Overview
- Morgan Stanley projects roughly $1.58 trillion of global AI compute investment in 2027 and warns hyperscalers and neoclouds could burn about $600 billion in free cash flow that year.
- The bank expects Nvidia to capture roughly two‑thirds of S&P 500‑related AI capex — about $690–700 billion — while the company has helped assemble financing platforms with major asset managers to mobilize more than $500 billion.
- Market signals of strain are visible: unsecured bonds and shares of neoclouds such as CoreWeave have plunged into distress territory and at least 47 large U.S. data‑centre projects have been delayed, withdrawn or blocked.
- On‑the‑ground bottlenecks are growing: grid connection delays, local permitting pushback and skilled‑labor and power shortages mean analysts judge only about one‑third of the planned U.S. data‑centre pipeline as credible.
- The likely outcomes are concrete — tighter credit will pressure Asian hardware suppliers and semiconductor stocks, pension and institutional investors will take on new infrastructure risk, and many announced projects may be delayed, scaled back or impaired.