Enterprises Will Carry $1.8 Trillion of the $2.6 Trillion AI Bill
Enterprises, not consumers, will pay for the AI buildout. BCG’s October 29 analysis puts end-market spending on AI infrastructure and labs at about $2.6 trillion a year by 2030, excluding China, and assigns roughly $1.8 trillion of it to enterprises. Consumers account for $366 billion, hyperscalers’ internal use for $204 billion and emerging categories such as physical AI for $200 billion. Consumer AI apps, at around $37 billion, cover only about 1.4% of the requirement.
For a company planning its own AI budget, the number that matters is share of technology spend. BCG projects enterprise AI rising from 5% of tech spending today to 25% in 2030 — about $857 billion reallocated from existing budgets and $937 billion added on top. That is a reallocation story as much as a growth story: someone’s line item shrinks to fund it, and the CFO will ask whose.
On the supply side, BCG has five hyperscalers’ AI capex climbing from $116 billion in 2022 to a projected $866 billion in 2030, peaking at 4.0% of GDP against 6.6% for railroads in 1872 and 1.1% for the internet in 2000. The authors say the build is funded mainly from operating cash flow rather than debt, and that inference — 71% of lab costs by 2030 — makes the system more resilient. They also say the figures are not a revenue forecast: if demand fails to appear, price wars and losses would hit both the model and data center layers.
Other estimates run larger because they measure something different. Bain’s Technology Report 2026 counts $5 to $6.5 trillion to build roughly 150 gigawatts of new compute by 2030, and flags grid connection delays of more than four years. Gartner forecasts $2.7 trillion of worldwide AI spending in 2026 alone, up 49.5%.
The practical read: vendors are planning around enterprises renewing at scale. Tie each AI line item to an outcome you can name now, before the reallocation conversation arrives.