BofA expects hyperscale cloud capex to exceed $860bn in 2026 (up ~80% YoY) and approach $1.2trn in 2027 (up ~38%). Microsoft, Oracle, AWS and Google report a combined cloud backlog of more than $2.3trn, up 16% QoQ, providing high visibility for future spend. The current AI infrastructure expansion will chiefly boost demand for compute, storage, semiconductor equipment, power semiconductors and optical communications. As AI agents and inference workloads scale, server allocations of HBM, DRAM and

BofA expects hyperscale cloud capex to exceed $860bn in 2026 (up ~80% YoY) and approach $1.2trn in 2027 (up ~38%). Microsoft, Oracle, AWS and Google report a combined cloud backlog of more than $2.3trn, up 16% QoQ, providing high visibility for future spend. The current AI infrastructure expansion will chiefly boost demand for compute, storage, semiconductor equipment, power semiconductors and optical communications. As AI agents and inference workloads scale, server allocations of HBM, DRAM and SSDs should rise; storage demand could outpace GPU shipments. Capex exceeding operating cash flow will compress free cash flow at big tech firms, though they can access debt and equity markets. BofA characterises the cycle as a multi-year infrastructure build and says rising cloud orders improve visibility on AI investment returns.