Three winners from Taiwan and Singapore
RIDING THE SILICON SHIFT
By Assif Shameen (Asiaweek )
WHAT'S THE LATEST NEWS you heard about the semiconductor industry? That there is a huge glut and margins are wafer thin. That little new capacity is being built. That Taiwan companies are far behind the Koreans and Japanese. All wrong - at least in one important market segment. Foundries, which process silicon into wafers and then turn them into semiconductors, are operating at virtually 100% capacity. The world's Big Three - Taiwan Semiconductor Manufacturing (TSMC), fellow Taiwan company United Microelectronics Corp. (UMC) and Singapore's Chartered Semiconductor Manufacturing (CSM) - are growing so rapidly that they and other foundries are forecast to produce 35% of all semiconductor products in 2010, from just 7% today.
Call it the silicon shift. Companies like Motorola and Texas Instruments have decided that the billions needed to put up in-house foundries can better be spent on designing circuitry for cutting-edge mobile phones and other gadgets. So they are subcontracting the task of turning their ideas into actual application-specific integrated chips to companies that specialize in foundry work, which already sell silicon wafers and other products to others that make memory chips and microprocessors. That is the sweetest spot to be in the age of silicon-driven technology. "I see the foundry industry growing 22% per year on average over the next three years," says Gurinder Kalra of Morgan Stanley in Hong Kong. "We will see large companies close their [semiconductor] plants or sell them to the likes of TSMC and outsource more and more of their fabrication needs."
Analysts single out TSMC as the trend's key beneficiary. So far this year, it has bought two Motorola chip plants and 30% of Acer's semiconductor unit, gaining not only physical facilities but also chip-design libraries and more engineering talent. California-based research firm Dataquest says TSMC now accounts for 35% of the foundry business. (UMC is estimated to have 19% while CMS has 16%.) With seven fabrication plants, two more under construction in Taiwan and Singapore, and six in the planning stages, TSMC may increase its market share to 50% in three years. Profit last year topped $469 million on sales of $1.5 billion - the kind of fat margins enjoyed by the likes of Microsoft and Hong Kong property companies (before the Asian Crisis upended their bottom lines). " TSMC is a jewel of a company," says Chris Hsieh, electronics analyst with Nomura Securities in Taipei.
The man behind the phenomenon is genial, pipe-smoking Morris Chang, a former executive at Texas Instruments. He was lured to Taipei to head TSMC, a venture founded in 1987 by the Taiwan government and Philips Electronics of the Netherlands. At the outset, Chang decided to focus on foundry work, not on dynamic random-access memory (DRAM) chips like Korea's Samsung, Hyundai and LG Semicon. The DRAM makers made billions in the mid-1990s as demand for memory chips soared and supply - it takes years to put up a DRAM facility - got tight. But the entry of more and more players eventually led to a glut and DRAM prices plunged. Memory chips are now regarded almost as low-margin commodities.
TSMC is riding a more sustainable wave. As the first pure integrated-circuit foundry company (it supplies DRAM makers and other chip companies with silicon wafers and also manufactures semiconductors), TSMC forged relationships with electronics giants like Motorola and Acer. Its charter prohibits it from designing its own brand-name microchips, so the majors know there is no conflict of interest. Those ties and the company's track record are placing TSMC first in line as the outsourcing trend takes off. At the same time, TSMC is picking up business from the proliferation of independent semiconductor houses that do not have their own fabrication facilities. And it supplies DRAM makers with wafers, though that is only a small part of its sales.
Second-ranked UMC is also well-placed to perform strongly. Partly owned by Taiwan's Ministry of Economic Affairs, the company is merging with four of its subsidiaries and affiliates to cut costs and boost earnings. (It made $131.8 million on sales of $551.2 million last year.) UMC intends to wrest the leadership of the foundry business from TSMC within a few years. Company officials talk about UMC evolving into a one-stop shop where customers can bring a concept, see it translated into an actual product and brought to market in far less time than they thought possible. UMC will help design the microchip, purchase proprietary information to enhance functionality if needed, come up with a prototype and manufacture the approved product.
The third player, Chartered Semiconductor, has no plans to become the world No. 1 - it cannot match the $15-billion expansion plans of TSMC and UMC. In part because of its overdependence on memory chips, CSM lost $98.1 million on sales of $333 million last year. Its product mix is now biased toward the data and telecommunications industries. At least CSM has no other major rivals. New entrants need at least $2 billion to build a wafer-fabrication plant, which requires extremely clean rooms (dust and other microparticles can render a chip useless) and ultra-sophisticated photo-optical and other equipment.
They also need skilled knowledge workers. "The foundry industry isn't like DRAMs," says analyst Kalra. "It isn't about economies of scale from huge capacity increases. It is about service and expertise." Don't forget relationships. The Motorolas of the world must trust their sub-contractors, and that means foundries that have proved themselves capable of manufacturing flawless products on time have the inside track. The Big Three have compiled a track record for reliability and timeliness. The Taiwan companies are especially known for their flexibility. "They are very nimble," says Nomura's Hsieh. "They are able to switch quickly as the demands of their customers change."
Where does that leave non-foundry players? Not necessarily out in the cold. The foundries cannot make, much less design, all the myriad logic chips, memory chips and microprocessors for existing and future electronic gadgets. That leaves a lot of opportunities for independent semiconductor firms in Malaysia, Indonesia and the Philippines, not to say Korea's DRAM manufacturers, which are moving on to more capacious (but tinier) 256-megabit memory chips. "The whole global semiconductor industry has begun its upturn," says Kalra. If they play their cards right, the Big Three will not only remain big but also just that - three. |
