Infosys Says AI Could Be Half Its Work In Three Years, And Warns The Backlash Is Coming Here

In this storyInfosys

Artificial intelligence could account for almost half of Infosys' work within two to three years, chief executive Salil Parekh said at the Bloomberg New Economy Forum in New Delhi on Tuesday.

AI transformation work currently contributes over 8 per cent of revenue and is growing at a double-digit rate sequentially. Parekh said that if the pace continues, it would reach 20 to 30 per cent within a few quarters before approaching half over two to three years. The work includes building AI agents, coding and modernising legacy technology for clients.

Infosys hired 20,000 graduates last year and plans similar numbers again, training them in AI skills.

Parekh also flagged risks from the rapid expansion of AI infrastructure. He said there is backlash against AI build-out in many Western markets and that it may reach India as well. Asked whether physical infrastructure was keeping pace, he said data had become critical to the AI ecosystem and that stricter sovereign data and cloud requirements were restricting some capacity build-outs. He said the effects could become more visible over the next five to ten years.

The company has guided to revenue growth of 1.5 to 3.5 per cent in constant currency for FY27, citing large opportunities in AI services alongside continued competitive intensity and an AI productivity impact.

Infosys shares have risen around 8 per cent over five sessions, with Indian IT stocks lifted by HCLTech earnings and easing visa concerns.

Half of Infosys' work in three years, from eight per cent, is an extraordinary claim and the arithmetic deserves checking before it is repeated.

Parekh was careful about how he framed it. He said the AI business is small relative to the company and growing at a double-digit rate sequentially, and that extrapolating that pace produces 20 to 30 per cent within a few quarters and roughly half within two to three years. That is an extrapolation, stated as one, not a target.

The caution matters because the same company guided to revenue growth of 1.5 to 3.5 per cent for FY27. If AI goes from 8 per cent to 50 per cent of work while total revenue grows by low single digits, most of that shift is not new business. It is existing work reclassified, or existing work displaced, with AI services substituting for the application maintenance and development that has funded Indian IT for thirty years.

That is the reading Bain's profit-pool analysis supports, placing IT services in the cluster where delivery shifts wholesale to AI and margin migrates to whoever owns the model layer.

Reclassification is not failure. Infosys would rather be the firm doing the AI work than the one whose contracts it replaces. But a company whose AI revenue triples while total revenue grows 2 per cent has not found a new market. It has changed what it calls the old one.

The number to watch is not the AI share. It is whether total revenue growth moves with it.

The backlash warning is the more useful half of what Parekh said, and it is the part least likely to be repeated.

India has had almost no organised opposition to data centres. The United States has had the opposite: Data Center Watch counted 75 projects worth roughly $130 billion blocked or delayed in a single quarter, and more than 100 local moratoriums under consideration. Amazon abandoned NDAs with local governments this month and pledged $1 billion to host communities, explicitly to clear permitting. Scotland now requires an environmental assessment for every facility above 50MW.

The conditions that produced that opposition exist here in sharper form. Capacity is concentrated in Mumbai, Chennai and Hyderabad, on a grid where more than half of 191 monitored thermal stations were recently holding coal stocks measured in days, and in states where agriculture and industry already compete for water.

What India lacks so far is the disclosure that turns a local inconvenience into organised resistance. Amazon has now committed to publishing energy and water use annually. Once that number exists in one market, declining to publish it in another becomes a position someone has to defend.

For Infosys specifically the exposure is second-order rather than direct. It does not own the data centres. But AI services sold to clients depend on those clients being able to deploy, and sovereign data rules that restrict where capacity sits are already a constraint Parekh named. A services firm pivoting half its revenue to AI is making a bet on infrastructure it does not control.

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