India's IT Outsourcing Industry Was Forced Into Existence by a Foreign-Exchange Rule

Contents

1. A rule meant to save foreign exchange produced export orders
2. Why this rule grew an industry when most regulations don’t
3. Why it isn’t one of the easier explanations
4. The conditions under which the whole argument stops holding
5. A few other things worth remembering
6. Worth reading alongside
7. The one thing to take with you

The power of producing wealth is far more important than wealth itself.

— Friedrich List, The National System of Political Economy (1841)

Asianometry’s episode “The Birth & Rise of India’s IT Outsourcing Industry,” published 18 October 2026, pointed at something I had never considered: India’s IT export business got its first orders from a rule written to conserve foreign exchange. In 1974 Tata Consultancy Services (TCS), the Tata Group’s software arm, imported a Burroughs B1728 for $300,000 under a Software Export Scheme that let companies bring in hardware on one condition — earn back double the import price in software export revenue within five years. Indian software exports totalled only $4 million for all of 1980; by 2023 the figure was $194 billion. Two conditions had to hold for this to work: the work the rule forced on them (porting old software for Western clients) had real buyers, and the communications bottleneck had to come down, which it did when Texas Instruments ran a satellite link into Bangalore in 1985 and the Software Technology Parks finally opened in 1991.

A rule meant to save foreign exchange produced export orders

Set the stage first. In 1973 Indira Gandhi’s government, running a fiscal deficit, passed the Foreign Exchange Regulation Act (FERA), which clamped down on imports and exports and capped foreign ownership of joint ventures at 40%. The American computer maker Burroughs wanted into India and needed a partner; its managers got along with TCS general manager F.C. Kohli, and so they went with Tata. The following year TCS imported the B1728 under the Software Export Scheme: you may import the hardware, and you commit to earning back twice its price in software exports over five years.

The intent was to protect reserves. The result was that an Indian company got pushed out of the country to find customers. Among TCS’s first overseas engagements was a $24,000 project porting a hospital accounting package from Burroughs medium systems COBOL to Burroughs small systems COBOL. Porting work made up something like 20–40% of the early business of India’s IT exporters. It was drab, tedious, nobody wanted it, and every line had to be watched, because getting it wrong broke the client’s books.

I stopped the episode there for a moment. This reads less like visionary industrial policy and more like a group of people squeezing out through a gap in a door because the rules left them no other way.

Why this rule grew an industry when most regulations don’t

India had no shortage of regulations in that era, and most of them produced shortages and nothing else. Three conditions held here at the same time.

First, the work they were forced into had real buyers. Porting legacy code was the job nobody in a Western firm wanted, so handing it to someone else felt like relief. That matters: if a rule forces you to produce something no one will buy, you make junk to satisfy the paperwork and close when the five years are up. What TCS was pushed out to sell happened to be work Western enterprises were glad to pay to be rid of.

Second, the skill compounded and transferred. Years of porting build deep familiarity with operating systems and compiler environments. There is an underrated turn here: India restricted imports of finished software and left source code imports alone. UNIX source circulated widely in the 1970s, so minicomputer makers like DCM and HCL adopted it fast, especially the Berkeley variant — and in 1985 the Reserve Bank of India recommended UNIX as the standard operating system for the country’s banks, which pushed it further. By the time Sun’s Solaris, IBM’s AIX and HP’s HP-UX were everywhere in American corporate machine rooms, an Indian firm could develop on whatever UNIX box it owned and know the output would run at a customer site. A gap in a trade restriction produced a generation of programmers fluent in the same environment.

Third, somebody had to tear down the physical bottleneck. India’s phone system at the time was among the most archaic and overloaded anywhere; the waiting list for a landline ran half a million names and two to five years. Without communications, Indian firms had one option — put engineers on planes and sit them at the client’s desk. As late as the end of the 1980s, 75–80% of India’s export earnings still came from that practice, known in the trade as body shopping. One detail from the episode has stayed with me: before the links existed, code travelled back and forth in pouches, carried by human couriers on flights.

