What Will Happen to Indian IT?

The situation of the world is as such: Some countries are rich and other countries are poor. Being poor is quite unpleasant, and creates all sorts of problems; in general, it’s much more pleasant to be rich. (This is true of individuals just as it is of countries.) So countries that are poor will generally want to find ways in which they can become rich. The main way to do this, in this world of ours where some countries are much richer than others, is to sell things to the countries that are richer than you are.

So what do you sell?

The most straightforward thing to sell is simply what your land naturally produces — crops, or oil, or copper, or diamonds. Sometimes this works very well. There are a few cases where natural resources have made countries exceptionally rich. But usually it doesn’t work. Sometimes this is because most natural commodities (like, say, cocoa) just aren’t very scarce. And when natural resources are scarce, they bring with them all sorts of problems: overvalued currencies, corruption, weak institutions. There’s a reason the “resource curse” is such a famous concept. And even if you wanted to sell those resources, you’d still need to have them in the first place. A lot of countries don’t.

So selling natural resources usually isn’t a great bet. What about selling goods that you manufacture? There are quite a few countries that got rich doing that. You start out with a lot of cheap labor; you build a labor-intensive manufacturing sector, producing things like garments or circuit boards; and you gradually move up the value chain into more difficult and valuable things. This takes a long time, and requires a very difficult alignment of institutions and culture and circumstances, but when it works the results can be pretty extraordinary.

But it’s also very hard. Most poor countries have tried to become manufacturing powers over the last few decades; very few of them have succeeded. And there’s good evidence to suggest that it’s become much more difficult over the last few decades. As technology improves, labor-intensive manufacturing just isn’t as labor-intensive as it used to be and manufacturing employment in most developing countries is lower than it was a few decades ago.

So resource-led growth is usually a bad strategy; manufacturing-led growth is very difficult. What do you do then?

In the last few decades, a new path has become possible: services-led growth. The resource strategy sells raw goods; the manufacturing strategy combines those raw goods with human labor and sells the finished result; the services strategy simply sells the labor.

This wasn’t really possible, of course, before the internet. People worked where they lived and tended to produce things where they worked: It wasn’t feasible for someone living in Argentina to have much of a relationship with a firm in California. People might migrate for work and send money back home — in the early twentieth century, there were certain Irish towns where the majority of the income came from those who had migrated to the United States — but “remote work” wasn’t really a thing, and selling labor wasn’t really a strategy for economic development.

The internet changed that. All sorts of labor that once had to be performed in one place could now be performed anywhere. Not all jobs, of course: mechanics, appraisers, and police officers would still have to be based at home. But if you ran a software company, you could employ engineers in any part of the world with a sufficiently fast internet connection, and pay those workers a fraction of what their American or European counterparts might make. The same with customer service jobs and back-office functions in general. And so from 1990 onwards, there was a veritable explosion of “outsourcing” and “offshoring.”

There were a lot of people who complained about this, of course: Outsourcing wasn’t a great deal for American or European workers. But it was a great deal for a lot of people in poorer countries: Wages that were a fraction of American pay were still, by local standards, very good money. And the greatest beneficiary of all of this — the country that bet the most heavily of all on services exports — was India.

India is a massive country; and for a long time it was very poor. Unlike other once-very-poor countries in Asia, like China or South Korea, India never managed to develop a competitive manufacturing sector: Its share of global manufacturing exports today is about the same as Vietnam’s, despite having 14 times Vietnam’s population. But over the last few decades, India has managed to build a competitive service sector. As the largest exporter of IT and BPO services in the world, India is “the world’s back office.”

India’s services export industry employs only a small share of the population, perhaps five million people in a country of 1.5 billion. But IT is still crucial to the Indian economy. It accounts for roughly seven percent of GDP and about a quarter of India’s total exports; the consumption of its high-earning workers sustains a much larger ecosystem of domestic employment; and the services trade surplus it generates, about $190 billion in fiscal year 2025, substantially offsets India’s persistent goods trade deficit, which would otherwise pose severe balance-of-payments problems. Service exports are the primary mechanism by which tens of millions of Indians have entered the global middle class.

