Overview
- Bain & Company estimates the global AI industry must reach about $6 trillion in annual revenue by 2031 to justify current data-center build plans, leaving roughly $4.2 trillion of new revenue to be created beyond existing AI services.
- Bain projects $5 trillion to $6.5 trillion in data-center spending by 2030 and says annual AI infrastructure outlays could peak near $1.5 trillion by 2031, adding at least 150 gigawatts of compute capacity that will increase strain on energy systems.
- Goldman Sachs forecasts the five largest U.S. hyperscalers—Amazon, Alphabet/Google, Microsoft, Meta, and Oracle—will spend about $1.2 trillion on AI infrastructure in 2027 and estimates $7.6 trillion in cumulative hyperscaler capex from 2026 through 2031.
- Analysts identify clear bottlenecks that could slow or shrink usable capacity, including power generation and grid upgrades, transformer and water shortages, skilled labor limits, and constrained supply of high-bandwidth memory chips from vendors like Nvidia, SK Hynix, and Samsung.
- Financing is shifting into bonds, private credit, and special-purpose vehicles, which increases opacity about who bears losses and raises the risk of project write-downs that could affect banks, insurers, pension funds, local communities facing stalled projects, and power grids under heavier load.