The turn came in 1985, when Texas Instruments installed a satellite data link between the United States and Bangalore. India’s Department of Electronics went to bat for them hard; one veteran said proudly that they “broke 26 separate rules to accommodate TI’s Bangalore subsidiary and are willing to break more.” By law only the Department of Telecommunications could own and run earth station equipment, so the DOE wrote a paper agreement under which TI paid for the gear while the department operated it and rented it back. The DOE began promoting Software Technology Parks in 1986, though the parks did not actually open until 1991 because of infighting over who controlled the communications links. Once those links worked, what TCS called the “offshore model” could run: a small team goes to the client site to collect requirements and implement, while the bulk of development stays in India at a third of Western salaries. Onsite work fell from over 80% in the late 1980s to 59% by the late 1990s.

Remove any one of the three and the story collapses. Demand without skill gets you one order. Skill without bandwidth means you are selling plane tickets forever. Bandwidth without anything forcing you abroad means you serve the domestic market and stay there.

Why it isn’t one of the easier explanations

Several smoother versions of this history exist. I checked them one at a time.

Version one: IBM’s 1977 exit left a vacuum. IBM had been India’s dominant computer company since 1951, and the government came to feel it treated the country as a dumping ground for outdated machines. Talks on technology transfer ended on equity: FERA required dilution to 40%, Britain’s ICL complied, IBM treated 100% ownership as global policy and left. The episode is honest about what followed — in hindsight, the opportunities that actually emerged had little to do with what IBM had been doing, but the perception that opportunities existed was enough. So IBM’s exit explains why so many people jumped in. It does not explain where the orders came from.

Version two: cheap English-speaking engineers. Both of those were present the whole time, and India’s total software exports in 1980 came to $4 million, with TCS and Tata Burroughs taking over 60% of it and roughly nineteen other firms splitting the rest. Salaries at a third of Western levels turned into money only once the satellite links made the offshore model possible. A price advantage without a delivery channel is just a cheaper plane ticket.

Version three: Y2K saved India. Y2K did pay beautifully — about $2.3 billion for the Indian software industry in the late 1990s. TCS revenue went from $125 million in 1995 to $419 million in 1999 while holding 32% margins, and Infosys drew over 20% of its 1998 and 1999 revenue from Y2K work alone. TCS opened a dedicated Y2K office in Chennai where a thousand programmers churned through two million lines of COBOL a day on a semi-automated 17-step process. But being able to assemble a thousand COBOL programmers in 1998 was the output of twenty-five years of porting work. Y2K was the harvest.

Version four: the 1991 liberalization. It dismantled the License Raj, devalued the rupee and opened the door to foreign investment; Infosys went public in 1993 and built its campus with the money, while the flood of imported hardware pushed Wipro (previously a vegetable oils manufacturer) and HCL into the services model. All true, and the Software Technology Park policy preceded it by five years, with the link the offshore model needed already installed in 1985. Liberalization scaled the industry up from a starting point that came earlier.

The conditions under which the whole argument stops holding

This is the part I keep turning over. The capability that rule created was aimed at one specific scarcity: Western enterprises could not find or afford enough programmer hours. The industry’s rent came from that scarcity. When a scarcity moves, the capability built around it loses its place at the till.

The numbers have turned. Indian IT grew 8.4% in 2023, then decelerated to 3.8% in 2024, with incremental revenue falling from $19 billion to $9 billion. Headcount tells it more plainly: per Nasscom, the industry added almost 300,000 jobs in 2023 and only 60,000 in 2024, while the largest firms shrank. Infosys shed roughly 26,000 people in 2024 through attrition and reduced hiring, its first decline as a public company. TCS fell by 13,000 in 2024, its first drop since 2005, then laid off 2% — some 12,000 people — in July 2025, the largest in its history. Since December 2024 both Infosys and TCS have seen their share prices fall 40%, with multiples compressing alongside.

The episode points at AI, and draws a parallel that gave me a chill: the early work being handed to tools like Claude looks a lot like what the IT exporters did — porting software, 24-hour development cycles where you submit a long-running job before bed, assorted knowledge work. When Anthropic announced Claude plugins for legal, finance and sales tasks in early 2026, the hit landed on SaaS names and on Infosys and TCS too.

