Gartner has added up the AI buildout and found the revenue side short. In a report on coming AI market consolidation, the research firm estimates providers will spend nearly $1 trillion on AI infrastructure in 2026 and a further $6 trillion between 2027 and 2030. At a 12 percent return on invested capital, operators would need about $8.5 trillion of revenue over the period to cover spending and financing costs. Gartner forecasts enterprise and consumer AI revenue attributable to those data centers will total no more than $3.6 trillion.
The consequence it draws is consolidation and write-downs. Gartner’s report title warns capital write-downs could exceed $4 trillion by 2032, as newer GPUs strand older hardware, and it forecasts that 80 percent of neocloud operators, the GPU-as-a-service firms that rent AI compute, will disappear as standalone businesses by 2030, bought out or shut down. Hyperscalers are expected to weather the shake-out because AI is one business among several for them.
Gartner is not alone in the arithmetic. Bain & Company estimated last month that the industry needs $6 trillion in annual revenue by 2031 to fund the infrastructure demand implies, and Jefferies argued this week that open-source and Chinese models threaten the pricing that closed-model developers are counting on. The IMF and the Bank for International Settlements have raised parallel financial-stability concerns.
Treat the precise figures as one firm’s model, reported here from a single outlet’s account of a paywalled report. But the direction of the argument matters for names across this beat: if the payback gap is real, the first losses land on leveraged compute renters and their lenders, not on the hyperscalers signing the purchase orders.