The question now, though, is how long the Indian services export model has left to survive.

For decades, Indian IT and BPO prospered because of a very particular arbitrage: Western firms could hire Indian workers, so that Western firms would save money and Indian workers would earn high local wages. But now there’s something that can underprice even the workers of Hyderabad and Chennai. AI models are very good and, for the quality of work they’re able to do, very cheap: They can’t do physical work (yet!), but neither can Indian employees who do their work remotely. Western firms hired Indian workers because they provided lots of value for the cost: AI, it seems, could provide even more value, for even less cost.

So it wasn’t too long after the AI boom of the last few years started that people started to suggest the first victims of displacement would be the IT and BPO workers of India. Daniel Gross, an influential tech investor, suggested in January 2024 that “$250b of India’s GDP exports are essentially GPT-4 tokens … what happens now?”

Gross was being provocative; we now look back with some condescension on the now-obvious limitations of GPT-4 and models like it. But his underlying logic was serious. The Indian IT and BPO sector had been built on a simple proposition: Indian labor was cheap, English-speaking, and technically competent, and Western firms were willing to pay for it. AI seems to break that model entirely. In a world where AI models can write code, answer calls, and process insurance claims faster and cheaper than humans, what happens to the services export path? We can get more specific. The models are already capable of doing those things. So: What happens now?

As with all complex things, it’s hard to trace a single point of origin for the Indian IT sector. One could go back to the Nehruvian elite’s fostering of elite Indian technical education in the 1950s, or — if you wanted to be truly exacting — to the British imposition of English-language education in the 1830s. But the best place to start is with a disastrous policy decision in 1978.

For many decades after its independence, the Indian state was deeply distrustful of foreign investment, viewing it as a legacy of colonial exploitation and a threat to self-reliance. And so in 1973, the Indian government passed the Foreign Exchange Regulation Act, which mandated that no foreign entity could own more than 40 percent of an Indian company. This was bad news for multinationals generally, and eventually resulted in Coca-Cola leaving the Indian market and not returning until 1993; but it was particularly bad news for IBM, which was then the largest technology company in the world and by far the dominant player in the Indian computing market. IBM simply refused to dilute its stake in its Indian subsidiary. Discussions dragged on for several years, until a trade unionist named George Fernandes became the country’s Minister for Industries. Fernandes had led anti-computer protests in Bombay in the 1960s; and once in power he took a hard line against IBM. In June 1978, IBM abandoned the Indian market. About 800 IBM installations — large, bulky mainframes — were left stranded.

This was, it turned out, a perfect catalyst for the emergence of a domestic IT sector.

At the time, there wasn’t much of a computing industry in India. The Tata Group, India’s largest company, had a fledgling IT business called Tata Consulting Services; and there was a tiny computing startup called Hindustan Computers Limited, which was focused on building minicomputers for the Indian market. To fill the vacuum left by IBM, a few new players emerged. The Indian government created one company, the Computer Maintenance Corporation, to manage the abandoned IBM installations; and in the 1980s two new startups — Infosys and Wipro — entered the market.

For all of these players the beginnings were not particularly auspicious. Wipro had been founded as a cooking oil concern in the 1940s, and only pivoted to technology after its founder died and his young son was called back from a Stanford engineering degree. Infosys, for its part, waited a year to acquire its first telephone and three years to acquire its first computer; simply importing the minicomputer they eventually bought required fifty trips to Delhi.

It wasn’t long before all of these companies — TCS, HCL, CMC, Infosys, and Wipro — realized that the real money in information technology was not in hardware but in software. India had an abundance of excellent engineers, many of them graduates of the elite Indian Institutes of Technology system; and the wages they expected to earn were pitifully low compare to what American engineers might earn. And — a basic precondition for everything else — they spoke English. Why have your engineers build things and sell them, when you could just sell the engineering labor directly?

