AI and Indian IT Services: Threat, Tailwind, or Both?
Is AI a threat to TCS, Infosys and Wipro, or their next growth engine? A data-driven guide to the billable-hours model under pressure.
Quick Facts
| The industry at stake | India's $315 billion technology sector, roughly 6 million people employed directly |
| The one-day proof point | Nifty IT fell 5.9% on February 4, 2026, its worst day since March 2020, on AI-agent fears |
| The mechanism everyone is watching | "Deal deflation": AI lets the same work get done for less, and clients want the savings |
| The number that cuts both ways | TCS shed 23,460 jobs in FY26, and grew its AI revenue run-rate to $2.6 billion in the same year |
| The competitor AI didn't create | Global Capability Centers, now 2,100-plus in India, employing 2.36 million people |
| What we are not doing here | Picking a winner. This is a framework, not a stock tip |
What You'll Learn
By the end of this article, you will understand:
- Exactly how the Indian IT services business model works, from first principles, so you can judge the AI threat with precision rather than vibes
- The full bear case, with the specific mechanism by which AI attacks this business, not just headlines
- The full bull case, with the specific mechanism by which AI could become this industry's next major growth wave, the way cloud computing did a decade ago
- Real, disclosed numbers from TCS, Infosys, HCLTech, Wipro, Tech Mahindra, LTIMindtree, Persistent Systems and Coforge, not analyst guesses
- Why Global Capability Centers are a genuine structural threat that predates AI, and how AI is changing that competition
- A practical checklist of exactly what to track in each earnings season to know which side of this debate is winning
If you want the company-specific deep dive on the largest Indian IT franchise, we have that separately in our full TCS analysis and TCS's most recent quarterly results. This article is the sector-wide version: not one company, but the whole business model.
The Day the Question Stopped Being Theoretical
On the morning of February 4, 2026, Anthropic released Claude Cowork, an AI system built with eleven plugins designed to do actual business tasks: reviewing legal contracts, building financial models, running sales forecasts, tracking regulatory compliance. Nothing about a chatbot that writes emails. This was software aimed squarely at the structured, repeatable, document-heavy work that Indian IT services companies have staffed with hundreds of thousands of people for three decades.
The market's reaction was immediate and brutal. The Nifty IT index fell 5.9% in a single session, its worst day since March 2020 (Forbes). TCS fell nearly 7%. Infosys fell 7.37%. HCLTech, Tech Mahindra and Wipro fell between roughly 4% and 5%. Roughly ₹2 lakh crore in market value evaporated across the sector in one day. HCL founder Shiv Nadar's personal net worth reportedly dropped by nearly a billion dollars that afternoon.
This was not a slow-burning worry finally showing up in a quarterly earnings call. It was a single product launch, from a company that does not employ a single person in India's IT services industry, wiping out two lakh crore rupees of market value in one trading session. Whatever you believe about AI and Indian IT, the market has already told you this is not a hypothetical risk. It is a priced-in one, and the price can move violently on a single day's news.
This article is not going to tell you whether that reaction was correct. It is going to give you the tools to judge that for yourself, because the honest answer, built from what these companies have actually disclosed rather than what headlines claim, is genuinely both threat and tailwind, operating inside the same companies at the same time.
Part 1: How This Industry Actually Makes Money
Before you can judge whether AI threatens this business, you need to understand exactly what the business is, because "IT services" is a vague label for a specific, mechanical way of earning revenue.
The Billable Hour, Explained From Zero
The traditional modelBillable Hours (Man-Day Billing)The traditional IT services pricing model: a client pays for the number of people, and hours, a vendor deploys on a project, regardless of how efficiently the work gets done. AI-assisted tools that let one engineer do the work of three threaten this model directly, since the billable unit itself shrinks.See all terms in the glossary works like this. A client, say a US bank, needs software built, maintained or upgraded. Instead of hiring engineers directly, the bank pays TCS or Infosys to deploy a team, and that team's cost is billed by the hour or by the person-month, a unit the industry calls a man-day. If a project needs ten engineers for six months, the vendor bills for roughly ten times six months of labour, plus a margin.
