Gartner: $3.5 trillion of AI could lose its value by 2030

Research firm Gartner estimates that more than $3.5 trillion invested in AI infrastructure could be written down by the end of 2030. The maths is simple. AI players will spend nearly $1 trillion in 2026, then another $6 trillion between 2027 and 2030. To make all that pay off, they would need $8.5 trillion in revenue, while Gartner expects no more than $3.6 trillion.

The problem is that chips age faster than they pay back what they cost. They will be replaced by the next generation before they have covered their price. According to Gartner, more than 80% of “neoclouds”, those specialised GPU rental companies, will no longer exist independently in 2030. Hyperscalers, on the other hand, have other businesses to absorb the shock.