Bain's $6tn AI maths says staff productivity alone cannot pay for the data centre boom

The National reported on Tuesday that the AI industry needs to earn $6 trillion a year by 2031 to justify what it is spending on data centres. The figure comes from a new Bain & Company technology report. Bain forecasts that annual spending on AI infrastructure might hit $1.5 trillion by 2031. It also says capital spending should run at about a quarter of industry revenue, which it calls "an ambitious but reasonable percentage". At that ratio, the market has to approach $6 trillion a year. The headline number reads as a question for investors. For an operator, the more useful figures sit in how Bain splits it.

Productivity is the smaller bet

Bain splits the $6 trillion into three parts. New products, covering search, advertising, autonomy and physical AI, carry about $4.2 trillion. Consumer subscriptions and advertising add $200 billion to $400 billion. Enterprise productivity, the part most workers mean when they say "AI at work", needs $1 trillion to $1.4 trillion across software development, sales, marketing, customer service and IT operations.

David Crawford, chairman of Bain's global technology practice and the report's lead author, said: "The debate today is fixated on employee productivity. The economics of AI infrastructure demand trillions in new revenue beyond productivity gains."

Read that from inside a company. Crawford says the debate is fixated on employee productivity. Bain's arithmetic says cost savings alone cannot pay for the buildout. Most of the money has to come from new products: Bain puts that segment at about $4.2 trillion, and says some of the products and uses behind it "don't exist today". It names drug discovery, mental health and energy generation.

That changes what your employer is buying. A firm that uses AI only to trim costs is going after the smaller slice of Bain's estimate. The firms that pull ahead will use it to build new offers and win new markets. That work needs people who can spot a new market and sell into it. AI does the routine work. You do the thinking.

Absorption speed is the new variable

Bain's second finding is closer to the desk. It calls absorption speed, meaning the pace at which companies can put AI to work, the "new competitive variable". According to the report, leading AI labs are investing upwards of $9.75 billion in engineering models to help companies take up AI faster.

If speed of adoption is what separates companies, much of it is a human problem. Someone has to redesign the process, decide which tasks go to a system and which stay with a person, and get a team to trust the output and check it. The person who does that work controls how quickly the organisation moves.

The scale explains the pressure. Bain says data centres are doubling in size and cost approximately every 12 to 16 months. Figures from the research firm Epoch AI, cited alongside the report, follow Meta's Prometheus site in Ohio from 600MW at an estimated $24 billion in 2025 to a projected 9GW at up to $200 billion by 2030. Capital on that scale needs customers to show results. Expect your employer to feel that pressure and pass it down.

Your Next Move

• Pick one process in your team that AI could redesign end to end. Write a one-page plan: which tasks move to a system, which stay with a person, and who checks the output. Whoever writes that plan leads the team's AI adoption.
• Tie your AI work to revenue or a new offer as well as to hours saved. Bain puts new products at about $4.2 trillion and enterprise productivity at $1 trillion to $1.4 trillion. Work that feeds the first figure carries more weight in the room where budgets are set.
• If your organisation has started measuring adoption, read our guide on showing impact when your employer grades your AI use before your next review.