I have written before about the rough quarters ahead for India’s large IT players like Tata Consultancy Services (TCS), Infosys and Cognizant: from outcome-based pricing and shrinking fees to a thinning fresher pipeline. And the fact that smaller players are upstaging them on artificial intelligence (AI)-based business. Persistent Systems grew over 16% in the last quarter while Coforge grew revenue by over a third, partly by acquisition.
The obvious approach is acquisitions and the obvious objective is not more delivery capacity, since that is precisely what AI is commoditizing. The more interesting targets sit adjacent to the core business: firms that verify, audit and govern AI output rather than simply generate it.
I have argued that enterprises will increasingly need ‘forensic plumbers’ capable of tracing failures through AI-enabled systems, and that judgement, the ability to remove plausible but wrong explanations, is becoming scarce even as fluent answers become abundant.
An IT major that acquires two or three firms which specialize in AI code review, model risk validation or compliance-grade agent monitoring would be buying exactly that scarce resource at a moment when very few enterprises have built it internally.
Vibe coding tools can generate output quickly; almost nothing in the toolchain verifies it before it ships. Acquiring that verification layer rather than another generator of plausible pull requests is a fundamentally different and harder-to-commoditize business.
The same logic extends to vertical shops: a firm that has encoded a regulator’s expectations into software is worth more right now than one that generates unchecked code. Coforge’s and Persistent’s own acquisitions already show inorganic capability-building works; there is no reason the strategy stops at tier-two companies.
Plus, there also exists the strategic alliance route. TCS has struck a partnership with Anthropic to build a dedicated unit deploying Claude across its business, extending into TCS iON’s training programmes and Diligenta’s UK pensions operations. Infosys partnered with Anthropic to build agentic systems into its Topaz platform, and separately with OpenAI, as has HCLTech.
The logic is straightforward. Frontier AI labs need enterprise distribution, sector-specific compliance expertise and client relationships built over decades; they have world-class models, but comparatively little experience navigating a bank’s risk committee or an insurer’s regulator. Indian majors have exactly that experience, plus balance sheets and delivery organizations that can scale globally.
It is a genuinely complementary pairing, and explains why these deals are being announced now, faster than most observers expected.
The arrangement, however, has a specific risk that IT majors should price in before celebrating. Anthropic has already launched its own AI services venture backed by Blackstone, Hellman & Friedman and Goldman Sachs, aimed at mid-sized companies; OpenAI is pursuing a similar vehicle. Both moves signal that model providers see enterprise implementation, not just model licensing, as territory worth owning directly.
A partnership that looks complementary today can look like training data for tomorrow’s inhouse competitor once the lab has learnt enough about enterprise delivery. The majors have seen this pattern before, in a different form, every time a client used an outsourcer to learn a capability well enough to bring it inhouse. There is no reason for the pattern to reverse.
While the pursuit of acquisitions or lab alliances remains valid, that trend argues for pursuing the right ones. Alliances that grant an AI lab distribution and relationship access without building anything proprietary for oneself are like a bridge loan against one’s own long-term relevance.
Acquisitions that only add headcount replicate the overheads that outcome-based pricing is now punishing. The deals worth pursuing are the ones that leave the acquirer holding something that a client or a partner lab cannot easily replicate: a proprietary verification layer, a regulatory data-set built over years, or a vertical’s compliance muscle memory that took a decade to acquire and can’t be trained into an AI model overnight.
The firms sitting on the largest war chests have considerably more strategic space than their quarterly numbers suggest. The sector’s next few years will be determined by which company spends its cash resources on the kinds of capability that AI cannot quickly absorb instead of what AI is actively making cheaper to replace.
That is a narrower path than it looks from the outside. Not every IT major will find it before its cash pile stops offering an advantage and starts looking like the sign of a firm that couldn’t figure out what to do with its own success.
The author is a technology consultant and venture capitalist.