Bain: AI compute demand needs $6 trillion by 2031

Two new reports from consulting giants mark a shift in the AI adoption conversation. Until recently, the debate over AI safety has somewhat dwarfed the still urgent discourse over AI's macroeconomic impacts. The focus is now moving toward spending sustainability, as pilot programs pile up and agentic applications take off.

Bain & Company's report this week puts a hard number on the industry's ambitions: compute demand will require $6 trillion in annual revenue by 2031, with about $4.2 trillion expected to come from new markets and products that don't exist yet. Separately, McKinsey's report on American government entities describes a squeeze for buyers. For many governments, usage might have been essentially near-zero cost until now, according to McKinsey Senior Partner and Global Public Sector Leader Tim Ward — but very soon, it won't be.

The coming IPOs from Anthropic and OpenAI will surely bring more debate over whether their revenue can catch up to the labs' lofty valuations. Meanwhile, credit rating agency Egan-Jones wrote Wednesday that "the complete disruption of the economy is all but certain," in a note titled It's Over. Whether the spending proves sustainable appears to hinge on whether buyers keep paying as agents drive total bills higher even while per-token prices fall.