The problem, in the 1970s and ‘80s, was that this was simply difficult to do. All the way back in 1974 TCS had secured its first outsourced software contract, when a Detroit hospital had paid it to convert some accounting code from one dialect of COBOL to another. The task was simple enough. But TCS didn’t have any machines that could run the code, so its engineers developed the software on a secondhand Burroughs mainframe — acquired from the Life Insurance Corporation, whose communist unions had blocked its use for fear it would eliminate jobs — and wrote a filter to make ICL code run on Burroughs hardware. The finished code had to be physically shipped to the United States, since no data-communication links existed between the two countries.

TCS and the like still managed to secure outsourced engineering contracts; by 1987 Infosys was successful enough to have opened an office in Boston. But it was only around 1990 that the real money started rolling in.

There were two things that underwrote this great change of fortune. The first was physical infrastructure: undersea fiber-optic cables, laid in the early and middle years of the 1990s, made it much cheaper and much faster to transmit data between the various places of the world, and allowed for high-speed connections between India and the West. And the second was policy. In 1991, India faced a severe economic crisis, the culmination of a structural balance-of-payments deficit; at the peak of the emergency the Indian government was forced to airlift gold to London to secure emergency loans.

The ultimate response to the crisis, undertaken under the aegis of the reformist finance minister Manmohan Singh under the prime minister P.V. Narasimha Rao, was to dismantle the “Licence Raj” that had defined the Indian state for decades. The onerous regulatory regime into which the Indian IT industry was born was dismantled; foreign investment was liberalized; and the Indian economy was, more or less, opened to the world. It was this combination of technological connectivity with economic liberalization, however limited, that underwrote everything that followed.

By the late 1990s the Indian IT sector had truly begun to shine. The catalyst was Y2K. Much of the world’s software had been written in the 1960s and ‘70s, most of it in the programming language COBOL, which stored years as two digits (such that “1968” was “68” and so on) in order to save on expensive memory. But this meant that when 1999 rolled into 2000, there was a risk that computers might interpret the year (“00”) not as 2000 but as 1900. Fixing this required the painstaking rewriting of millions of lines of COBOL: a grueling and labor-intensive endeavor, made more difficult by the fact that COBOL had been in decline among American developers for years.

But COBOL was still strong among Indian programmers, who had spent years dealing with clients like the Detroit hospital. And so in the late 1990s Indian programming talent enjoyed a sudden explosion of demand. By this point newer players, like Tech Mahindra and Cognizant, had joined the game; it was a gold rush for the writing and rewriting of dusty enterprise codebases. TCS alone remediated roughly 700 million lines of code. In 1999, Infosys became the first Indian company to be listed on the NASDAQ. India’s total software exports surged from $1 billion in 1997 to $6.2 billion by 2001.

It did not take too long for the humble startups of the 1970s and ‘80s to become economic giants. The 2000s and 2010s were a fantastic time for Indian IT and BPO: Broadband connectivity continued to improve; Indian IT firms were newly credible to Western purchasers thanks to the Y2K experience; and Western companies, under pressure to save money after the dot-com bust in the early 2000s and then again after the financial crisis of the late 2000s, found that offshoring work to India was an easy way to cut expenses. TCS crossed $1 billion in annual revenue in 2003, the first Indian IT company to do so; by 2012 it had surpassed $10 billion. Since its listing in 2004, TCS compounded revenue at roughly 15 percent a year. Infosys grew at a similar clip, reaching about $6 billion by fiscal year 2011.

And it was not only Indian IT firms that grew. Western companies increasingly set up their own captive operations in India — what are now called Global Capability Centers. India today hosts somewhere between 1,700 and 2,000 of these centers, the offshore arms of companies like JPMorgan, Goldman Sachs, Google, and Microsoft, employing roughly 1.9 million people.

