Forecast: 80 percent of neoclouds cease operations by 2030

According to forecasts by Gartner market researchers, the arms race in AI infrastructure could lead to significant write-downs and the cessation of operations for most neocloud providers. The analysts expect infrastructure spending to reach around one trillion US dollars in 2026 and 6 trillion US dollars from 2027 to 2030. However, the expected demand for AI performance from companies and end customers will not be sufficient to generate returns on all expenditures at this investment level.

“By 2030, AI data center operators would need to generate cumulative revenues of $8.5 trillion to cover announced capital expenditures, financing costs, and a 12 percent return on investment, which we assume is the threshold for investment capital,” explains Gartner analyst Will Sommer. However, Gartner estimates the total potential revenue from AI business with corporate and private customers for this period at only up to 3.6 trillion US dollars.

Is the Neocloudocalypse Coming?

AI infrastructure providers are also holding hardware subject to rapid innovation cycles, which depreciates its value. According to Gartner, write-downs are almost unavoidable for those who fail to achieve sustainable returns on capital. By 2030, around 3.5 trillion US dollars in investments in assets could be worth less than booked on the balance sheets. By 2032, the impairment risk could be 4 trillion US dollars.

According to the forecast, the neocloud segment, i.e., cloud infrastructure providers like Nebius and CoreWeave that focus on AI workloads, will be hit particularly hard. Gartner expects radical market consolidation here. By 2030, 80 percent of neoclouds would have to cease independent operations due to excessive cost pressure and an overly one-sided business model. Gartner estimates that the market for AI-optimized IaaS will grow from 66 billion US dollars in 2027 to 174 billion in 2030 – too little to recoup investments before the next GPU generation.

It looks completely different for hyperscalers. According to Gartner, Amazon, Meta, Google, and others have enormous and broadly diversified revenue streams, allowing them to continue their AI investments for years. Whether they will actually do so if demand is clearly too low is another question. In any case, Gartner assumes an unbridled spending boom for its scenario.

A forecast by the consulting firm Bain from the end of September sees a very similar discrepancy between the investment level and the realistically achievable revenues. By 2031, the AI business must have become a market with an annual turnover of almost 6 trillion US dollars to continue investing so much. Consumer and enterprise demand alone are likely to yield significantly less, according to Bain – a massive boom in future fields like physical AI with robotics is needed.

(axk)