AI infrastructure faces potential write-downs above US$3.5 trillion
Gartner also forecasts that more than 80% of neoclouds, which rent AI computing capacity, will no longer operate independently by 2030.
Maintaining the current pace of AI infrastructure investment could leave more than US$3.5 trillion exposed to write-downs by the end of 2030, Gartner warned on 14 October. The risk concerns assets that cannot generate enough income to recover their cost.
The research firm estimates that AI providers will spend nearly US$1 trillion on infrastructure in 2026 and another US$6 trillion between 2027 and 2030. To meet those costs, cover financing and achieve an assumed 12% return on invested capital, data center owners would require US$8.5 trillion in revenue through 2030.
Gartner puts the associated revenue from business and consumer AI at no more than US$3.6 trillion during that period. It also expects owners of substantial AI infrastructure to collect less than half of AI revenue, noting that most AI-native providers have asset-light operations.
The firm identified particular risks for neoclouds, businesses that rent computing power built for AI. It forecasts that more than 80% will lose their independent status by 2030, citing cost pressures and insufficiently distinctive business models. By contrast, hyperscalers could draw on varied revenue streams to support investment for years.
Spending on AI-optimized cloud infrastructure is forecast to reach US$174 billion in 2030, up from US$66 billion in 2027. Gartner senior director analyst Will Sommer said that increase would not be enough: most AI computing providers would fail to recover their investment before newer chips replace those currently in use.