And soon the same model was replicated for customer service and back-office work: It turned out there were countless simple functions that American and European businesses were eager to have done abroad for less money. In 1997, a General Electric executive named Pramod Bhasin opened GE Capital International Services in the city of Gurgaon, with 20 employees processing GE’s car loans; within four years the unit had expanded to 12,000 employees. In 2005 GECIS became an independent company and changed its name to Genpact; today it employs about 145,000 people.

And on top of this massive outsourcing boom, a small crop of massive companies emerged. By 2021, TCS, which in the early 2000s had absorbed CMC and consolidated its place as the dominant player in Indian IT, was valued at more than $200 billion; Infosys was valued at $100 billion; HCL and Cognizant at $50 billion; Wipro at $30 billion; and Tech Mahindra north of $20 billion. The IT companies were among the most valuable companies in India. Collectively, these six companies employed millions of people, with millions more employed at smaller players, or directly by Western firms.

In absolute terms, of course, the IT sector’s few million employees were a drop in the bucket: India has more than 250 million people working in agriculture alone, with hundreds of millions more employed in low-end informal services that have nothing to do wiht software. But the IT sector proved to be an invaluable anchor for the rest of the Indian economy. An IT worker paid $23,000 a year might have been ludicrously cheap by American standards; but he still earned several times India’s per capita GDP. His consumption — the cooks and maids he hired, the flights he bought, the money he infused into his town or his ancestral village — proved to be massively helpful for Indian growth in the 2000s and 2010s.

And indeed the Western stereotype of IT jobs as demeaning sweatshop labor was wrong. IT jobs were in extraordinarily high demand, and so the labor market for IT workers was extremely tight: In the fourth quarter of 2022 attrition at Infosys hit 27.7 percent, simply because employees had so many options that they could cycle between firms for 30 or 40 percent salary jumps. And an entire social infrastructure emerged around the dream of the Indian IT job. In the matrimonial pages of The Hindu, one of India’s most popular newspapers, the most desired type of job was “software engineer.”

The absolute peak of the IT model came in 2022. The pandemic and the resulting glut of digital spending had been a massive boon for TCS, Infosys, and their ilk; and accordingly they hired at a pace never seen before. TCS alone added nearly 100,000 employees in a single year; Infosys hired 85,000 graduates. The going was good.

On November 30, 2022, OpenAI released ChatGPT, a new chat interface for its GPT series of AI models. GitHub Copilot, which used a modified version of GPT-3, had become generally available the previous June. Initially, the capabilities of these tools were limited: The models would make simple mistakes of logic, or forget what it was doing, or confabulate something. But over time it became clear to everyone that the models would be very good at a wide range of economically valuable tasks. The work in which the Indian IT firms had specialized — routine code generation, application maintenance, and testing — could be done at an extremely low marginal cost, far below what even Infosys or Cognizant could offer.

The doubts arrived almost immediately. In February 2023 — less than three months after ChatGPT’s release — J. P. Morgan published a note warning that generative AI would be a “deflation driver” for legacy IT services, and that consulting firms like Accenture and Deloitte would gain market share over Infosys and Wipro. India’s own 2024 Economic Survey, presented to Parliament by the Finance Ministry, cautioned that AI “casts a huge pall of uncertainty as to its impact on workers across all skill levels,” and warned that employment in the BPO sector was “estimated to decline considerably in the next ten years.” By early 2026, the mood had turned to something closer to panic. At the India AI Impact Summit in New Delhi in February, the venture capitalist Vinod Khosla declared that IT services and BPO would “almost completely disappear within five years.” Less than two weeks later, the New York Times ran a long piece suggesting that AI is “starting to shrink."

Investors’ expectations about the future prospects of the Indian IT industry have certainly soured: TCS, Cognizant, and Infosys have each shed about 30 percent of their market capitalization over the course of 2026. And — like many American tech companies — the Indian IT giants have started to shed workers. In July 2025, TCS announced the largest layoff in the history of Indian IT, releasing 12,000 employees in a single exercise. In 2024, headcount at Infosys fell for the first time in 22 years, and graduate hiring fell from 85,000 in 2022 to 12,000 in 2024. Wipro cut 25,000 jobs over two years; Cognizant announced “Project Leap,” a restructuring expected to cut about 15,000 jobs.