This is why headcount has always mattered so much to these companies' investors. Revenue growth has historically tracked headcount growth closely, because the product being sold is, fundamentally, human hours.
A few more pieces of vocabulary make the rest of this article click into place:
A large bench used to be an unavoidable cost of doing this business: you keep spare people ready because a client can ask for a new team next month. AI-assisted delivery is one of the few forces that can shrink the bench a company needs without shrinking its revenue, which is exactly why bench size has quietly become an AI story, not just an HR story.
Why This Model Is Structurally Vulnerable to AI, In Theory
Here is the part that makes 2026 different from every previous technology scare this industry has weathered. Cloud computing, mobile, and cybersecurity all created new work for IT services firms to do. Agentic AIAgentic AIAI systems built to complete multi-step tasks on their own, such as reviewing a contract, building a financial model, or handling a customer ticket end to end, rather than just answering a single question. This is the category of AI tool most directly aimed at the repetitive, structured work that IT services firms have traditionally staffed with people.See all terms in the glossary, AI systems built to complete multi-step tasks on their own rather than just answer a question, is the first technology aimed directly at doing the existing work itself, not creating new work around it. A tool that can review a contract, flag the risky clauses, and draft a summary is not asking TCS for help. It is replacing the three junior associates who used to do exactly that.
That is the whole bear case in one sentence: if the billable unit is human hours, and AI reduces the number of human hours a given piece of work requires, the model's own foundation shrinks.Whether that theoretical vulnerability is actually playing out, and how fast, is what the rest of this article investigates with real numbers.
Part 2: The Bear Case, Built From What Companies Have Actually Said
The Mechanism Has a Name: Deal Deflation
The clearest, most concrete evidence of AI eating into this business comes not from an analyst's model but from a sitting CEO describing it in his own words. HCLTech's CEO, C Vijayakumar, told analysts during the company's Q4 FY26 earnings call, held in April 2026, that AI-driven productivity is directly shrinking deal sizes: "something that was a $100 million deal could now be an $80 million deal because of the deflation" (Forbes India; NewsBytes).
Deal deflationDeal DeflationWhen a contract that would once have been priced at a certain value shrinks because AI-assisted delivery lets the vendor do the same work with less effort, and the client demands a lower price to match. A concrete example an IT services CEO has cited publicly: a deal that would have been $100 million a few years ago might be priced at $80 million today.See all terms in the glossary, HCLTech's own term for this, is currently estimated by the company at a 2% to 3% annual drag on revenue, with contracts now requiring 25% to 30% more effort from the vendor to execute at the lower agreed price, according to the company's own disclosures. This is not a rumor or an analyst's worst-case scenario. It is guidance from the company itself, built into how it is telling investors to model FY27.
Sit with what this means structurally. AI is not simply making these companies less competitive against outside disruptors. It is making them compete against themselves: the same AI tools that win them new deals are also the reason clients now expect to pay less for the deals they win. Productivity and pricing power are pulling in opposite directions inside the same contract.
Everest Group, an IT services research firm, frames the same shift more starkly. Partner Yugal Joshi has described agentic AI as directly automating the man-day billing model itself, calling it a structural threat to the roughly $300 billion global outsourcing industry, with engagement timelines shrinking and billing shrinking alongside them. Everest's own research found that only 19% of enterprises have actually redesigned their procurement, governance and contracting structures to support AI-led delivery, meaning 81% are still trying to buy AI-augmented work through contract templates built for the old labour-counting model. That mismatch is exactly where deflation pressure like HCLTech's comes from: old contracts, new productivity, and someone has to eat the difference.