Which runs straight into a question a lot of people are holding: something I own is down 40% — is that mispricing or a structural change? What I do is ask what rent the business collects, then ask whether that scarcity still exists. The Indian IT giants say generative AI will enhance their work rather than replace it, because enterprise customers still need integration, governance and trust, and rising productivity creates savings that can be passed to clients with some margin retained. There is something to that, and notice what it does: it moves the source of the rent from programmer hours to having someone willing to sign for the outcome. The first gets eaten by cheap tools. The second depends on who the client calls when things break. The episode’s own image is a crossing — they are wading through a fast-flowing river now, and we will see whether they all make it through intact.

The same question travels to any industry a new technology has swept across, including yours. When people come to you, is it because you hold a capability they cannot get, or because it is inconvenient for them to do it themselves right now? The first keeps paying for a while. The second lasts exactly as long as the inconvenience does. What that 1974 rule gave TCS was five years of pressure to get out of the country. What this moment gives the rest of us is pressure to work out which kind of rent we collect.

A few other things worth remembering

• The term “body shopping” describes the stage the industry was at. It carried a poor reputation, and early-1990s changes to the US B-1 business visitor visa plus worries about top talent hopping off made it harder to run. What pushed it down was the satellite links, not the disapproval.
• Tata nearly built computer hardware. In 1977 the government finally granted Tata a licence to manufacture Burroughs machines, and the Tata Sons office decided against it, setting up Tata Burroughs to import and sell them instead. The episode’s verdict: they took the easy way out, and India lost a shot at making hardware itself.
• Infosys started on $250 from a founder’s wife. In July 1981 N. R. Narayana Murthy left Patni Computer Systems with six colleagues. They could not afford a computer of their own, so every cofounder except Murthy was sent to sit at client sites in America and write code there.
• The offshoring jobs debate never settled on the data. Business Week ran “The New Global Job Shift” as a February 2003 cover story, citing a Forrester estimate that 3.3 million jobs and $136 billion in wages would move offshore, and the issue burned through the 2004 US presidential election. Economists’ objection is methodological: there is no way to establish whether a given Indian engineer’s work replaced work an American would have done, and those same years carried a post-bubble recession, rising US worker productivity, and a growing Indian industry buying more American services.
• One customer carried half of Wipro’s software revenue. Through the 1990s, GE accounted for 50% of Wipro’s software revenue and 20–30% of Infosys’s, and in mid-2005 the Wall Street Journal reported GE still made up 15% of TCS’s overseas revenue. For a long stretch, the industry’s credibility was borrowed.

Worth reading alongside

• Asianometry, “The Birth & Rise of India’s IT Outsourcing Industry,” 18 October 2026. The episode credits a coauthor, Zoso/Felix of the Zoso Project; the host says he often disagrees with him but has to think it over first, which is a recommendation I like.
• Nasscom’s annual figures on Indian IT revenue and employment, for the 2023–2024 growth and headcount shifts.
• Business Week, February 2003 cover story “The New Global Job Shift,” the opening document of the offshoring debate.
• TCS and Infosys public filings, for the Y2K-era revenue and margin figures and the 1999 Infosys ADR issuance.

The one thing to take with you

A constraint can force a capability into being, and that capability pays only while the scarcity it was aimed at still exists. The 1974 rule — import a computer, earn back twice its price in exports — produced a generation of engineers who could move old software onto new platforms, and that skill paid for fifty years because “the West cannot find enough affordable programmer hours” lasted fifty years. That scarcity is shrinking now, so the capability’s place at the till is moving.

Here is something I tried that needs only a sheet of paper. Write down three things someone recently came to you for help with — fixing a laptop, reading a contract, watching their kid, sorting a shift roster all count. After each one, write a sentence answering why they didn’t do it themselves. Then sort the answers into two piles: “they can’t” and “it’s inconvenient for them right now.” Put the paper away and look at it again a week later to see whether the inconvenient ones are still inconvenient. The first time I did this, two of my three landed in the inconvenient pile, and one of those vanished two months later because the person installed an app.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.