Gartner Puts $3.5 Trillion of AI Infrastructure Spending at Risk of Write-Down by 2030

The math that the AI infrastructure build-out runs on has always had a gap in it. Gartner put a number on that gap this week, and it's large enough to matter to anyone who owns, operates, or finances a data center focused on AI workloads.

In a report published October 14 and reviewed by TechTarget, Gartner estimated that more than $3.5 trillion in AI infrastructure investments could be exposed to asset impairment by the end of 2030, with that figure potentially climbing past $4 trillion by 2032 if current spending trajectories hold.

The mechanics are straightforward. According to Gartner's own figures reported by TechWire Asia, AI providers are on track to spend close to $1 trillion on infrastructure in 2026 alone, followed by another $6 trillion between 2027 and 2030. To earn a 12% return on invested capital from that cumulative outlay, the industry would need to generate roughly $8.5 trillion in revenue over the same period. Gartner projects that enterprise and consumer AI revenue attributable to data centers will reach no more than $3.6 trillion through 2030. That's a $4.9 trillion shortfall before you factor in depreciation.

Gartner senior director analyst Will Sommer, whose projections were cited by TechTarget and Stocktwits market reporting, identified neoclouds as the most exposed category. Neoclouds are the generation of compute providers that built their businesses specifically around GPU-dense, AI-optimized infrastructure, rather than growing out of broader cloud businesses with diversified revenue. Sommer put the failure rate plainly: "We [take the] position that 80% of neoclouds will cease independent operations by 2030." He was careful to note that ceasing independent operations includes acquisition and partnership, not only outright failure, but the consolidation pressure is real regardless of the exit path.

The structural problem Gartner is naming isn't new to anyone who has watched hardware investment cycles before. AI computing infrastructure depreciates against a rapid chip replacement cycle. A GPU cluster bought today to run current-generation inference workloads won't hold its book value if the next architecture generation makes it significantly cheaper to run the same workloads in two years. When an asset can't generate the returns it was bought to produce, companies have to mark down its value, and that hits reported earnings directly.

Hyperscalers sit in a different position. As Gartner noted, AWS, Azure, and Google Cloud carry massive diversified revenue bases that can absorb AI infrastructure investment losses in ways a pure-play neocloud simply can't. The hyperscalers have a longer runway to wait for AI revenue to materialize. Gartner's own analysis suggests hyperscalers could reach a 12% return on invested capital by 2032, two years after the window closes for many neoclouds.

The market reacted quickly to the report's release. Shares in publicly traded neocloud operators including CoreWeave, Nebius, and IREN fell on the news, according to Stocktwits market coverage, which cited the Gartner forecast directly.

What the report doesn't do is predict that AI demand itself collapses. Gartner's own numbers show AI-optimized infrastructure-as-a-service spending rising from $66 billion in 2027 to $174 billion in 2030. That's meaningful growth. The problem isn't that the market stays flat. The problem, as Hosting Journalist noted in its coverage of the report, is that revenue growth and capital recovery are different calculations, and the industry has been treating them as though they're the same one.

For enterprise buyers, the consolidation forecast has a practical implication that doesn't get discussed enough in the vendor pitches: neocloud providers that lose independent operations mid-contract create a migration problem. The compute disappears or changes hands; the workloads still need to run somewhere. That's a procurement risk worth pricing now, before a write-down announcement forces the conversation.

Sources cited:
- TechTarget / IT Infrastructure News (https://www.techtarget.com/it-infrastructure/news/366652181/Gartner-forecasts-massive-AI-infrastructure-write-downs)
- TechWire Asia (https://techwireasia.com/2026/10/gartner-ai-infrastructure-spending-risk/)
- Gartner Newsroom (https://www.gartner.com/en/newsroom/press-releases/2026-10-14-gartner-says-over-us-dollars-three-and-a-half-trillion-of-artificial-intelligence-infrastructure-spending-could-be-exposed-to-impairment-by-the-end-of-2030)
- Stocktwits / Markets (https://stocktwits.com/news-articles/markets/equity/crwv-nbis-iren-stocks-tumble-after-gartner-forecasts-80-of-neocloud-market-will-cease-to-exist-by-2030/cZxfxbXRBqZ)
- Hosting Journalist (https://hostingjournalist.com/news/gartner-warns-of-3-5-trillion-ai-infrastructure-impairment-risk-by-2030)