatmap poll preferred not to have a data center built near them.
The deployment of infrastructure is piecemeal, depending on private and local incentives, and is constrained by varying permitting regulations across states and counties. Beyond electricity supply, shortages of labor and materials are impacting construction timelines. Up to 40% of projects due for completion in 2025 and 26 are at risk of falling behind schedule due to these shortages. The challenges underscore a critical test for the US in balancing its leadership aspirations in AI with the practicalities of massive infrastructure development and environmental concerns.
