US Cloud Providers Commit Up to $690 Billion for 2026 AI Buildout

3 articles · Updated · InfoWorld · Oct 16

$660 billion to $690 billion in 2026 capex would nearly double what the five biggest US cloud and AI infrastructure providers spent a year earlier, with Amazon alone projected at $200 billion.
Two-thirds of that spending is aimed at compute—GPUs and custom accelerators—while the rest goes to data centers, cooling, networking and increasingly power infrastructure, including on-site generation where grids fall short.
An $80 billion Azure backlog underscores why the buildout is accelerating: hyperscalers say they are supply-constrained, not demand-constrained, as enterprises rush to secure AI capacity and hardware.
Shortages of processors, RAM and cloud capacity are likely to persist through the next two years, supporting vendors such as Dell, HPE and IBM as silicon and memory supply is absorbed by AI clusters.
By 2028 and 2029, though, efficiency gains or weaker-than-expected AI adoption could leave some 2026-committed capacity underused, echoing past overbuild cycles in fiber and early cloud.