Only A Quarter Of AI's $4.7 Trillion Prize Comes From Productivity

Bain & Company estimates that $4.7 trillion of global corporate profits are at stake from artificial intelligence between 2025 and 2035, in a study published on 8 September that scored 92 sectors.

The figure covers profits created in new categories, shifted between competitors, or lost by firms that move slowly. Bain puts it at roughly a fifth of total projected global corporate profits in 2035, against $1.4 trillion of profit shifts from the Internet over the twenty years to 2015.

The firm expects AI to structurally transform 71 per cent of sectors, compared with 41 per cent during the Internet era. Its explanation for the difference is that the Internet was a distribution technology that cut the cost of reaching customers, while AI is a production technology that cuts the cost of making the product.

Three forces account for the total. Innovation, meaning new AI-enabled products and business lines, contributes about $2.2 trillion or 47 per cent. Redistribution of existing profits as cost structures and capabilities change contributes roughly $1.3 trillion, around 29 per cent. Productivity gains account for about $1.1 trillion, close to 24 per cent.

Bain groups sectors into four clusters. Technology foundation, covering semiconductors, data centres, cloud and servers, has $1.5 trillion at stake on what Bain calls unavoidable demand. Rewired, at $1.5 trillion, spans healthcare, pharma, payments, enterprise software, logistics and manufacturing, where it expects leadership gaps to open between fast and slow adopters within a few years. Augmentation accounts for $1.3 trillion in sectors whose business models remain intact. Revolution accounts for $0.3 trillion.

The revolution cluster includes IT services and customer support, where Bain says AI can replace parts of the human delivery model and profit migrates to whoever owns the AI layer.

The breakdown matters more than the headline figure, and it contradicts how most companies are spending.

Of the $4.7 trillion, productivity accounts for about $1.1 trillion. Just under a quarter. The remaining three quarters comes from innovation, at roughly $2.2 trillion, and from profits moving between competitors, at about $1.3 trillion.

That is not how AI is generally being bought. Most enterprise deployment is aimed at doing existing work faster and cheaper: summarising documents, drafting code, handling support tickets. Those are productivity projects, and Bain's arithmetic says they address the smallest of the three pools.

The larger money sits in things that did not previously exist, and in taking share from firms that adapt slowly. Both require changing what a company sells rather than how efficiently it produces it, which is a far harder institutional task than deploying a copilot.

There is a further implication in the redistribution figure. Roughly $1.3 trillion simply moves between companies, which means for every firm gaining there is one losing. Industry-level studies tend to present AI as additive. Bain's accounting says a meaningful share of it is a transfer, and the transfer is from incumbents who move slowly to competitors who do not.

That is the uncomfortable half of the number, and it is the half that will show up in individual companies' results rather than in national productivity statistics.

The smallest cluster is the one Indian readers should look at first.

Revolution carries $0.3 trillion, a rounding error beside the other three. It covers IT services and customer support, and Bain's description of it is unusually blunt: the delivery model shifts wholesale to AI, and profit migrates to whoever owns the AI layer. That is the sector India built its services economy on.

The evidence is already visible in filed accounts. MathCo's revenue rose 23.7 per cent in FY26 while profit fell 94 per cent, because employee costs rose 33 per cent to deliver that growth. TCS grew revenue 4.6 per cent for the full year. eGain's AI-related revenue rose 20 per cent while total revenue rose 3. In each case the AI work is real and the economics of delivering it are worse than the economics of what it replaces.

Bain's framing explains why. If AI collapses the cost of production, a business whose product is production capacity sold by the hour faces falling prices for its core output. Owning the AI layer is where the margin goes, and the AI layer is owned in Santa Clara and San Francisco.

The $1.5 trillion technology foundation cluster is the other side of the same coin, and it is also where India has the least presence. Applied Materials committing $5 billion to India and Tata building at Dholera are attempts to buy into that cluster, with the first commercial fab still years from output.

The honest reading is that the largest profit pools sit where India is either a customer or a late entrant, and the pool India currently occupies is the one Bain expects to shrink.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.