JPMorgan expects global AI capital expenditure to reach $870 billion by 2026, with investment logic shifting towards commercial monetization.

A report from JPMorgan on July 20 estimates that global AI-related capital expenditures will reach nearly $870 billion in 2026, a year-on-year increase of 77%, with hyperscale vendors contributing about $750 billion. In 2027, Google's capital expenditure is expected to be about $300 billion, Amazon's about $300 billion, and Meta's about $200 billion. The five major tech giants’ bond financing will rise from $40-50 billion in 2022 to about $190 billion in 2026, and Google completed an $85 billion equity financing in June. JPMorgan believes that the logic of AI investment is shifting from competing on capital expenditures to competing on commercial monetization, and that, in the medium term, the risk-reward ratio for hyperscale vendors is better than for AI hardware. Semiconductor revenue growth excluding storage this year is expected to exceed 30%, but the hardware sector has recently corrected by about 15%. Google, Microsoft, Meta, Amazon, and NVIDIA will successively announce their financial results, and capital expenditure guidance will be a key indicator.