Gartner: $3.5T of AI infrastructure faces write-downs

Gartner's math

Per the October 14 release, AI providers will spend nearly $1 trillion on AI infrastructure in 2026 and another $6 trillion from 2027 to 2030 (Gartner). To cover that, plus financing costs and a 12% return on invested capital (ROIC), AI data center owners would need $8.5 trillion in cumulative revenue through 2030.

Gartner forecasts that enterprise and consumer AI revenue attributable to data centers will reach no more than $3.6 trillion in that period. Less than half of it will go to providers that own significant infrastructure, because most AI-native providers run asset-light models.

What follows

Missing returns on fast-depreciating assets would almost certainly force write-downs. If the investment trajectory continues, more than $4 trillion of asset value could be at risk through 2032, and the 12% ROIC would not be reached before the end of 2032 at the earliest.

Neoclouds most exposed

Hyperscalers have revenue broad enough to fund AI for years, but neoclouds must earn returns on AI-optimised IaaS soon. Gartner sees that market growing from $66 billion in 2027 to $174 billion in 2030, too little for most providers to recoup their investment before the current chip generation is replaced. Analyst Will Sommer told Fierce Network the problem is not demand but monetising it amid price competition (Fierce Network).

What it means for buyers

Short term, mostly good news: price competition keeps compute cheap. Longer term, expect consolidation, so if you rely on a smaller GPU cloud, plan for exit and portability. The report landed in a nervous week for markets, with chip stocks on a losing streak.