Hyperscalers to Spend $667 B on AI in 2026, Squeezing Free Cash Flow
Consensus forecasts now project hyperscaler capital expenditures of $667 billion in 2026, a $127 billion increase since Q4 that implies 62% year-over-year growth and will consume roughly 92% of operating cash flows. Hyperscalers including Meta Platforms will reinvest nearly all free cash flow into AI compute, data centers and networking, cutting buyback allocations from 43% to 16%.
1. Surge in Hyperscaler Capital Expenditures
Consensus estimates now put combined hyperscaler capex at $667 billion in 2026, up $127 billion since the start of the fourth-quarter earnings season, implying a 62% year-over-year increase in spending on AI infrastructure.
2. Effects on Meta’s Cash Flow
Capex slated for AI compute, data centers, networking and power capacity is set to consume about 92% of operating cash flows, forcing Meta Platforms to divert nearly every internally generated dollar back into expansion.
3. Shareholder Returns Come Under Pressure
As reinvestment escalates, hyperscalers have cut buyback allocations from roughly 43% of cash flow at the start of 2023 to just 16% today, reducing near-term shareholder distributions for firms like Meta.