Overview
- JPMorgan’s midyear update published Thursday projects total global AI and data‑center capital spending of about $5.5 trillion through 2030 and raises its debt financing estimate to roughly $4.1 trillion.
- The firm expects hyperscalers to sharply ramp investment, forecasting about $650 billion of capex in 2026 and more than $1.1 trillion in 2027 which will drive most demand for servers, GPUs and memory.
- JPMorgan documents 122 gigawatts of planned new data‑center capacity for 2026–2030 and notes construction already reached an annualized $40 billion run rate by mid‑2025, creating heavy demand for electricity and cooling.
- Analysts flag near‑term risks from high leverage and elevated loan‑to‑cost ratios above 85 percent, plus supply bottlenecks in high‑bandwidth memory and optical networking that could slow deployment or hurt returns.
- Markets are adapting with new finance and operating models such as long‑term compute contracts, leasing and repurposing high‑power Bitcoin‑miner sites to host AI servers, and these moves will shape who benefits from the supercycle.