Via racksolutions.com
AI-driven construction spending hit $45.1B a month by December 2025, and the bottlenecks are just getting started
The AI infrastructure race has a spending problem, and it is not that companies are spending too little. Monthly US data-center construction spending hit $45.1 billion by December 2025, an 85% jump from just two years prior.
Amazon Web Services, Microsoft, Google, Meta, and Oracle are collectively projected to spend approximately $1.15 trillion in capital expenditures from 2025 through 2027, according to Goldman Sachs estimates.
The driver is AI compute demand, which requires far denser and more power-hungry hardware configurations than conventional cloud workloads. Deloitte forecasts that US AI data center power demand will grow from 4 GW in 2024 to 123 GW by 2035, a more-than-thirtyfold increase in roughly a decade.
The supply chain is not keeping up #
Constructing a data center requires gas turbines, transformers, switchgear, cooling systems, and miles of specialized cabling, all sourced from manufacturers with long production queues. US gas turbine production runs at around 10 GW annually when allocated to commercial operation dates. Transformer lead times have stretched dramatically, and switchgear manufacturers are operating at capacity.
Between 30% and 50% of data center capacity planned for 2026 could face delays tied to construction and electrical grid bottlenecks.
Industrial real estate as an unlikely beneficiary #
Every gigawatt of data center construction generates approximately 2 million square feet of industrial and warehouse demand, with roughly 80% of that footprint tied to long-term operational needs rather than construction staging.
Capacity under construction in the Americas reached 25.3 GW as of early 2026. Pre-commitment rates on new data center capacity sat at nearly 89%, meaning almost all space coming online was already spoken for before the first server rack arrived. Vacancy rates fell to around 4.2%.
Site selection is now driven less by traditional factors like tax incentives or labor costs and more by one overriding question: is there power available, and when.
What this means for investors and the broader economy #
A data center that cannot receive its transformers on schedule cannot generate revenue on schedule. When financing costs are factored in, a six-month delay on a large campus development can materially erode project returns.
Power-rich states and municipalities that have proactively invested in grid capacity are attracting data center investment at a pace that compounds their advantage. Areas without that infrastructure are effectively locked out of the current cycle, regardless of how aggressively they compete on tax breaks or permitting speed.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our