The threat is not in the export data. India's services earnings just posted their biggest jump on record. It is in two leading indicators — and it is a balance-of-payments story, not an unemployment story.
Indian IT in an AI Era
- Net services receipts hit $216.6bn in FY2025-26, up $27.8bn in a single year. They cover 64% of India's goods trade deficit on their own, and 107% once remittances are added.
- On every measured outcome, AI has so far increased India's services dollars. India's back-office exports sit 27% above their pre-AI trend.
- The warning signs are elsewhere: net hiring has collapsed from ~450,000 (FY22) to 135,000 (FY26) against a growing revenue base, and chief executives now name "AI deflation" worth 3–5% of revenue.
- IT is ~1% of India's workforce but ~8% of GDP. That asymmetry is why an AI shock here would show up in the rupee long before it showed up in the jobs data.
1Start with what services actually do for India
India runs a very large deficit in goods and a very large surplus in services. That is not a curiosity; it is the load-bearing structure of the external account. In FY2025-26 the merchandise trade deficit was $337.3 billion. The current account deficit was $25.2 billion — 0.6% of GDP. The distance between those two numbers is what services and remittances do.
| Balance of payments, US$ billion | FY2024-25 | FY2025-26 |
|---|---|---|
| Merchandise trade deficit | −286.9 | −337.3 |
| Net services receipts | +188.8 | +216.6 |
| Net secondary income (remittances) | +123.5 | +143.6 |
| Primary income (investment income paid out) | −48.3 | −48.2 |
| Current account balance | −22.9 | −25.2 |
Two ratios follow, and they are the reason this article exists. Net services alone offset 64% of the goods deficit. Net services plus remittances offset 107% — they more than cover the goods gap. What actually produces a current account deficit at all is the $48.2 billion that leaves as investment income to foreign holders of Indian assets.
So when people ask whether AI threatens Indian IT, the question that matters for the country is narrower and sharper: does it threaten the dollar receipts that make a $337 billion goods deficit survivable at 0.6% of GDP?
2The measured data says AI is helping, not hurting
This needs stating plainly before any argument about risk, because it is the strongest evidence available and it points the other way.
India's gross services exports reached $421.3 billion in FY2025-26, up from $387.5 billion. Telecom, computer and information services were $206.6 billion — 49.0% of the total; business services another $124.2 billion, or 29.5%. Net services receipts rose $27.8 billion in one year, the largest increase on record.
The Global Capability Centre build-out points the same way. India now hosts 2,117 GCCs across 3,728 units, employing about 2.36 million people and generating an estimated $98.4 billion in FY2025-26 — a 32% increase in centre count since FY21, with 506 of the Forbes Global 2000 operating one. Nearly half of all GCCs established since FY21 were designed AI-first from inception.
And in the segment supposedly most exposed, Naukri's JobSpeak index recorded BPO/ITES hiring up 21% year on year in January 2026.
3Where the warning actually shows up
Two indicators have turned, and neither is visible in the export total.
First, the headcount elasticity of revenue has collapsed. The industry added roughly 450,000 people in FY2021-22. In FY2025-26 it added 135,000 — while industry revenue was expected to cross $315 billion and direct employment to reach about 6 million, up 2.3%. Revenue keeps compounding; the hiring that used to accompany it does not.
| NASSCOM basis | FY2024 | FY2025 | FY2026E |
|---|---|---|---|
| Industry revenue (US$ bn) | 268.8 | 282.6 | >315 |
| Total headcount (million) | 5.67 | 5.80 | ~6.0 |
| Net headcount additions | 60,000 (old basis) | 126,000 | 135,000 |
| Memo: net additions, FY2021-22 | ~450,000 | ||
Second, and more consequentially, price is being given away. This is the channel by which AI can reduce dollar earnings even while volumes rise, and it is now on the record from the people running the companies.
