The AI industry needs about $6 trillion a year by 2031 just to cover the data centers it is building right now.
That is the headline number from Bain & Company's Global Technology Report, published Tuesday. The consultancy estimates that current AI products and services will bring in somewhere between $1.2 trillion and $1.8 trillion of that total, leaving a gap of roughly $4 trillion to $5 trillion a year. Bain argues that today's conversation about AI's payoff is too narrow, focused mostly on employee productivity gains rather than the scale of revenue actually required to justify the infrastructure buildout. The report does not say where the rest of the money is supposed to come from.
That gap matters because it puts a number on a worry that has mostly stayed vague: the data centers, chips, and power contracts going up today assume a level of AI revenue nobody has demonstrated yet. If productivity tools alone cannot close a multi-trillion-dollar hole, the industry needs new products, new customers, or new pricing that do not exist yet.
None of this proves the buildout will not pay off eventually. But it is worth remembering that the firm making this forecast also sells strategy advice to the companies building those data centers, and consultancies have a long history of pairing big scary numbers with big future consulting engagements.