What happened
PwC released its inaugural Global Data Centre Outlook on September 2, 2026, projecting that global investment in AI infrastructure will reach a record "US$31.6 trillion through to 2050," with annual data-centre capex rising from roughly $800 billion in 2026 to $1.8 trillion by 2050. Modeled by Oxford Economics across 46 countries and territories, the outlook finds the US will capture 48% ($15.1 trillion) of cumulative investment, with ICT equipment's share of spend rising from 70% today to 93% by 2050 as AI compute requires recurring upgrade cycles rather than one-time build-out. The report runs two scenario tests: tighter export controls could cut cumulative investment to ~$25.5 trillion (roughly $6 trillion less than baseline) by disrupting chip supply chains, while a digital-sovereignty scenario redirects capital toward domestic infrastructure without materially reducing total spend. PwC identifies power availability, connectivity, security, policy certainty, and GPU access as the five factors that will determine where capital lands.
Why it matters
This is the first long-range, scenario-tested capex forecast tying AI compute demand directly to geopolitical trade-policy outcomes, giving boards and infrastructure investors a quantitative basis for assessing exposure to export-control and sovereign-AI policy shifts.
Action needed
Incorporate the export-control and digital-sovereignty scenarios into capital planning and data-centre siting strategy reviews.