The Headcount Numbers Are No Longer Ambiguous
For years, "AI will hurt IT services headcount" was a prediction. In FY26, it became a disclosed fact, though a more nuanced one than the headlines suggested.
Full-year change, FY26 (year ended March 2026), where disclosed. TCS and Infosys resumed net hiring in the following quarter.
TCS's headcount fell from 607,979 at the end of FY25 to 584,519 at the end of FY26, a net reduction of 23,460, the largest such cut in the company's history (Zee Business). The company had earlier flagged plans to reduce roughly 2% of its workforce, over 12,000 people, concentrated at mid and senior levels, and its own HR leadership has said the total decline cannot be attributed to that programme alone.
Across the sector, the pattern was similar even where the numbers were smaller. Infosys's headcount slipped by 532 sequentially in the June 2026 quarter to 328,062. Tech Mahindra cut 863 employees in the same quarter. Wipro's headcount technically rose, but the addition "disappears once you strip out an acquisition," meaning organic headcount was flat to down (PrimeInvestor sector review). Sector-wide, gross hiring across Indian IT firms fell from a five-year average of roughly 230,000 people a year to about 170,000 in the year ending March 2026.
None of this is happening because these companies are shrinking. Revenue kept growing through FY26 at every major firm. That is precisely what makes this a genuine AI signal rather than a downturn story: companies are doing more revenue with fewer people, which is the textbook definition of a productivity shock hitting a labour-priced business.
The Competitor That Predates AI, and That AI Is Now Arming
Global Capability CentersGCC (Global Capability Center)An in-house technology and operations unit that a multinational company sets up directly in India, staffed with its own employees, instead of outsourcing that work to a vendor like TCS or Infosys. A GCC competes with IT services firms for the same talent and increasingly for the same higher-value work.See all terms in the glossary are not an AI phenomenon. Multinational companies have been setting up in-house technology and operations units in India for over two decades, initially to cut costs by doing internally what they used to outsource. What has changed is the scale and the ambition of what these centers now do.
Source: NASSCOM and Zinnov's India GCC Landscape research, widely reported across Indian business media in 2026.
The critical shift NASSCOM's own 2026 commentary highlights is not GCC headcount growth, it is a move "from delivery engine to enterprise nerve center": GCCs are increasingly taking on product ownership, AI-led transformation, and business decision-making, not just the routine, lower-value cost-arbitrage work they started with. That is precisely the higher-margin, higher-skill work that IT services vendors have spent the last decade trying to move up into themselves. NASSCOM projects the GCC count could exceed 3,500 by 2030, generating over $110 billion annually.
Every rupee of work a multinational chooses to do inside its own India GCC is a rupee it never sends out to bid for at TCS, Infosys or Wipro. AI does not create this competitor. It makes the competitor more capable, because a GCC with fewer people can now attempt more ambitious, AI-augmented work in-house than it could have five years ago.
Part 3: The Bull Case, Built From the Same Companies' Own Numbers
AI Revenue Is Real, Disclosed, and Growing Fast
Every argument above is true. It is also incomplete, because it describes only the threat AI poses to the old way these companies made money, and says nothing about the new way they are already making it. Unlike the vague "AI-powered" marketing language that saturates every sector right now, most large Indian IT firms now disclose a specific, audited AI revenue figure every quarter, which lets you actually check whether the bull case has substance or is just a narrative.
TCS and HCLTech report an annualised AI run-rate; Infosys and LTIMindtree report AI as a share of quarterly revenue. Figures as disclosed by each company for the periods shown.
Sources: TCS Q1 FY27 results, HCLTech Q4 FY26 and Q1 FY27 disclosures, Infosys Q1 FY27 official results.
Notice which two companies are missing from that chart. Wipro and Tech Mahindra have not disclosed a specific AI revenue figure as of their most recent results, even while both talk extensively about AI in their investor commentary. That gap is itself information: a company confident in its AI-linked revenue generally wants to show you the number. Treat undisclosed AI revenue the same way you would treat any other unverifiable claim on an earnings call, as marketing until proven otherwise.
