Do We Really Need a (High Volume) Wafer Fab ?

By Dennis Posadas

Reposted from KatipunanBGN© with explicit permission from the author. 

A copy of this article and others similar to it may be found at http://www.geocities.com/katipunanforum/index.html  

In the early days of the chip industry, the battlecry was “Real Men Have Fabs”. Nowadays, with every country in Southeast Asia aspiring to build a fab, we might see a situation where excess fab capacity exists, and no one makes any money. Consider these statements in BusinessWeek …

            “In Shanghai alone, plans have been announced for about a dozen plants that by 2005 will be capable of making a half-million wafers a month….Like CSMC, many of them are foundries, or plants dedicated to contract manufacturing--an industry pioneered by Taiwan Semiconductor Manufacturing Co. (TSMC ). Other big plants are in the works in Beijing , Shenzhen, and Leshan in Sichuan Province ….From Malaysia to South Korea , huge wafer fabs built by ambitious tycoons and governments sit idle or severely underutilized. Even foundry king TSMC is operating at less than half-capacity. A glut of new Chinese plants could depress prices.” (BusinessWeek, “ China ’s Chip Binge”, January 21, 2002 )

In other words, if its cheaper to rent something, why build it ? Of course, I am oversimplifying the situation here, but if you have a situation where you have ten or more wafer fabs trying to outdo each other in a price war, why join them ? Wouldn’t it be better to work the situation to your advantage ?

“TSMC accounted for half of that volume, and Chang predicts that Taiwanese foundries will process 40% of the world's wafers by 2010. Meanwhile, hoping to emulate Taiwan 's success, China , Malaysia , and Singapore plan to build 15 wafer fabs, five of them 300-mm behemoths destined to serve as foundries. As a result, SEMI expects foundries, located mainly in Southeast Asia , to be churning out half of all chips by decade's end.” (BusinessWeek, “Chips on Monster Wafers”, November 11, 2002 )

Which brings me to my second point. While it may not be a good idea to build a volume manufacturing wafer fab here in the Philippines (from a business perspective, if the assumptions above hold true), it is definitely a good idea to put our money in building up the chip design capability AND to setup a university prototype/training wafer fab facility.

Now before we worry about the costs of setting up even a prototype/training wafer fab facility, because of the glut in the world economy, we actually have a situation where it is a buyer’s market out there for fab equipment. If ever there was a time to setup a prototype/training wafer fab, it is NOW. Firesales and bankruptcies are going on in the developing countries, with equipment going for unheard of prices.

The current and severe IC downturn has put enormous pressure on semiconductor makers to cut their manufacturing costs--a trend that has propelled the used or secondary fab-equipment markets, especially in China .

The used chip-equipment market has been around for years. To complement their tool procurement efforts, semiconductor makers, IC-packaging providers, and chip-testing houses over the years have procured a range of low-cost, used fab equipment from third-party companies, brokers, and even Web-based auction sites. ( November 19, 2002 Semiconductor Business News)

For those who would argue that all we need is the chip design capability, and our value added is just to sell the design, that is actually just skimming the surface. While there are customers for that sort of work, most people look for a product (and someone to blame for that product). In short, most people want the actual part, not just the design. What we should aim for are fabless chip companies with niche products, not just pure design companies. This is where the training fab comes in. We can get away with utilizing the contract fab foundries, but if we don’t have a training fab, our relationships w/the contract foundries might not be productive.  We can insure that our design engineers at least know how a fab works, therefore allowing them to design better products. We will also be training the people who need to deal with the customers and the fabs, the failure analysis and the chip marketing folks.

Now before anyone argues that these older generation fab equipment are already useless, consider this. You can actually run MEMS on older equipment such as those used for 6” wafers. And a lot of these used but perfectly good equipment are sitting out there at rock bottom prices. Get the picture ?

Just in case you didn’t know, here is some good news. The University of the Philippines EEE Department has been training their students to design chips for some time now. With a corporate sponsored Microelectronics Lab, these students are already sending their designs for fabrication at TSMC in Taiwan , after they’ve designed it with industry standard chip design software. Just a few meters away at the UP National Institute of Physics, they are growing optical devices like SiGe and GaAs on an actual low volume fab equipment donated by the World Bank. Over at the UP Metallurgical Engineering department, they also have a low volume wafer fab facility called the Shono lab (named after a retired Sophia University professor). Already, other schools are also trying to setup their own VLSI design capability, like Ateneo, DLSU and Mapua. But these need to be complemented with a training fab facility.

In order for us to play in the higher value added arena of semiconductors and electronics, we need to have chip design startups from these university training programs, and invest in at least one local prototype/training wafer fab facility. We can cycle engineers and technicians through this training fab facility, so that if our local upstarts have any yield problems dealing with the contract manufacturing fabs, they would know how to handle it – even if they have to fly off to these neighboring fabs a few hours away. In short, we need a prototype/training wafer fab to be able to take advantage of the excess fab capacity in the region, not because we want to have our own high volume wafer fab.

The Diliman-Katipunan area (see separate feature) is a good place to start. We already have some equipment and expertise we can use as a base to start a “Silicon Alley”, one that we can position against Bangalore , and other regional technology centers. While we may have to move it elsewhere in the future, for the moment, Diliman-Katipunan is the most suitable area to start something that can run immediately.