Market research firm Gartner warned that the AI bubble could burst. It projected that if AI infrastructure companies keep investing at the current pace, they may have to book more than $3.5 trillion in losses by the end of 2030.

According to a recent report by Techzine, AI companies are expected to spend nearly $1 trillion on infrastructure in 2026. They plan to invest an additional $6 trillion from 2027 to 2030.

AI data centre operators expect cumulative revenue of $8.5 trillion by 2030.

But Gartner estimated that at the current growth rate, revenue from AI applications for businesses and consumers will total only $3.6 trillion. That would leave a gap of nearly $5 trillion.

That could increase the risk that companies will need to reduce the book value of infrastructure.

Gartner estimated that if the current investment trend continues, more than $4 trillion in assets could be subject to write-downs by 2032. It forecast that the industry would also struggle to achieve a target return on invested capital (ROIC) of 12 percent before the end of 2032.

The hit is expected to be concentrated among neo-cloud companies. According to Network World, hyperscalers have far more diversified business structures than neo-cloud companies and can also withstand large-scale investments for years to raise market share.

Gartner predicted that as costs rise and differentiation becomes harder, more than 80 percent of neo-cloud companies will be unable to operate independently by 2030.