New Delhi: As a Tamil brahmin, I grew up in a social milieu where for the longest time (starting the late ‘80s up until recently) the biggest aspiration of families was that their children should become software engineers and go work in the US. Photos of cousins standing surefooted in their Nikes, leaning casually against their shiny new Toyota Corollas were meant to motivate legions of younger ones to find the shortest and fastest route to the US.
Then came the offshoring boom, which turned sleepy Indian cities such as Bengaluru into bustling tech cities. Youngsters no longer needed to go to the US, but could work for large American corporations via an Indian IT services company. Even as other sectors of the economy failed to create sufficient or meaningful jobs, the IT industry turbocharged not just dreams and prosperity of Indian households, but the economy itself. From about $130 million in software/IT exports in 1990 to $283 billion (including hardware) in FY25 (as per Nasscom estimate), India’s IT industry hasn’t just created the world’s largest middle class, but also earned India global respect as a country of coders.
So, the country owes a huge debt of gratitude to the FC Kohlis, Narayana Murthys, Azim Premjis and Shiv Nadars who turned India into the world’s tech services hub and business back office.
Here’s the thing, though: the combined market cap of the top six Indian IT companies (TCS, Infosys, HCL, Wipro, LTIMindtree, and TechMahindra) was about $224 billion as of yesterday. Anthropic, maker of AI assistant Claude, said mid-Feb that its valuation had shot up to $380 billion after a $30-billion fund raise. Indian IT companies are cash machines; Anthropic doesn’t expect to break even until 2028, burning roughly $3 billion a year. Yet, just one small AI company in the US is worth more than the biggest Indian IT companies combined.
One could argue that AI valuations are perhaps in the bubble territory, but the unmistakable message in so much money chasing AI is that it is the most profound technological shift we are witnessing as humans. Just consider what happened to the IBM stock after Anthropic said a few days ago that its Claude Code could modernize COBOL codebase in a few quarters instead of years. Out go the army of consultants and coders. The IBM stock dropped 13% on the news—its biggest single-day drop since October 2000. In February alone, IBM’s market value was down 27%!
This wasn’t a one-off AI anxiety attack. Earlier in February, Claude Cowork plugins had rattled a wide swathe of software stocks, including ThomsonReuters (single-day drop of 16%), LegalZoom (20%), FactSet (10% plus), database analytics giant RELX, and the Indian IT pack as well.
While Claude Code is unlikely to ‘modernise’ COBOL on its own (software engineers will evaluate Claude’s assessment and recommendations and then implement) and Claude Cowork is not going to make some sophisticated software obsolete. For instance, even Anthropic says that a central feature of its Claude Code Security is mandatory human review. It does not automatically apply any changes, send API requests, or even validate how bad a software vulnerability is in live environments.
Therefore, while investors are reacting to the fact that AI could broadly upend current business models, they aren’t factoring in the many complexities involved in deploying and running diverse software and of varying complexity and vintage. Most simply, there are huge safety concerns around sharing proprietary code and data with third-party AI systems. Then, every industry has its own regulatory requirements, which add another layer of complexity. In other words, the TCSes and Infys will continue to be relevant.
That said, there is no doubt that India’s IT giants cannot continue working as they have so far. Their business models will need to hange to the emerging reality of AI. (As an aside, I don’t buy into Citrini Research’s 2028 ‘scenario’ of Indian IT industry getting wiped out.) The bread-n-butter Time & Material model, where the client pays for the number of manhours at an agreed per-hour price, was already beginning to fray. In the age of AI, it becomes a perverse model, where greater efficiency means lower billable hours. If an AI agent can do in 10 hours what a human takes 100 hours, then the billing model can no longer be effort-based, but must be value- or outcome-based to allow the service provider to capture appropriate reward.
Already there are reports of clients renegotiating contracts, with an eye on extracting AI-led productivity gains from vendors. How much will be the drop in contract rates? One wouldn’t know until the IT companies report details of their new contracts. The fact is, IT companies that can quickly adapt to this new reality and move away from their linear, more-headcount-more-revenue model will be the ones that not just survive, but thrive in the new age of AI.