NASSCOM's own Strategic Review 2026 says providers “re-engineered revenue models, moving away from FTE delivery to outcome-based, risk-sharing constructs as AI-driven productivity materialises,” and that hiring will “shift from volume to skill mix, reflecting greater AI-driven productivity gains passed through to clients.” That last clause is the industry body conceding both halves of the problem: the productivity is real, and it is being handed to the customer as price rather than retained as margin.
HCLTech's chief executive C. Vijayakumar put a number on it in April 2026, saying “AI deflation” would take 3–5% off revenue in the coming year — while his company grew 6.0% in dollar terms (3.9% in constant currency) and added staff. TCS's K. Krithivasan acknowledged the same pressure, calling it “degrowth.” Infosys's finance chief Jayesh Sanghrajka told analysts in July 2026 that “client expectation on productivity, along with high competitive intensity is resulting in softer increase in price versus our expectations.”
| Company | FY26 revenue, US$ | YoY reported, % | YoY const. currency, % | Headcount, 31 Mar 26 | Net change |
|---|---|---|---|---|---|
| TCS | 30,017M | −0.5 | −2.4 | 584,519 | −23,460 |
| Infosys | 20,158M | — | +3.1 | 328,594 | +5,016 |
| HCLTech | ~14,700M | +6.0 | +3.9 | 227,181 | +3,761 |
| Wipro (IT services) | 10,478M | −0.3 | −1.6 | 242,156 | +8,810 |
| Tech Mahindra | 6,385M | +1.9 | +0.6 | 147,623 | −1,108 |
| Cohort | — | — | — | — | −6,981 |
Read the cohort the same way. Three of the five — Infosys, HCLTech and Wipro — added people in FY2026, together +17,587. The cohort’s net −6,981 is TCS’s −23,460 offset by everyone else’s hiring, with Tech Mahindra shedding a further 1,108 — and even that is an internal mix shift (IT −5,232, business process services +4,694) rather than a clean reduction. “The industry is shrinking its workforce” is not what the filings say. “TCS is” comes closer.
4The layoff that officially was not about AI
In late July 2025 TCS announced it would cut about 12,000 roles — some 2% of its workforce — over FY2026, concentrated in middle and senior grades, with severance, extended health cover and outplacement. It was the largest such announcement in the sector's history, and the market read it as the first big AI displacement in Indian services.
The company said otherwise, and the denial is unusually specific. Chief executive K. Krithivasan:
Take him at his word and the story is still awkward. TCS ended the year down 23,460 net — roughly double the announced cut, because gross hiring did not offset separations — while simultaneously hiring around 40,000 freshers. Indian IT unions alleged the layoffs breached the Industrial Disputes Act. A company can be telling the truth that it does not need fewer people in aggregate and still be demonstrating that it needs different people, faster than it can retrain the ones it has. For an economy whose comparative advantage was supplying large numbers of trainable graduates, that distinction is not reassuring.
5What is genuinely not known
The honest position is that FY2026 data cannot distinguish between two very different futures, and anyone claiming otherwise is guessing.
Volumes up, prices down, headcount flat is exactly what an industry looks like when it is simply becoming more productive — delivering more work with the same people and passing some of the gain to customers. It is also exactly what an industry looks like just before dollar revenue rolls over, when price concessions finally outrun volume growth. The same data supports both readings.
There is no credible measured estimate of AI-driven job or revenue displacement in Indian BPO. The figures being quoted are consultancy estimates, corporate statements, or inferences from hiring slowdowns — not measurement. An EY analysis putting entry-level IT role losses at 20–25% is an estimate, not a survey.
ICRIER's task-based study of AI and India's labour market, which uses KLEMS, PLFS and occupational data, concludes that displacement risk in employment-intensive sectors is “limited for now” and that in knowledge services AI “may augment capabilities.” It contains no quantitative estimate of IT or BPO export displacement. I could find no RBI Bulletin article on AI and India's labour market at all.
