CAN India's IT industry do to the West's IT giants what Wal-Mart has done to rival retail firms, or Dell to computer makers? The Indians talk a good game. “The productivity growth of Indian IT services is the highest in the world,” says Mr Narayanan at Cognizant. He should know: one-third of his firm's employees are in America and two-thirds in India. Nandan Nilekani, the chief executive of Infosys, goes further. “Almost everything that is done can be done by us faster, cheaper and better,” he says.

The argument for an Indian takeover of the world goes something like this. Like Dell and Wal-Mart, companies such as Infosys, TCS, Wipro and Cognizant source their offerings from poor, cheap countries. Wal-Mart has grown by adding Chinese-made toys, clothing and household appliances. Dell has added printers, hand-held devices and televisions to its line of made-in-Asia computers. In the same way, predict the Indian firms breezily, they will grow by adding new lines of IT services, offering global standards or better but produced at Indian costs. Investors understand this, say the Indians. Accenture's revenue is 14 times that of Infosys, but the American firm's market value is only one-third higher than that of its Indian competitor.

IBM and Accenture have been recruiting in India to lower their costs in areas where the Indian firms have grown fastest, such as maintaining popular business-software packages. But these global firms are so large (IBM employs 340,000 people; Accenture 100,000) that hiring even 10,000 extra staff in India has made little difference to their overall costs, most of which are still incurred in rich, expensive economies, the Indian firms point out gleefully. “The multinationals will never be able to restructure their costs fast enough to shift their centres of gravity,” says Arindam Bhattacharya of the Boston Consulting Group in New Delhi.

Moreover, because the Indian firms know India better than their American and European rivals do, they can grow (and are indeed growing) more quickly and more cheaply in India than anyone else. This will lower their costs even further. “We're adding close to 5,000 people in India this year,” says Mr Narayanan. “No American company can do that.” However, Accenture may recently have grown far more quickly in India than it can easily manage—though it bristles at the suggestion that it is finding India unusually difficult.

Wal-Mart sells commodities, such at microwave ovens at $28. In commodity businesses, the firm with the lowest price, which is often achieved by selling at the highest volume, wins the most customers. But not everything the IT industry sells is a commodity.

Broadly, the industry has three layers. The bottom one consists of businesses that have clearly become commodities. These are ruled by common standards, as in IT hardware manufacturing (where high-volume, low-cost Dell operates). A lot of this has moved to Asia.

The top layer is made up of tailored, bespoke technology services. Accenture, for instance, advertises work it has done for a large Australian casino to introduce a tracking technology, called Radio Frequency Identification, to improve the way the casino handles the 80,000 bits of staff clothing it has dry-cleaned every year. IBM is working with an American limousine-fleet company to introduce the same mathematical models the airline industry uses to route aircraft. Atos Origin, the European IT-services firm, is working with a British government agency, the Vehicle and Operator Services Agency, to equip its inspectors with hand-held computers to help them decide which passing vehicles to check. Because these services are tailored to meet the needs of individual customers, they are likely to continue to be provided close to the IT industry's biggest customers in America, Europe and Japan.

That leaves a large block of services sandwiched in the middle. These services are on their way to becoming commodities as shared standards spread. The ready adoption of a small number of business-software packages sold by firms such as SAP and PeopleSoft, for instance, is making the maintenance and even the installation of such software increasingly routine as these popular packages are becoming de facto standards. It is this large middle layer of services that is currently feeding the rapid growth of Indian firms such as TCS and Infosys.

Champions of the Indian firms look at the industry's employees and see a large bulge of people offering this middle layer of IT services, with a thinner sliver of business consultants doing the bespoke work on top. This makes them think that it should be far easier for the Indian firms to move up to that top layer by hiring consultants in America and Europe than for western IT firms to shift most of their employees (and their costs) from rich countries to poor ones. “About 20% of our value is added near our customers in America and Europe and 80% here in India,” says Infosys's Mr Murthy. “If IBM wants to replicate this, it needs 80% of its employment in less developed countries as well.”

