Gartner Puts $3.5 Trillion Write-Down Tag on AI Infrastructure Buildout

The math has been nagging at infrastructure watchers for two years. Gartner put a number on it Tuesday.

In a report published October 14 and reviewed by TechTarget's IT Infrastructure desk, the research firm estimated that AI providers are on pace to spend nearly $7 trillion on data centers through 2030, roughly $1 trillion in 2026 alone, with another $6 trillion projected between 2027 and the end of the decade. To clear a 12% return on that capital, the industry would need cumulative revenues of $8.5 trillion. Gartner's own revenue forecast for AI over that period tops out at $3.6 trillion. The gap is not subtle.

That leaves more than $3.5 trillion in infrastructure assets exposed to impairment by 2030, a figure Gartner said could climb past $4 trillion by 2032. The report's title, cited by Crypto Briefing in its coverage, skips any hedging: "Prepare for AI Market Consolidation: Capital Write-Downs Could Exceed $4 Trillion by 2032."

The category Gartner is most worried about is neoclouds, the wave of AI-compute-specific providers, companies like CoreWeave and Nebius, that scaled on GPU density and customer demand for AI-optimized infrastructure-as-a-service rather than diversified cloud revenue. Will Sommer, senior director analyst at Gartner, told TechTarget that the firm's position is that 80% of neoclouds will cease independent operations by 2030. He was careful to clarify that "ceasing independent operations" covers acquisition and partnership, not just outright failure, but consolidation at that scale still represents a structural reset.

The mechanism isn't hard to trace. Sommer noted in remarks also reported by Benzinga that because most AI-native providers run asset-light models, less than half of total AI revenue will actually land with companies that own significant physical infrastructure. That's the core squeeze: the people bearing the capital risk aren't the people capturing most of the revenue.

There's a hardware depreciation problem layered on top. As TechTarget noted in its write-up of the report, neocloud providers are holding assets subject to rapid innovation cycles. If a provider can't achieve sustainable returns before the current GPU generation is superseded, the book value of that hardware doesn't hold. Sommer put it plainly in remarks cited by Seeking Alpha: "The growth in spending on AI-optimized IaaS will be too low for most AI compute providers to recover their AI investments before the current generation of AI semiconductors is replaced."

Hyperscalers, Amazon, Google, Microsoft, are in a different position. As noted by TradingView's coverage of the report's market impact, their revenue diversification insulates them from the neocloud math. They'll likely absorb capacity or acquire providers that can't hold on independently, which is consistent with Sommer's framing of how the consolidation plays out.

Markets didn't wait for the nuance. CoreWeave's stock dropped roughly 15% by midday Wednesday, with other neocloud names following, according to TradingView's reporting.

The Gartner number is a model, not a certainty. As Dealroom noted in its own summary, outcomes depend heavily on demand trajectories, power availability, and pricing dynamics. What the report does is put a credible floor on the conversation about whether this buildout pencils out, and for most neocloud operators, the answer, at current spend rates, is that it doesn't.

Sources cited:
- TechTarget IT Infrastructure News (https://www.techtarget.com/it-infrastructure/news/366652181/Gartner-forecasts-massive-AI-infrastructure-write-downs)
- Crypto Briefing (https://cryptobriefing.com/gartner-ai-infrastructure-write-down-risk/)
- Benzinga (https://www.benzinga.com/markets/equities/26/10/62381272/crwv-nbis-iren-in-focus-gartner-warns-80-of-neoclouds-will-cease-independent-operations-by-2030)
- TradingView / Seeking Alpha (https://www.tradingview.com/news/seekingalpha:ba53b19f6094b:0-coreweave-collapses-after-gartner-predicts-80-of-neoclouds-will-evaporate-by-2030/)
- Dealroom (https://app.dealroom.co/news/note/gartner-over-80-of-neoclouds-may-cease-independent-operations-by-2030-as-3-5t-of-ai-infra-faces-impairment-risk)
- Heise Online (English) (https://www.heise.de/en/news/Forecast-80-percent-of-neoclouds-cease-operations-by-2030-11487979.html)