Gartner: $3.5T in AI spend may be written off if value never shows

Gartner used its latest forecast to put a hard number on a soft worry: by the end of the decade, more than three and a half trillion dollars of the money pouring into AI data centers may never be recovered. The arithmetic is blunt. To cover the capital already announced, plus financing and a reasonable return, those facilities would have to generate on the order of eight and a half trillion dollars in revenue through 2030, yet Gartner’s own forecast sees barely more than three and a half trillion flowing back.

The gap is not a rounding error, and it lands hardest on the operators with no other business to lean on; the firm expects most standalone neocloud providers to be gone by 2030, while the hyperscalers ride it out on profits earned elsewhere. Spending, meanwhile, keeps climbing - close to a trillion dollars this year and several trillion more still to come.

Strip away the scale and this is the same question every finance chief now faces one invoice at a time: is the money going into AI actually coming back out as value? That is the entire reason Uprovd exists. We meter what AI costs, attribute it to an owner and an outcome, and make the return provable - so spend is defended by results rather than written off as a bad bet.