The Argument Nobody Disputes: Someone Has to Build the Plumbing
The strongest version of the bull case is not "AI won't hurt IT services." It is a more specific claim: enterprises overwhelmingly lack the in-house skill to actually deploy AI into decades-old, messy, custom-built legacy systems, and that integration work is exactly what Indian IT services firms have spent thirty years getting good at.
A large bank does not have a single "AI system." It has hundreds of separate applications built over decades, in different programming languages, holding data in incompatible formats, connected by fragile custom integrations that only a handful of people fully understand. Deploying an AI agent that can, say, automatically process a loan application requires that agent to talk to all of that legacy infrastructure safely, without breaking compliance rules, audit trails or existing workflows. That is systems integration work, and it is precisely the muscle Indian IT services firms have built at enormous scale.
The deal evidence backs this reading. Infosys closed $3.6 billion in large deal TCV in Q1 FY27 alone, with 61% of it "net new" business rather than renewals, and $14.9 billion in large deal TCV for the full FY26 year (Infosys official results). Wipro's large deal bookings reached $1.63 billion in Q1 FY27, up 12.9%, anchored by thirteen separate multi-year AI-enabled contracts. Tech Mahindra's new deal wins hit $1,078 million, up 33% year on year, alongside EBIT growth of 53.3%.
Routine, structured, repeatable tasks: basic testing, documentation, Level 1 and Level 2 support, standard contract review, template-based coding.
The pure "staff augmentation" contract, where a client simply rents headcount by the month with no specialised deliverable.
Integrating AI safely into legacy enterprise systems that were never designed to talk to it.
Redesigning entire business processes around AI, not just bolting a chatbot onto the old one, the kind of work TCS's $800 million SKF deal represents.
Data governance, security, and compliance work needed before any regulated company can deploy agentic AI at all, an area Indian IT firms already dominate in banking and healthcare.
Routine, structured, repeatable tasks: basic testing, documentation, Level 1 and Level 2 support, standard contract review, template-based coding.
The pure "staff augmentation" contract, where a client simply rents headcount by the month with no specialised deliverable.
Integrating AI safely into legacy enterprise systems that were never designed to talk to it.
Redesigning entire business processes around AI, not just bolting a chatbot onto the old one, the kind of work TCS's $800 million SKF deal represents.
Data governance, security, and compliance work needed before any regulated company can deploy agentic AI at all, an area Indian IT firms already dominate in banking and healthcare.
The Industry Has Been Here Before, and It Did Not Die Then Either
It is worth remembering that this is not the first time a technology shift was supposed to end Indian IT services. Around 2015 and 2016, the rise of public cloud computing triggered near-identical predictions: cloud would let enterprises manage their own infrastructure directly, cutting out the outsourced infrastructure-management contracts that were a major revenue line for Indian IT vendors at the time. Some analysts at the time forecast cloud adoption could slice as much as 15% off the earnings of the top outsourcing players.
It did not play out that way. What actually happened is that Indian IT firms pivoted from managing clients' own data centers to managing clients' cloud migrations and multi-cloud environments, a service line that barely existed before and became a major growth driver through the following decade. The technology that was supposed to eliminate the business instead created a new, larger version of it. That is not a guarantee AI will follow the identical script, generative and agentic AI genuinely attacks the billing unit itself in a way cloud never did, but it is a legitimate reason not to treat "this time the disruption is real" as automatically true just because it is being said with more urgency this time.
The honest historical lesson is narrower than "IT services always survives." It is this: when a technology threatens to eliminate demand for a service, the companies who win are the ones who move fastest to sell the technology itself as a new service, rather than the ones who simply defend the old pricing model. Cloud migration was that pivot for the 2015 disruption. AI implementation and AI governance appear to be the equivalent pivot being attempted right now, and the disclosed AI revenue numbers above are the evidence of how far along each company actually is.
