Should the State subsidise a semiconductor fab?


Business Standard, 15 March 2006


The `Semindia' project has made waves in recent months. A bright future is held out for India in making integrated circuits (chips). However, there is no case for the government to get involved in such a project through equity capital, subsidised financing, or land. The government must restrict its activites to facilitating public goods such as infrastructure. What factories are built once these public goods are in place is best left to the private sector.

In the 1980s, there was a great scare in the US when Asian chip vendors were taking away market share in memory (DRAM) chips. It was argued that DRAMs were of strategic importance for the US. I have grown to be cynical about the word `strategic'. When someone says that DRAMs are strategic or oil supplies are strategic, that is often a code word for saying that we should suspend common sense and agree with whatever is about to be proposed.

Big government swung into the game in the US, with a project named `Sematech'. Many years later, we can evaluate the results. The memory chip business moved first to Japan and then to South Korea and Taiwan. It is now understood to be the low-margin, capital-intensive, commoditised end of the business. There is little profit to be made in making memory chips. The high profit end of the business is in CPUs, and this stayed in the US anyway owing to the top-end human capital which was in the US. Looking back, the moral of the story is: there was no role for industrial policy in the US on the subject of fabs.

This is yet another example of one big lesson that economists have understood for a while now: that "Industrial Policy" is a bad idea. Industrial policy involves a government that runs `corporate welfare programs' for some firms. The government is supposed to choose what industries are "good", and what management teams are "good", to put public money behind them. But does a government know more than an entrepreneur about what the future holds? And can a government ever choose the best management teams to back? Picking industries and picking teams are things that a venture capitalist does, not a government.

There is no part of technology that is as glittering and impressive as a semiconductor fab. Any politician who is taken for a tour of one will be highly impressed. If the politician is not well versed in economics, he is not allergic to industrial policy. It is then easy to talk him into putting public money behind such a project "of strategic importance to the nation". This adds up to the worst of situations: where a politician writes a cheque, spending someone else's money, on something he doesn't understand.

Public sector participation for a fab in India is an entirely dubious idea. One problem is capital intensity. A fab costing $3 billion might employ a few thousand people at most. It is a capital intensity of roughly Rs.4.4 crore per employee. In contrast, the mainstream services exports work in India has a capital intensity of atmost Rs.0.1 crore per employee. With a capital intensity that is roughly 44 times bigger than labour-intensive services work, it is not obvious that India has a natural comparative advantage in the fab business.

A particularly unhappy feature of the Semindia project is talk of 1200 acres of land. A standalone fab costing $3 billion occupies 25 acres (e.g. TI's DMOS 6 fab in Dallas in Texas). A full mini-city with a fab costing $3 billion occupies 75 acres (e.g. Intel's Fab 18 in Kiryat Gat in Israel). It is hard to explain politically allocated land of 1200 acres for Semindia.

In the budget speech last month, Mr. Chidambaram promised support for the semiconductor industry: "I propose to use the existing vehicles of viability gap funding and the India Infrastructure Finance Company Limited (IIFCL) to create a window to provide equity participation and/or viability gap funding to the new ventures". It is not clear to me what public goods are produced when someone builds a fab. It is just a factory. Just as GOI should not be involved when someone builds a factory that makes trucks, GOI should not be subsidising a factory that makes semiconductor chips.

All this is not to say that fabs are not a delicious prize. India should, of course, be thrilled if a serious semiconductor firm like Intel, TI, AMD or IBM voluntarily chooses to build a fab in India. This will depend on what we do on roads, water, electricity, and airports. A politician suffering from fab envy should do his part, which is to foster roads, water, electricity and airports.

While fabs have very little employment, they require top quality Ph.Ds. This is where (say) Israel wins hands down over India - they have world class universities while we do not. If a minister is worried that India does not have fabs, he should visit Israel to learn how they manage their universities better than we do. The business case for a fab should only be judged by private sector equity investors, never by a politician, and the fab should always have to buy land from the private sector at market prices.

If the government does want to foster high technology in India, interventions need to be designed in a way which have a public goods character. The most effective way to do this is to emulate the Defence Advanced Research Projects Agency (DARPA) of the US. DARPA gave out thousands of small grants to scientists to build defence-related technologies. DARPA contracts were awarded by a meritocratic system to the best teams, regardless of whether they were at companies or universities. Contracts had to yield tangible deliverables, which helped to keep the pressure up. DARPA tended to fund a large number of small projects, which spread the risks and increased the chances of reaching a few geniuses.

A DARPA-style meritocratic mechanism for giving out research contracts is the best way for government to foster high technology. There are enormous spillovers of knowledge to the economy when teams in the country learn new things. There is a public goods aspect here. As an example, the global Internet grew out of technologies built on DARPA contracts, and US citizens have reaped good returns to the public investment in DARPA funding. Such an approach is cheap when compared with the financial outlays involved in a fab.


Back up to Ajay Shah's 2006 media page
Back up to Ajay Shah's home page