AI needs $6T in annual revenue to justify data centre boom

Bain & Company projects the AI industry must generate about $6 trillion in annual revenue by 2031 to justify the massive capital now going into data centres. It breaks that revenue into roughly $4.2 trillion from new product development (search, advertising, autonomy, physical AI), $1-1.4 trillion from enterprise productivity gains (software development, sales, marketing, customer service, IT) and $200-400 billion from consumer services like subscriptions and advertising. Annual AI infrastructure spending could reach $1.5 trillion by 2031 for new facilities, GPU, memory and networking upgrades; if capex runs at roughly 25% of revenue - a level seen among cloud providers - the market needs to expand to the $6 trillion scale. Bain also highlights “absorption speed” as a competitive variable, noting top AI labs are investing billions in engineering models to help firms adopt capabilities faster.

Data-centre sizes and costs are accelerating, roughly doubling every 12-16 months, with Epoch AI estimates showing Meta’s Prometheus project growing from 600MW/$24bn in 2025 to as much as 9GW/$200bn by 2030. Scaling raises challenges: grid upgrades, chip and component bottlenecks, talent retention, regulatory and public pushback over resources and noise. Several governments (UAE, Saudi, EU, South Korea, US) are backing capacity buildouts, making sovereign infrastructure and public-private partnerships key strategic and investment levers as the industry searches for the new markets and innovations to deliver the necessary trillions in value.