US AI infrastructure investment projected to reach $10.3 trillion by 2032

The United States is entering an unprecedented period of investment in artificial intelligence infrastructure, with spending on data centres, power systems, networks and specialised equipment projected to reach $10.3 trillion between 2025 and 2032.

The estimate comes from Columbia University economist Stijn Van Nieuwerburgh in research presented through the Brookings Institution. The projected spending would amount to an average of 3.63 per cent of US gross domestic product each year over the period.

The scale of the projected investment is being compared with some of the largest infrastructure buildouts in US history. Brookings estimates that railroad investment averaged 2.24 per cent of GDP between 1870 and 1890, while highway investment averaged 1.13 per cent between 1956 and 1973. Telecommunications and fibre investment averaged 1.10 per cent between 1996 and 2003.

Earlier infrastructure programmes included canal investment, which averaged 0.66 per cent of GDP between 1836 and 1841, and electrification, which averaged 0.50 per cent between 1905 and 1925, according to the same comparison.

The AI projection covers a broad range of infrastructure. It includes data centre buildings, power systems, networking infrastructure and specialised chips and other equipment required to develop and operate AI systems.

The forecast reflects expected investment rather than money already committed or spent. The Brookings analysis also highlights uncertainty around how quickly planned data centre capacity will actually be built, meaning the final level of investment could differ from the current projection.

The spending boom is already reshaping capital investment among major technology companies. Recent analysis cited by Brookings points to rapidly rising expenditure by companies including Amazon, Microsoft, Alphabet, Meta and Oracle as they expand computing capacity and related infrastructure.

The scale of the buildout is also raising questions about how it will be financed. Brookings notes that an increasing share of AI infrastructure financing is moving through structures such as joint ventures, private credit, special-purpose vehicles and lease arrangements, creating additional questions around transparency and exposure to changes in AI demand.

For now, the $10.3 trillion figure represents a forecast of where the US AI infrastructure buildout could go by 2032. If the projection is realised, its annual share of the US economy would exceed the levels recorded during several previous major infrastructure investment cycles.