IT services and Bengaluru

The usual explanation of Bengaluru runs in one direction. The Indian Institute of Science opened in 1909, the engineering colleges followed, the graduates stayed, and therefore the software industry grew where the engineers already were. It is satisfying because it makes the city the author of its own fortune. It also gets the order wrong.

In 1980 the city held a dense concentration of people who could build, calibrate and repair precision instruments, and almost no way to sell that skill abroad. What changed between 1985 and 1995 was not the supply of engineers. It was a short run of administrative decisions that made it possible to move work across a border at a predictable cost — taken in a city that already had the machine shops. Industry first, people second, rents third.

A base built for something else

Hindustan Aircraft was set up here in 1940, overhauling aircraft and later building jet engines under licence. Bharat Electronics followed in 1954, the National Aeronautical Laboratory in 1959, the headquarters of the space programme in the same decade, and a handful of defence research establishments. The result was an aerospace town, not a software cluster.

Reading that base as a software cluster in waiting is the standard mistake. What it left behind was narrower: a labour market of people who had worked to a tolerance and knew what a calibration certificate was; dealers in components and test equipment; and an administrative habit — a public laboratory could raise a purchase order for an imported instrument, and the file did not die in a drawer. Software export in its first decade needed exactly that: someone who could keep a room of imported machines running, and a purchase order that cleared customs.

For ten years the scarce input was the wire

Offshore work has an awkward property: the customer has to read the code while it is being written. In the mid-1980s an international data circuit here was a state monopoly, leased from VSNL, priced beyond what a small firm could earn, and slow to install. The bottleneck was not desks. It was bandwidth to the border.

In 1985 Texas Instruments opened a design centre here with a captive satellite earth station — the first such private link approved for a foreign-owned unit. The interesting part is not that a chip company arrived. It is that an approval existed: a file in the Department of Electronics establishing that a private firm could send work out of the country over its own dish. One exception is a precedent; a policy is a product.

The 1993 rule that let park units lease dedicated satellite links through VSNL turned the exception into the product. It did not make bandwidth cheap. A circuit cost more in a month than a dozen junior salaries, which is why its unit of organisation was never the office but the campus: one dish, one generator, one wall, a thousand desks behind it.

STPI's real invention was an accounting category

The Software Technology Parks scheme, run from 1991 by a body under the Department of Electronics, is usually described as a tax break. Register a unit inside a park and you got single-window clearance, duty-free import of capital goods, a ten-year exemption on export profits, 100% foreign equity, and an export obligation in foreign exchange.

The more consequential part was definitional. India's export regime was built for goods: something left the country, a shipping bill was filed, foreign exchange came back against it. Software left the country as a file over a satellite link, with no shipping bill to file. The scheme manufactured a paper equivalent — a registered unit, a periodic statement of foreign-exchange earnings, a legal fiction in which an export had occurred. Once the export could be counted, it could be exempted, financed, collateralised and promised to a client in another time zone.

The same year delivered an unplanned subsidy. In July 1991 the rupee was devalued in two steps, by roughly a fifth against the dollar. For firms billing in dollars and paying salaries in rupees, that is a cut of a fifth in the wage bill in the currency the client uses — delivered in three days, mentioned in no policy document. The devaluation was not aimed at software. Software was simply best placed to collect it.

The firms came first and the graduates second

The companies that became the industry were founded before the labour market existed. Tata Consultancy Services was incorporated in 1968, selling programmers to foreign clients by the person-month. Infosys started in 1981 in Pune with ten thousand rupees and moved to Bengaluru in 1983, listing in 1993. Wipro was a vegetable-oil company that moved into computers in the late 1970s. The business was mostly body-shopping: send two engineers to the client's site and bill their time. That model exports people. It needs an airport, not a campus.

The shift to offshore delivery — work done here, reviewed there — is what created local employment at scale, and it depended on the link and the tax category above. The demand shock that filled the seats was the year-2000 problem: from about 1997 the world needed a very large amount of legacy code inspected, and India was the cheapest place with enough people able to read it. The export line roughly doubled between FY1999-2000 and FY2001-02. The NASSCOM–McKinsey study of 1999 gave the industry its projections and, more usefully, its vocabulary.

Engineering capacity followed the wage premium rather than causing it. Karnataka's intake into engineering colleges expanded steeply through the 2000s, a decade after salaries had already risen. Which is why the "talent pool" argument inverts the causality: the pool was filled by the industry's paychecks, and by families who could see what a good college seat was worth.

Rents are the last variable

Bengaluru's urban population went from about 4.1 million in 1991 to about 8.5 million in 2011, and the desks were placed in a sequence that drew the map. Electronics City was laid out in the late 1970s by the state electronics corporation, twenty kilometres south of the centre, then open land. The first international tech park opened in 1994 on the eastern edge. Whitefield and then the Outer Ring Road absorbed the next two waves. Each decision moved the rent gradient outward and turned a commute into a migration.

The park form is the industry's answer to a city that could not supply what it needed. A campus buys its own water, runs its own generators and buses, hires its own security, because a supply that fails for two hours is not a nuisance when the client is asleep in another time zone and expects a build by morning. The firm internalises the infrastructure. The city absorbs the traffic, the tankers and the sewage, and books the export earnings.

The same arithmetic that selected the city can unselect it. Bengaluru wages rose faster than in any comparable delivery location; work moved to Pune, Hyderabad and Chennai, then to Manila and Warsaw for voice, and to smaller Indian towns for back-office work. A place chosen because a set of rules made it the cheapest legal option can be unchosen when it stops being cheap. The engineers who arrived in the 2000s and bought flats on the ring road were betting that the arithmetic was permanent. It was not permanent. It was about twenty years.

This is one of the Modern India entries in About India. The pillar's argument — that each of these systems has a stated reason and an actual one — is on the pillar page.