The Reskilling Effort Is Real and Large
More than 2 million technology professionals in India have now been trained in AI skills, with 200,000 to 300,000 specifically in advanced AI capabilities, according to NASSCOM's 2026 Strategic Review, part of an industry-wide response that has been underway since well before the February 2026 stock selloff made the threat front-page news.
Whether reskilling at this scale is enough to offset the productivity gains AI delivers to clients is exactly the open question this whole debate turns on. But it is worth being precise about what the bear case gets wrong when it treats these companies as passive: the scale of retraining, the specificity of disclosed AI revenue lines, and the double-digit growth in AI-linked deal bookings are not the behaviour of an industry in denial.
Part 4: The Sector Is Not One Story, It's Two
Averaging TCS and a mid-cap specialist like Persistent Systems into one "Indian IT services" verdict hides the most useful signal in this entire debate: size and specialisation are producing genuinely different outcomes.
Q1 FY27 (quarter ended June 2026) reported figures, US dollar terms unless noted. Coforge and Wipro figures include the effect of recent acquisitions.
The largest firms face a structural growth-math problem that has nothing to do with AI and everything to do with size: to grow revenue by even a few percentage points, TCS needs to find billions of dollars in new business every year, which mathematically requires either enormous new deals or a huge volume of smaller ones. A mid-cap firm doing $450 million a quarter needs a fraction of that to post double-digit growth.
1 Vertical depth creates switching costs AI cannot easily replicate. Persistent Systems' Q1 FY27 results included a record $1.15 billion in TCV, anchored by a single $650 million, six-and-a-half-year strategic agreement built around large-scale AI-enabled engineering work, its 25th consecutive quarter of revenue growth. Mphasis, concentrated almost entirely in banking and financial services, reported that 42% of new TCV wins in a recent quarter contained AI components, with its BFSI pipeline growing 45% year on year specifically because banks face intense regulatory pressure to adopt AI for fraud detection and risk modelling, a domain where deep, narrow expertise beats a generic AI tool.
2 This is the same switching-cost moat covered in our guide to economic moats, applied to software. A bank that has embedded a specialist vendor's fraud-detection AI into its core workflow does not switch because a cheaper alternative exists. It stays because migrating that workflow, retraining staff, and re-certifying compliance is enormously expensive and risky, exactly the dynamic we described in our broader picks-and-shovels map of AI infrastructure, where we first flagged Persistent and Mphasis as companies whose AI revenue passes the "specific numbers, not just AI mentions" test.
Read growth-rate comparisons like the chart above carefully. Coforge's headline 33% growth is real, but a large share came from consolidating the Encora acquisition, which alone contributed over $100 million of the quarter's $592 million in revenue. Strip out acquisitions and the organic growth story is more modest, though still healthy. Always separate acquisition-driven growth from organic growth before treating a number as a verdict on AI, on any company, in any sector.
Part 5: What to Actually Track Each Quarter
If you can track only one number, track the gap between revenue growth and headcount growth. A widening gap, revenue rising while headcount falls or stays flat, is the single cleanest evidence that AI productivity is showing up in the numbers rather than remaining a slide in an investor presentation. It is exactly the same discipline we recommend in how to read annual reports: find the number that cannot be spun, and watch it every quarter.
Once you have a view on which companies are actually converting the AI story into durable growth, the next question is whether the market has already priced that in. Our guide to valuation and when to buy covers exactly that discipline, which matters enormously here given how violently this sector can reprice on a single day's news, as February 4, 2026 showed.
Key Takeaways
-
The billable-hour model is genuinely, structurally exposed to AI, and this is not speculation anymore. HCLTech's own CEO has quantified deal deflation at 2% to 3% annually, with specific contracts shrinking from roughly $100 million to $80 million because AI lets the same work get done with less effort.
