To achieve a 12% return on their AI infrastructure investments through 2030, data centre operators would need to generate $8.5 trillion in revenue, according to analysis by Gartner. However, the research firm forecasts that only $3.6 trillion in AI-related revenue tied to data centres will materialise during that period. The gap between these figures puts more than $3.5 trillion of spending at risk of impairment, Gartner warned on 14 October, assuming AI providers maintain their current investment trajectory. When assets fail to generate sufficient returns, companies must write them down on their balance sheets.
The outlook for independent cloud operators offering AI computing capacity looks particularly bleak. Gartner predicts that more than 80% of neoclouds—firms specialising in renting computing infrastructure designed for artificial intelligence—will lose their independence by 2030. The firm attributes this consolidation to mounting cost pressures and the absence of differentiated business models.
All of them want to be the full-stack. They can't all be.
Will Sommer, Gartner analyst, to Bloomberg
The revenue gap
Gartner's senior director analyst Will Sommer outlined the scale of the challenge in the firm's latest research. AI providers are expected to invest nearly $1 trillion on infrastructure in 2026 alone, with a further $6 trillion allocated between 2027 and 2030. To cover these expenditures, associated financing costs, and deliver a 12% return on capital—the threshold Gartner uses as an investment hurdle rate—operators would require $8.5 trillion in cumulative revenue. Yet the firm projects only $3.6 trillion in business and consumer AI revenue connected to data centres during this window.
Our research indicates that enterprise and consumer AI revenue will be insufficient to support profitable returns on all planned AI capital expenditures during this period.
Will Sommer
Most AI-focused providers operate on asset-light models, meaning Gartner expects less than half of AI revenue to flow to firms that own substantial infrastructure assets. If current spending patterns persist, asset write-down exposure could exceed $4 trillion through 2032, and the 12% return target may not materialise until the end of 2032 at the earliest. Sommer collaborated with Mark McDonald and John Lovelock on the research underpinning these projections.
Neoclouds are the most exposed
Large hyperscalers possess diversified revenue streams that can sustain AI spending for extended periods, according to Gartner. Neoclouds face a tighter timeline: they must generate returns on their AI-optimised infrastructure quickly to remain viable. Spending on such infrastructure is projected to climb from $66 billion in 2027 to $174 billion in 2030—growth that Sommer characterised as insufficient. Most AI compute providers will struggle to recoup their investments before newer chip generations render current hardware obsolete.
Bloomberg identified CoreWeave and Nebius as specialist providers facing greater exposure than major cloud operators such as Microsoft, Alphabet, and Amazon. CoreWeave experienced sharp declines in its share and bond prices this week amid concerns about its AI-related debt.
Many neoclouds aspire to offer the higher-margin software and services that traditional cloud providers deliver, Sommer noted to Bloomberg. However, few are likely to succeed in this transition.
For every dollar that's spent on compute, only a fraction of that is going to the owners of compute.
Will Sommer, to Bloomberg
When speaking with The Register, Sommer offered a more nuanced assessment of the sector's prospects.
Not everyone is in trouble. Some will outperform, some will underperform, and some will have to exit the market.
Will Sommer, to The Register
Other warnings
Gartner's analysis aligns with earlier findings from Bain & Company, which estimated that the AI sector requires $6 trillion in annual revenue by 2031. This week, the International Monetary Fund also raised concerns about circular AI transactions.
Source: The Next Web
