The so-what

India's IT-services majors are living through the most consequential repricing event in the sector's history, and investors still appear to be modeling it as a cyclical slowdown rather than a structural shift. TCS booked $2.3 billion in annualized AI revenue by Q4 of its just-closed fiscal year, HCLTech's Advanced AI line hit $620 million, and Infosys reported $275 million in AI-linked revenue in its Q3 print 1. At the same time, TCS revenue actually declined 0.5% and Wipro's slipped 0.32% in dollar terms for the fiscal year, with five major Indian IT firms cutting a combined 6,981 jobs 1 4. The headline growth numbers look weak. The underlying mix shift — from linear, headcount-billed delivery to AI-enabled, outcome-priced delivery — is the real story, and it is happening faster than most sell-side models assume.

Deflation is real, but it is being captured, not just suffered

HCLTech's own chief executive has put a number on the pricing compression: a modernization deal that would have been worth $100 million two years ago is now worth roughly $80 million, even though it requires 25–30% more delivery effort to close because clients demand AI-augmented, outcome-based contracts rather than time-and-materials staffing 1. Kotak Securities estimates 12–15% of sector revenue is exposed to direct AI-driven displacement, with a 3–5% deflationary drag baked into FY27 guidance across the tier-1 pack 1. TCS CEO K. Krithivasan has framed the company's own target explicitly: arrest the degrowth in legacy delivery while AI revenue scales fast enough to be "net accretive" 1. That is the crux for investors — whether AI revenue growth rates can outrun legacy price deflation before margins get squeezed. So far, deal volumes suggest they can: Infosys closed $14.9 billion of large-deal total contract value in its fiscal year with 55% coming from net-new business, evidence that AI capability is winning incremental wallet share even as per-unit pricing falls 1.

NASSCOM's numbers show the export engine is still running, just on a different fuel

India's National Association of Software and Service Companies projects the broader tech industry — services, products, GCCs and BPM combined — will hit $315 billion in FY26 revenue, up 6.1% year-on-year, with $246 billion of that from exports and IT services alone contributing roughly $149 billion 2 3. Crucially, NASSCOM sizes the AI-specific services opportunity at $10–12 billion for FY26, still a modest slice of the total but one management teams across the sector describe as the fastest-scaling line item they run 3. About 85% of technology service providers surveyed by NASSCOM now have agentic AI platforms live or in pilot, and the industry has trained more than two million professionals in AI skills, including 200,000–300,000 in advanced capabilities — a retraining program at a scale few other national tech workforces are attempting simultaneously 3. NASSCOM's more provocative claim is forward-looking: agentic AI could open $300–400 billion in additional addressable spend for technology services globally by 2030, spanning legacy modernization, AI operations, cybersecurity and governance — a pool Indian vendors are positioning to capture disproportionately given their existing client relationships and delivery infrastructure 3.

The Infosys-Anthropic deal is the clearest signal of where margin defense is headed

In February, Infosys and Anthropic announced a strategic collaboration to build enterprise-grade AI agents for regulated industries — telecommunications first, then financial services and manufacturing — integrating Claude models and the Claude Agent SDK with Infosys's own Topaz AI platform 5 6. The structure matters more than the announcement: rather than reselling a foundation model API, Infosys is embedding Anthropic's agent tooling into vertical-specific offerings it can price on outcomes — claims processing, compliance reporting, network operations modernization — precisely the multi-step, judgment-heavy workflows that are hardest for a hyperscaler to commoditize with a general-purpose tool 5. Infosys has assembled a similar multi-front stack with OpenAI, Google Gemini, Microsoft, AWS and NVIDIA, a deliberate hedge against being locked into a single model provider's roadmap or pricing 1. Rivals are converging on the same playbook: TCS's largest 2025 acquisition, Coastal Cloud, and a wave of tuck-in deals at HCLTech and Infosys pushed calendar-2025 Indian IT M&A value to $5.7 billion, roughly double 2024's $2.9 billion, as vendors buy vertical expertise and AI-native talent rather than build it from scratch 1.

Where the real risk sits

The more skeptical read, laid out sharply in recent Forbes coverage, is that hyperscaler-native tools like AWS Transform are attacking the highest-margin part of the legacy business directly — cloud and mainframe modernization — by monetizing the cloud consumption destination rather than the services-led migration journey, and have already moved 1.1 billion lines of customer code with dramatically compressed cost and timeline versus conventional services delivery 4. TCS Chairman N. Chandrasekaran's counter-thesis is that governance, oversight and accountability for AI systems becomes the new annuity revenue once modernization itself is commoditized 4 — but that market is unproven at anything like the scale of the maintenance work being displaced. For investors, the next two-to-three quarters of earnings (Infosys reports Q1 FY27 on July 23) should be read less for topline growth and more for two ratios: the pace of AI/agentic revenue as a share of total, and the gap between deal TCV growth and headcount growth. A widening gap is the clearest evidence yet that India's tech-export economy is being restructured around AI-augmented margin rather than volume-based labor arbitrage, with real implications for which vendors — and which GCC and outsourcing-adjacent equities — re-rate first.

What to watch

Investors tracking this theme should watch three concrete markers into year-end: whether AI-linked revenue growth rates (currently running well above group averages at all four majors) continue outpacing legacy-segment deflation, whether net headcount additions turn positive again as NASSCOM's projected 135,000 FY26 net job adds materialize, and whether hyperscaler-native modernization tools capture enough enterprise mainframe and legacy-application work to structurally cap services pricing before Indian vendors' own agentic platforms scale to defend that revenue pool 1 3 4.