-
The market has already shown you how violently this can reprice. The Nifty IT index fell 5.9% in a single day on February 4, 2026, after a single AI product launch, wiping out roughly ₹2 lakh crore in sector market value.
-
Headcount data now confirms the productivity story is real, not theoretical. TCS cut 23,460 jobs in FY26 while growing revenue; sector-wide gross hiring fell from a five-year average of roughly 230,000 a year to about 170,000.
-
AI revenue is equally real and disclosed, not marketing. TCS's AI run-rate reached $2.6 billion, HCLTech's Advanced AI revenue grew 62.1% year on year in a single quarter, and large AI-linked deal bookings are growing across nearly every major firm.
-
Global Capability Centers are a structural threat that predates AI and that AI is now amplifying. India hosts over 2,100 GCCs employing 2.36 million people, increasingly doing higher-value, AI-led work that used to be outsourced.
-
Size changes the calculus. Large firms face a harder growth-math problem independent of AI; mid-cap, vertically specialised firms like Persistent Systems and Mphasis are showing faster growth and AI-linked deal traction, built on switching-cost moats that resemble what we described in our economic moats guide.
-
This industry has faced an identical "the technology will kill us" narrative before, around cloud computing in 2015, and pivoted into selling the disruptive technology as a new service line instead of dying from it. Whether AI follows the same script is the single most important open question for anyone holding these stocks, and the evidence in this article is what will answer it, one quarter at a time.
Sources
- Forbes, "Indian IT Stocks Nosedive As Fear Of Anthropic's New AI Tool Goes Global," February 4, 2026. Stock price and Nifty IT index moves on the day of the Claude Cowork launch.
- Rest of World, "Why Claude Cowork is a math problem Indian IT can't solve," 2026. Context on the man-day billing model and industry expert commentary.
- Forbes India, "HCLTech profit rises 4.2%; AI revenues at $620 million," 2026. The CEO's deal deflation quote and Advanced AI revenue figures.
- NewsBytes, "HCLTech says AI causes 2% to 3% annual deal deflation." Confirming the company's own deflation estimate.
- Zee Business, TCS headcount coverage, FY26 results. TCS's exact headcount figures.
- Infosys official Q1 FY27 IFRS results release. Revenue, margin, AI revenue share, and deal TCV figures.
- PrimeInvestor, "Indian IT Sector Rally: What Q1 FY27 Results Actually Tell Us." Sector-wide headcount and margin comparison across all major firms.
- NASSCOM and Zinnov, India GCC Landscape reports and NASSCOM Technology Sector Strategic Review 2026, as reported across Indian business media. GCC count, employment, revenue, and sector-wide reskilling figures.
- Persistent Systems, official Q1 FY27 results release. Revenue, TCV, and EBIT figures.
- Coforge, official Q1 FY27 results announcement. Revenue growth and the Encora acquisition contribution.
A note on sourcing. Every figure in this article is attributed to a specific company disclosure, official earnings result, or a named research organisation such as NASSCOM, Zinnov or Everest Group, current as of the periods stated. AI revenue definitions are not standardised across companies, so figures are not always directly comparable, TCS and HCLTech disclose an annualised run-rate while Infosys and LTIMindtree disclose a share of quarterly revenue, and this article flags that distinction wherever it applies. Sector-wide estimates like GCC counts and NASSCOM's national figures are best available estimates from industry bodies, not audited numbers, and should be treated accordingly.
Disclaimer
Nothing on this site is investment advice. All content is for educational and informational purposes only. Do your own research and consult a registered financial adviser before making any investment decisions.
Finished reading? Mark this article to track your learning progress.
Software Engineer, Self-Taught Investor
Software engineer who started learning about money in 2016 after a layoff coincided with a new home loan. Went from bank deposits to mutual funds to picking stocks in India and the US, learning through YouTube, screener.in, TradingView, and the hard way. Still learning. This site is her notes made public — for education and sharing only, not financial advice.