And this has been a disaster for the graduates on whom the IT firms have so long depended. IIT Delhi, the flagship university of the IIT system, saw its placement rate fall to 61 percent in 2025; across all 23 IIT campuses, roughly 8,000 students went unplaced, more than double the figure from two years earlier.

But there’s little sign, at least for the moment, that Indian IT companies are actually losing revenue because of AI. Revenue growth has slowed — from double digits a few years ago to low single digits in nominal terms — and overall headcount has been roughly flat. But there’s no sign of a collapse just yet. Margins at each of the major IT players have expanded over the last few years: In 2025, Infosys posted an operating margin of 21.1 percent, its best in three years. And, much like American tech companies, they are rebranding themselves as “AI-native”: TCS reported that it is already earning $2.3 billion in cash it labeled as “AI revenue.” A lot of this new AI business is concrete enough: A typical engagement might involve a Fortune 1000 company asking Infosys or TCS to build a retrieval-augmented generation tool on its internal data, or to integrate large language models into its existing enterprise workflows. The IT firms, in other words, are not being replaced by AI; they are being hired to implement it.

So the Indian IT firms are refashioning themselves to be somewhat more like American tech companies: lighter on labor and heavier on profit. And indeed they’re quite well-positioned to accomplish this pivot. India’s IT majors enjoy longstanding relationships with countless legacy enterprises; they’ve managed to gradually upskill themselves, moving from low-end programming gruntwork to higher-value work spanning cloud migration, cybersecurity, data analytics, and enterprise consulting; and in many cases they have more visibility into the work they do than any of their clients. In the future, if AI obviates the need for human programmers, Indian programmers will be as irrelevant as American ones; but in the immediate term, Infosys and the like are well-geared to expand their margins. Mohit Joshi, the CEO of Tech Mahindra, has described it as a shift from a “pyramid” to a “bulging middle”: fewer juniors at the base, more senior generalists working alongside AI tools, and a much smaller overall workforce generating more revenue per head. Revenue per employee at TCS rose to nearly $50,000 in fiscal year 2025; at Infosys, $60,000; at HCLTech, over $61,000.

So — at least for now — the Indian IT industry’s business is not collapsing; only its labor model is breaking. HCL’s CEO, C. Vijayakumar, suggested in February 2025 that “the time is out for that model”: The company would seek to “generate twice the revenue with half the workforce.” The decoupling of revenue and employee count “will happen and in the next few years, it will be much more pronounced.”

But this might not be the case in a few years.

If there’s any single type of company threatened by the prospect of powerful AI with superhuman coding abilities, it’s the companies selling cheap software labor. It’s for this reason that the huge semiconductor rally of the last year has been accompanied by a slump in the valuations of the Indian outsourcing firms: TCS, for example, is down more than 30 percent. It’s not that the IT giants are shriveling right now; but public markets seem to believe that the net present value of their cash flows are a good deal lower than they were a few months ago.

What if that comes to pass? What would that mean for the Indian IT dream?

The grim scenario is not hard to sketch. As AI models improve, the revenue that Indian IT firms earn from Western clients gradually declines: The work that once required fifty engineers now requires perhaps 10. (And even that is being generous.) So India sees its service exports, built up over decades of IT sector expansion, stagnate and eventually decline. At the same time, Indian firms themselves become consumers of AI: Through the medium of tokens they effectively import the outputs of American-owned models and American-owned compute. So India’s exports fall and its imports rise.

And perhaps this balance-of-payments erosion is mirrored by a weakening economic situation at home. The high-earning workers of the IT sector now sustain an enormous downstream economy of domestic consumption; as the ranks of the IT sector thin one could see the contraction ripple outwards into the rest of India’s economy, touching the many service workers whose livelihoods depend on their consumption.

For the tens of millions of Indians for whom IT was the pathway to a better life, and for the Indian economy more broadly, the question is a simple and disturbing one: What then?