Officialdom has moved too, and the direction is worth noting. The Economic Survey 2024-25 devoted a chapter to “Labour in the AI Era” and warned that AI “can result in large scale labour displacement, especially at the middle- and lower-quartiles of the wage distribution,” adding that “a labour-surplus country like India cannot afford protracted labour displacement.” One year later, the 2025-26 Survey struck a markedly calmer tone — fears of mass displacement “have not materialised so far,” AI “is more likely to complement human labour” — while proposing an AI Economic Council to monitor the risk. Two Surveys, twelve months apart, on the same question.
6Why this is a currency story, not a jobs story
Here is the asymmetry that most coverage misses.
| Exposure | Value, % | What it means |
|---|---|---|
| Direct tech-sector employment | ~6.0mn | Against ~562mn employed nationally — about 1.1% of the workforce |
| Services exports as share of GDP | ~10 | $421.3bn against roughly $4.3trn |
| Software + business services share of services exports | 78.5 | The AI-exposed slice is about 7.7% of GDP |
| Share of the goods deficit covered by net services | 64 | 107% including remittances |
A sector that is 1% of jobs but ~8% of GDP and covers two-thirds of the goods trade deficit cannot produce a mass-unemployment event. It can produce a balance-of-payments event. If services receipts stall while the goods deficit keeps widening — and it widened by $50 billion in FY2026 alone — the adjustment happens in the exchange rate and the capital account, not in the unemployment rate.
That is the opposite of how the AI-and-India debate is usually framed, and it is what the Economic Survey's own argument implies: precisely because services exports impose no hard employment constraint, an AI shock to them would show up as a currency problem long before it showed up as a jobs problem.
Implications
- Watch price realisation and revenue per employee, not headcount. Headcount is a lagging, politically noisy indicator that companies manage. Price is where AI deflation lands first, and three chief executives have now said so on the record.
- Watch whether GCC growth keeps offsetting vendor pricing pressure. Captive centres and listed vendors are now roughly half the export base each. The balance of payments only cares about the sum — which means the vendors can be in trouble while the country is not, for a while.
- Treat import substitution as balance-of-payments insurance, not industrial nostalgia. If the services surplus is the thing covering a $337bn goods deficit, then every dollar of that deficit which can be removed by making things domestically is a dollar of reduced dependence on a receipt stream whose pricing power is visibly eroding. That is the real link between this story and the chemical, fertiliser and textile substitution analyses on this blog.
- Do not claim the displacement has happened. It has not, in any measured series. The case for concern rests on leading indicators and executive testimony, which is enough to act on and not enough to assert.
Sources & method. Balance of payments: RBI, Developments in India's Balance of Payments, Q4 2025-26; the published lines reconcile to the reported current account in both FY25 and FY26. Services export composition: RBI, cited in a Rajya Sabha written reply, and Economic Survey 2025-26. Industry aggregates: NASSCOM Strategic Review 2024, 2025 and 2026 — note NASSCOM's own restatement of FY2024 after its GCC landscape study, and that current-year figures are estimates. Company data: Infosys Q4 FY26 fact sheet filed with the SEC (Form 6-K); TCS (BSE-filed) and Wipro (SEC Form 6-K) figures are from company filings; only HCLTech's headcount figure is press-reported (its own release states only "227,000+"). Executive quotations: NASSCOM Strategic Review 2026 (verbatim); Vijayakumar and Krithivasan April 2026, Sanghrajka and Parekh Infosys Q1 FY27 call, 23 July 2026 (press transcripts). Counter-evidence: Richard Baldwin, IMD, analysis of WTO commercial services statistics, 30 July 2026; Zinnov–NASSCOM India GCC Landscape 2026; Naukri JobSpeak, January 2026. Policy: Economic Survey 2024-25 (“Labour in the AI Era”) and 2025-26. Structural analysis: ICRIER, Interactions of Artificial Intelligence with India's Labour Market. Figures described as estimates or press-reported are labelled as such in the notes; where a primary source could not be opened, that is stated rather than substituted.
About this article: Researched, written and edited by Umashankar Triplicane Dwarakanathan, with AI research assistance; every figure is meant to trace to the primary source cited. See the Editorial Policy for how sourcing, AI use and corrections work.
