PwC's inaugural Global Data Centre Outlook, released on 2 September 2026, puts the cumulative cost of building AI infrastructure at $31.6 trillion through to 2050, a figure that exceeds the current size of the entire US economy.
The report, modelled by Oxford Economics across 46 countries and territories, projects that annual data centre capital expenditure will rise from roughly $800 billion in 2026 to $1.8 trillion by mid-century. If AI adoption accelerates beyond PwC's baseline, cumulative spending could reach $50 trillion over the same period.
PwC notes that the scale dwarfs earlier infrastructure build-outs, including railways, electrification and the rollout of the internet. Unlike those cycles, which concentrated costs upfront, AI infrastructure investment is expected to keep growing because chips, servers and networking equipment require replacement every few years.
The US is forecast to capture the largest share, at $15.1 trillion, or nearly half the global total. Asia-Pacific follows at $8.2 trillion, with China and India identified as major sources of incremental demand, driven by large populations and expanding digital economies. Europe is projected at $5.6 trillion, the Middle East at $1.1 trillion, and Africa at $255 billion.
Hardware, particularly AI chips from companies such as Nvidia, will account for the bulk of spending. PwC estimates that ICT equipment's share of total investment will rise from 70 percent today to 93 percent by 2050, as recurring upgrade cycles, rather than land or construction, define the capital pattern.
Power supply, data sovereignty and semiconductor supply-chain stability are flagged as the key variables that could alter the trajectory. PwC's report notes that supply-chain disruptions alone could reduce global investment by nearly 20 percent.
