I have spent enough years sitting across IT managements to recognise when a sector is arguing with itself. For two decades the Indian IT story was simple. You sold labour arbitrage at scale, you added digital and cloud as a second chapter, and you hoped the client’s budget cycle stayed kind. That model is not dead – at least I don’t think so. It is just no longer the whole story. What is unfolding now is messier, and far more interesting. Clients are no longer buying seats. They are buying outcomes with the uncomfortable assumption that AI will do a large part of the work. That shift is rearranging the pecking order inside the industry, and it is doing so faster than the old large-cap versus mid-cap shorthand can capture.
Look at the last three years without the noise. Growth at the very top of the listed universe has been modest, sometimes barely visible in constant currency. Mid-tier firms have grown at a different speed. That is not an accident of a single quarter. It is what happens when the buying pattern changes. Large transformation programmes still go to firms that can sit at the CIO and CTO table. But a growing slice of AI-led engineering, data modernisation and industry-specific work is being awarded to companies that can move without a 400,000-person org chart. The market is quietly pricing that divergence.
Among the large caps, the more honest conversations I have heard are not about defending last year’s revenue line. They are about whether the firm can still be relevant when productivity is no longer a slide in an investor deck but the client’s opening demand. One of the more visible rebuilds has been at Tech Mahindra. After a long stretch of being treated as the lagging large cap, the firm has spent the last three years trying to look like a different company: large-deal muscle, a tighter industry mix, AI in delivery rather than in the brochure. Last-twelve-month large deal wins have crossed the $4 billion mark. Client satisfaction and NPS have moved off their lows. About 84% of the client-facing workforce is now enabled on AI, and the firm claims productivity gains north of 40% through AI-led delivery. The market cap gap with the rest of the peer set has narrowed sharply. As per maths borrowed from a note (and verified, of course), the average premium that peers commanded over TechM has compressed from around 5.5 times in early 2023 to closer to 2.8 times by August 2026. That is not a victory lap. It is a reminder that in this sector, reputation is now a lagging indicator. You can lose a decade of relative standing, and you can also claw some of it back if the operating system actually changes.
ALSO READ:Â AI Disruption Is Exponential, Businesses Need A Margin Of Safety: HDFC AMC’s Navneet Munot
The FY27 aspiration the company has put out is the one every largecap now recites in some form – grow faster than the peer average, get margins back to an industry-standard band, make cash and predictability boring again. Whether that lands will depend less on another slogan and more on whether large deals already won convert into revenue without the usual leakage. I have seen enough ‘transformation programmes’ in this industry to know the slide is easy. The quarter after the mega-deal is hard.
The more aggressive rewrite, though, is happening one rung below.
Coforge has spent the last few years behaving like a firm that does not intend to remain mid-tier for the sake of the classification. FY26 revenue grew close to 29% in dollar terms. EBIT margin expanded 370 basis points to 14.4%. Twenty-one large deals were signed through the year. The next-twelve-months executable order book has been running well above $1.7 billion and, more recently, above $2 billion as the Encora integration has started to show in the numbers. The public stance of the CEOs in the one rung below companies, the likes of Nitin Rakesh, Sandeep Kalra or Sudhir Singh, has been consistent and blunt. Generative AI is a commodity. Applied AI, sitting inside a bank’s workflow or an insurer’s claims stack, is the differentiator. Coforge’s $2.3 billion Encora acquisition was the capital-markets expression of that view – buy an AI-native engineering firm with Silicon Valley DNA, fold it into domain-heavy verticals, and try to build a $2 billion AI-led engineering, data and cloud engine inside a larger house that still wants to be $5 billion well before FY30.
Persistent has chosen a still more binary path. The Nagarro offer – roughly 1.27 billion euros, a steep premium to the undisturbed price – is the largest bet the firm has ever placed. Combined, the two businesses are a $2.9 billion AI-led engineering platform with more than 46,000 people across 40-plus countries. Europe, which was high-single-digits of Persistent’s revenue, is meant to move towards the low twenties. ERP and customer-experience muscle that Persistent did not have at scale comes in with Nagarro. A separate $650 million-plus, six-and-a-half-year net-new deal in the US was announced almost in the same breath. The stated destination is $5 billion by FY31.
I do not know yet whether that deal will be remembered as the moment Persistent bought the geography it could not build fast enough, or as the moment a good compounder took on more integration than the stock could digest. The first-day market reaction was unkind, which is usually the market’s way of saying the strategy may be right and the timing of cash and culture is still unproven. That is a fair scepticism. Buying a European engineering franchise at a 140% premium only works if the combined entity can sell larger, multi-region AI programmes that neither firm could win alone. Scale for its own sake is the old IT disease. Persistent’s own history of tuck-ins suggests they know that. Nagarro is not a tuck-in.
These risks are a different risk from the one largecaps carry. Largecaps worry about deflation of the old book. The likes of Persistent and Coforge are underwriting integration risk and the hope that proactive, non-RFP deals can keep pricing from collapsing. Large five-year AI-led transformation wins in Europe and other parts were the kind of ticket mid-tiers were not supposed to win five years ago. They are winning them now. In a sense, the mid-tiers are trying to become more like the largecaps in geographic reach and deal size, without inheriting the bureaucracy. And within largecaps, there are examples like Tech Mahindra which are looking wired differently from their peers currently.
Amongst others, Hexaware is building the GCC into an agentic hub, Mphasis is already sitting on the large-deal table the giants used to keep, with FY26 TCV near $2.1 billion, average deal size moving toward $75 million, and close to two-thirds of recent net-new work tagged AI-led. Above them, HCLTech is turning bookings and a $1.14 billion European workplace mandate into a case that infrastructure and the digital estate still matter, while LTIMindtree is buying European consulting scale cheaply through Randstad’s tech book. And then companies like Intellect are talking of AI-led products that eventually have the ability to take margins higher in due course.
While we differentiate between the caps’ ability to win globally, there is an India overlay that I cannot ignore. This country is no longer just the back office of global IT. It is becoming a place where AI infrastructure, engineering talent and enterprise process knowledge sit in the same time zone. TCS’s HyperVault campus in Hyderabad is the large-infrastructure expression of that. The mid-tier acquisitions of American and European engineering firms are the talent-and-client expression of the same idea. One is building the power and cooling. The other is buying the right to sit in the room where the model is applied to a claims engine or a factory line.
An investor would be flirting with risk if the thought is that the old largecap names disappear. Scale still wins certain conversations. I also do not think the mid-tier outperformance of the last two years was a fluke. When a buying cycle fragments, the firms that can specialise and still show up with a serious AI delivery stack take share. Coforge and Persistent have spent real money to stay in that conversation. A few largecaps, Tech Mahindra among them, have spent the last three years trying not to be left out of it.
The sector that emerges from this will not look like the Nifty IT of 2018. It will look more like two races run on the same track. One is scale plus AI, fought among the names everyone already owns. The other is engineering plus AI, fought among names that used to be called midcaps and are now trying to price themselves as the new core. Both can compound from here. The ones that confuse headcount with relevance will not.
ALSO READ:Â Indian IT’s AI Story Is Changing, But BofA Flags A New Earnings Risk
About the Author
