S & T » Science

Updated: September 18, 2013 19:04 IST

Soon, chips made in India

V. Sridhar
Comment (1)   ·   print   ·   T  T  

Union Cabinet has given ‘in-principle’ approval for two international consortiums to establish fabs

The Union Cabinet’s announcement last week that it had given ‘in-principle’ approval for two international consortiums to establish India’s first semiconductor fabrication plants was like a mood elevator for a patient in deep depression. The gigantic scale of investment, about $8 billion (more than Rs. 50,000 crores), could be just what the doctor ordered for the Indian economy. But industry insiders, while euphoric about the “historic step,” point to several “imponderables” on the road ahead.

The global semiconductor industry has been going through a downswing in the last few years. “This is not a surprise because, historically, the industry is known to be cyclical,” says Aninda Moitra, president and CEO, Applied Materials India.


Mr. Moitra is thrilled at the prospect of the establishment of the fabs because his company is a global leader in the fab equipment business — the machinery that goes into the making of the fab.

The ballpark figure for the cost of the equipment that goes into a semiconductor fab is about 55 per cent of the entire capital cost.

Three tiers

The global semiconductor industry, with sales of around $300 billion in 2012, can be broadly divided into three tiers. At the apex, populated by the likes of Intel, IBM, Toshiba and Samsung, are companies that are fully integrated. They not only design, but also make the chips that end up in the devices they make. This ensures that rivals do not commoditise the chips they make easily, which ensures them high returns for the huge investments sunk in.

In the middle are the “fabless” designers of chips, who outsource the manufacturing to the specialised chipmakers. In this group are companies like Qualcomm, Broadcom and Nvidia.

At the third level are the “foundries,” which have established huge capacities for a variety of chips that they produce for clients. Companies such as Taiwan Semiconductor Manufacturing Company, the biggest independent foundry, and Globalfoundries are in this league.

The late entrant’s dilemma

A late entrant such as India faces several dilemmas entering the high-stakes game of chip manufacturing.

The first is the obvious question of where to start. In the wafer business, the level of sophistication lies in manufacturing larger slices of wafers on which the chips are seated.

The current state of the art is at 300 mm, but by the time the Indian fabs get their act together, wafers may progress to 400 mm. At the level of the chips, the game is to make them smaller, while accommodating more transistors, and adding functionality. Indications from industry are that the Indian fabs will start with the 65 nm to 45 nm range.

Is it worth it for a late starter to enter a race at a point when the others have progressed even further in that time? Actually, this question is based on a ‘false’ ideal. If India does not begin now, it will be even further away three years later — which is when the fabs start churning out the chips if they were to start the project right away. So, rephrasing the question, asking how much would the Indian industry gain by starting now, rather than later, knowing fully well that there would be less catching up to do three years later, is more relevant.

P.V.G. Menon, president, Indian Electronics and Semiconductor Association (IESA), believes plenty can be done three years later with the 65nm to 45 nm range, even if the “cutting edge and high value chips” have moved to the 22nm to 20 nm range by then. In three years, the Indian fabs are likely to “load” only about 30 to 40 per cent of their capacity, catering to relatively “low end” products such as the Akash tablet and the lower end of the telecom sector’s needs. “At a loading of 50 to 60 per cent, the Indian fab would still be losing money. But at 80 to 90 per cent, it will start making profits, huge profits,” he observes.

He believes that a targeted capability of 65 nm to 45 nm is “well positioned.” However, the catch in this is the assumption that the fabs will start rolling out in 36 to 42 months from now. But if things go well, the development of the ecosystem and the learning-by-doing gained would make it possible for India to try and approach the cutting edge.

Incentives a must for success

A gigantic project such as this would obviously need government intervention at multiple levels, just as it has been in the case of every semiconductor superpower. One of the biggest myths of the business is that the business has grown simply on the backs of private enterprise. The industry’s evolution in Taiwan, China, Korea, the U.S. and Europe prove the contrary.

In India too, subsidies and incentives are bound to play a role in the success of the venture. Industry sources reckon that the Rs. 50,000-crore expenditure would perhaps need State support — including viability gap funding like many discredited road projects in the country — of around Rs. 12,500 crore. This may be in the form of outright subsidies, tax concessions or even direct government support, but it is obvious that State support is critical to the project.

“We need to see the semiconductor business as strategic, as important as national security,” says Mr. Moitra.

I have been reading news of such type from past 3 years, dont think the consortia who have committed would be ale to realise their initiatives considering the infrastruture required and the skilled manpower required in this highly specialised foundary business.

from:  Gaurav Pruthi
Posted on: Sep 17, 2013 at 18:07 IST
This article is closed for comments.
Please Email the Editor



Recent Article in Science

NASA's Dawn spacecraft heads toward the dwarf planet Ceres as seen in this undated artist's conception released on January 22, 2014.

NASA spacecraft nearing its second stop — dwarf planet Ceres

A NASA spacecraft is preparing to rendezvous with the largest object in the asteroid belt located between Mars and Jupiter.The Da... »