The services industry did not fail India. It succeeded so thoroughly that it became the air.
India's IT services industry is one of the great economic success stories of the last three decades: roughly a quarter of a trillion dollars a year, five and a half million jobs, and one of the largest sources of foreign exchange the country has. Everyone agrees. That agreement is the trap.
India's economic ceiling is not only a shortage of talent or capital. It is a habit of selling hours instead of owning products, held in place by an incentive system that rewards the hour. That habit paid for a generation. It is about to become unaffordable.
The country that codes for the world earns roughly fifty to sixty thousand dollars per engineer per year. Measured per engineer, a global product company's revenue is several times that. An Indian product company such as Zoho earns only a little more per engineer, on the order of sixty-five to seventy thousand dollars. The difference is not the person. It is the contract, and the market the contract sells into. The model has a human shape too: an engineer between projects, paid but unbilled, waiting for an allocation code that decides city, technology, and client.
India codes for the world, but the code belongs to the client. Under the standard services contract, the deliverable is work for hire. The client owns the code, the design, the intellectual property. Under the default contract, the vendor is paid for the hours and cannot reuse what it built; some contracts let the vendor keep background tools, but the default transfers ownership. An engineer may solve the same problem for five different clients, and the firm captures none of the compounding.
India codes for the world, but the revenue scales with people. A services firm's revenue is billing rate times billable hours times utilized staff. There is no inventory and no compounding asset. When the project ends, the revenue stops unless another project replaces it. Every new dollar of revenue requires roughly a dollar of additional engineer time. Gross margins in pure services work run twenty-five to thirty-five percent. In product software, seventy to eighty, because the marginal cost of one more user is close to zero.
India codes for the world, but it spends less than one percent of its GDP on research: roughly six to seven tenths of one percent, against two point four in China, three and a half in the United States, four point eight in South Korea, five and a half in Israel. Foreign companies file a large share of the patents in India, and Indian residents are a minority of applicants in most recent years. The country that codes for the world does not own the science that the code runs on.
India codes for the world, but its largest firms grow by adding engineers, not products. TCS, Infosys, Wipro, HCLTech, and Tech Mahindra together employ well over a million people and generate combined revenue in the range of seventy-five to eighty billion dollars a year. Their revenue per employee is a fraction of what a product company's is. The pyramid is the business model: a wide base of freshers doing implementation and testing, a thin top of architects and client managers. The pyramid is also the ceiling.
India codes for the world, but the thing it sells is the thing a machine can now compress. AI coding tools can compress routine coding work. In a 2022 study, developers completed a specific coding task about fifty-five percent faster with an AI assistant. Under a time-and-materials contract, that is a revenue cut. Under a fixed-price contract, the client learns the new speed and renegotiates the next one downward. The client captures the gain. The vendor absorbs the loss. The billable hour, the unit the whole model is priced on, is the unit under attack.
The strongest case against this argument is not that the numbers are wrong. It is that the habit was never a habit. It was a rational response to what India actually had. Three decades ago the country had a large pool of English-speaking, technically trained engineers and almost none of the things a product company needs: patient capital, a domestic market large enough to fund version one, courts that move at the speed of business, a safety net that lets a twenty-six-year-old bet five years on an idea that might not work. Services were the highest-return use of what existed. The industry took the talent and turned it into dollars, and those dollars became apartments, engineering colleges, plane tickets abroad, and eventually founders. Five and a half million jobs and roughly two hundred billion dollars a year in exports are not a rounding error. Many of the founders of the product companies that do exist came out of services. The services industry was the ladder, and the people who climbed it are the ones now building products. To call the whole thing a trap is to mistake a ladder for a cage.
The steelman is right that the barriers were real. Capital was thin. Domestic markets were small. IP enforcement and contract courts were slow. A salaried job was the rational choice for a graduate supporting a family. But the barriers have moved, not vanished. Early-stage capital in India was thin for decades, and Indian founders still often incorporate abroad to reach global money and customers. That is a market-structure fact, not a character trait. India still spends under one percent of its GDP on research. Its domestic enterprise software market is still small next to the United States or Europe. Those gaps are real. What has changed is that they are no longer the whole story. Indian product companies exist now, and they are growing. Founders who built them came out of services, which means the talent was always there. Global venture money reaches Indian founders, even if many still route through a Delaware or Singapore holding company. Indian product revenue is orders of magnitude smaller than two hundred billion dollars in IT services exports, so the ownership gap is still enormous. But it is no longer zero. If the only problem were missing inputs, the product share would track the arrival of capital and talent. It has not, or not fast enough to explain the gap by inputs alone. The firms that dominate the industry have product arms. They are real. They remain a small share of revenue. The shape of the firms has not changed. That is consistent with an incentive system that rewards the hour. It is not proof of it, but it is the pattern the numbers show.
The path does exist, and it is walked by more than one founder. A founder from a town in southern India left a US engineering career, took no venture money, stayed profitable, and built a product company from a village. A helpdesk product born in Chennai listed in New York. An API platform and a payments company came out of Bengaluru. They prove the product path is walkable. They do not prove it is easy, or that it has changed the shape of the industry. The southern founder argues that Indians should own products rather than rent out their brains, that services scale linearly with headcount and so are a ceiling, that AI will hollow out code for hire. His company ships dozens of products from rural offices, sells globally, and employs thousands. It is proof that the path exists. It is also not proof that it is easy. The same founder has drawn criticism for public statements well outside business, for the rural model's pay scales, for the limits of a no-funding stance. The product path exists. It is not a template. It is a set of stubborn choices, and stubbornness is not a national strategy.
The cost of the habit is not abstract. It shows up in the engineer between projects, paid but unbilled, waiting for an allocation code that decides city, technology, and client. It shows up in the fresher who trained for months on a campus and then waited for a project. It shows up in the pyramid itself: a wide base of young engineers doing implementation and testing, the work most exposed to automation, billing hours that a machine can now produce faster. The growth engine of the industry was to hire thousands of freshers, train them, bill them. That engine stalled. Headcount at the largest firms fell in fiscal 2024 for the first time in roughly two decades, by a combined sixty thousand or more. Fresher intake at the largest firm dropped from over a hundred thousand to roughly forty thousand in two years. Attrition, which peaked above twenty-five percent in 2022, fell to twelve to fifteen percent. In a services firm, low attrition is not a sign of contentment. It is a sign that there are fewer new projects to leave for.
India has proved it can build products. What remains is whether a habit that paid for a generation can be unlearned by a generation that still remembers the pay. The services industry did not fail India. It succeeded so thoroughly that it became the air: five and a half million jobs and roughly two hundred billion dollars a year in exports are not a rounding error. The gain is real. The ceiling is real. You cannot ask a country to climb down a ladder it is still standing on. Whether that ladder still reaches anywhere is the question it now has to live with.