JP Morgan says Q2 AI company revenues beat expectations and the economic
sustainability of the AI capex cycle has improved versus six months ago. Its
internal credit research projects $5.5 tln of AI-related capex in 2026–2030;
external estimates reach up to $10 tln. At the midpoint (~$7.5 tln), JP Morgan
models combined revenue for AI cloud, model providers and new neocloud services
at $1.6 tln by end-2026; with 10–20% annual growth thereafter, 2030 revenue
would be $2.5–3.0 tln. Demand is concentrated in enterprises: large firms’ AI
spend as a share of total opex+capex is expected to rise from 4.5% over the past
12 months to 5.8% over the next 12 months — against roughly $30 tln of global
large-enterprise opex+capex this implies about $1.7 tln of AI spend. JP Morgan
says the bearish claim that AI infrastructure far exceeds monetization capacity
has weakened, but margins, depreciation, utilization, pricing power and
intra-supply-chain returns still require validation. (JP Morgan report, Aug. 20)