This analysis neglects several important points. Perhaps the most crucial of these is that patterns of demand in the IT industry have shifted in the past, and may well do so again. Ten years ago customers spent a much bigger chunk of their IT budgets on computer hardware than they do now. Between 1993 and 2001, calculates Catherine Mann of the Institute for International Economics, spending on software and services grew by 12.5% a year, nearly twice as fast as hardware spending, pushing the share of software and services in overall expenditure from 58% to 69%.

As Ms Mann points out, the movement of IT hardware manufacturing to low-cost Asia helped to finance this shift in demand, because falling hardware prices freed up money to spend on software and services. Likewise, thinks Ms Mann, the migration of commodity IT services to low-cost places such as India will leave companies with more money to spend on the top-end bespoke services, which will help to expand this category of work.

If the world's IT giants want to remain big, they will have to change to meet changing demand. IBM has already performed this trick once. At the beginning of the 1990s, the company was mainly a hardware manufacturer. By the end of that decade, it had shifted much of its weight into IT services. Now, says IBM's Mr Harreld, the firm needs to move its high-cost employees into tailored services as commodity services migrate offshore.

Mr Harreld predicts that demand for such bespoke services will grow strongly, and that it will be many years before everything the IT industry sells becomes a commodity. To support his argument, he turns to Carlota Perez, an economic historian. In her book, “Technological Revolutions and Financial Capital”, Ms Perez traces five boom-and-bust cycles of technological innovation: the industrial revolution; steam and railways; steel, electricity and heavy engineering; oil, cars and mass production; and information technology and telecommunications.

In each age, argues Ms Perez, a phase of innovation, fuelled by hot money, has been followed by a financial bust, and then by an extended period in which the technology is deployed properly. Having just emerged from its bust, the information age is only at the beginning of this long deployment period, says Mr Harreld. Proper deployment, he argues, will require a large number of people working close to the industry's customers, in the way that IBM is doing for its limousine-fleet customer, or that Atos Origin is doing for Britain's vehicle-safety agency.

Two questions remain. The first is how long it will take for the large middle layer of services to become a commodity. If this happens too quickly, companies such as IBM, EDS and Accenture may find themselves overwhelmed by the pace of change, just as IBM nearly found itself ruined by the shift of IT manufacturing overseas in the early 1990s.

Of the three giants, EDS is in the weakest position. Having struggled with financial troubles and management turmoil at home, it has done little so far to counter the threat from Indian competitors, who are eating into large chunks of its business. Other smaller IT-services companies, such as BearingPoint and Capgemini, may also struggle with the shift of services abroad.

Most services in the middle layer, however, are likely to move offshore at a fairly manageable speed. That is because the IT organisations of most large companies tend to be a tangled mess of overlapping systems which go wrong so often that, as a practical matter, it will be hard to move IT work anywhere without fixing the systems first. To illustrate this point, Mr Harreld produces a diagram showing the different systems of one of IBM's customers, along with their interconnections. It is so intricate that it might pass for the design of a semiconductor chip. IBM itself runs 17,000 software applications, a figure that Mr Harreld thinks can comfortably shrink to 10,000 in due course.

The other big question is how easily companies such as Wipro, TCS and Infosys can expand into that upper crust of bespoke services that Mr Harreld predicts will flourish close to the industry's customers in rich countries. The Indian firms have lots of cash to spend: the cost of an Indian programmer is so much lower than an American one that Wipro and Infosys are earning fat profits on lines of business that may be only just profitable for big western companies. So far, the Indians have spent their money cautiously, making small acquisitions and hiring the odd western consultant from rival firms.

If they are serious about taking on companies such as IBM and Accenture, the Indian firms will have to act more boldly. Yet buying or building people businesses of this kind is notoriously difficult. Time and again, and in all sorts of industries, from banking to telecommunications, America's and Europe's best managers have tried and failed miserably. Moreover, the competition is well entrenched. IBM, for example, has built up good relations with its customers over decades. The Indian companies may yet find that the only thing they can do faster and better on their rivals' home turf is to lose their shirts.

Copyright © 2004 The Economist Newspaper and The Economist Group. All rights reserved.