Slow business doesn't discourage Asia's silicon foundries
08/01/2001
Pieter "Pete" Burggraaf, Senior Technical Editor
overview
Just over a year ago, foundry manufacturing epitomized the leading edge of semiconductor manufacturing technology, and Asia was the hotbed of foundry building activity. Has this changed? Leading executives comment on how the industry downturn in 2001 has affected Asia's silicon wafer foundries, particularly their business levels and building plans.
In light of the semiconductor industry's current downturn, what has happened to the rosy view for Asian silicon-wafer foundries that we saw less than 12 months ago?
Our last "foundry watch" report [1] characterized foundry business as struggling to meet a 40% increase in wafer demand in 2000, with capacity booked for several years out.
All this was driving new fab construction at a blistering pace, particularly in Asia. On the technology side, Asian silicon foundries entered the new millennium leading the industry to 300mm wafers (Fig. 1) and 130nm plus device rules.
Just last October, Semico Research (Phoenix, AZ) was predicting that the worldwide silicon-wafer foundry business, which is dominated by Asian manufacturers, would grow 17% in 2001. At the time, it looked as if silicon wafer foundry demand would grow from 12.7 million 200mm equivalent wafers in 2000 to 15.1 million wafers this year.
 Figure 1. A Trecenti fab worker displays "the foundry industry's first 300mm wafer." (Source: UMC) |
Well, there is no question that silicon-wafer foundry business is slumping along with the negative-growth that the rest of the industry is seeing. Several times since late last year, Semico Research has adjusted its forecast for worldwide foundry services. Part of the reason is an oversupply of wafer fabrication capacity that has caused some large IC manufacturers to pull production back into their own fabs (Fig. 2). Semico's most recent forecast data show that the demand for processed wafers from third-party manufacturers (i.e., foundries) will decline from 12.7 million 200mm equivalent wafers in 2000 to 12.0 million wafers in 2001, 14% of the industry's total wafer demand (Table 1 on p. S12). This decline is also documented by Semi's tracking of virgin silicon-wafer shipments (see "Silicon wafer shipments in early 2001").
The current forecast by Semico of a 5.7% downturn is the worst slump ever in the short history of the silicon-wafer foundry business (see Table 1). Previously, this sector suffered its biggest setback from 1997 to 1998 when business fell 0.5%.
Some bright spots
There is one bright spot: According to Semico's chip foundry analyst Joanne Itow, in the current downturn, foundry-processed wafer prices have not taken a beating, as many in the industry originally feared. "Wafer prices have not changed that much this year compared to 1997 and 1998. During the last foundry slump, leading manufacturers engaged in a price war to maintain their market shares and keep their plants full."
The future is also bright, according to Semico. Itow projects silicon-wafer foundry suppliers may see a recovery in the second half of 2001, with the likelihood of a boom period starting in 2002. "Next year, the silicon foundry business will surge in terms of unit volumes," she says. "Right now numbers are negative for 2001, but I believe the market will hit bottom in the third quarter of this year. Things will look better in the fourth quarter, and 2002 is going to be a good year." According to Semico Research, the coming silicon-wafer foundry rebound will be driven by both the communications and PC systems. "We are bullish about the communications market," says Itow.
The foundry view?
To get a firsthand picture, Solid State Technology recently asked a few key silicon foundry executives at Asia's top three silicon-wafer foundries how the current downturn in the semiconductor industry has affected silicon foundries, particularly in regard to business levels and building plans. (We also looked at some foundry upstarts. For more information, see "Silterra Malaysia" on p. S12 and "Korean foundry Dongbu — 300mm Fab 2" on p. S16.)
In Singapore
Obviously, the sudden and surprisingly sharp downturn, particularly in the communications market, has taken a major toll on the foundry industry in terms of sales and fab-utilization rates. Singapore's Chartered Semiconductor Manufacturing, for example, has reported that net revenues were US$206.7 million in 1Q01, down 13.3% compared to US$238.4 million in 1Q00. This significant drop in revenues was due to lower shipments to the communications and consumer end markets. Compared to 4Q2000, Chartered's net revenues were down 35.1% as a result of the broad weakness experienced this quarter in all market segments.
Chartered's wafer shipments in 1Q01 were 166,400 200mm equivalents, a decrease of 20.8% compared to 210,100 wafers in 1Q00 and a decrease of 31.9% compared to 244,300 wafers shipped in 4Q00.
 |
Figure 2. 2001 distribution of foundry silicon-wafer processing. (source: Semico Research)
"Chartered has been heavily weighted in the communications industry," says Itow. "In comparison, Taiwan Semiconductor Manufacturing Co. Ltd. (TSMC) and United Microelectronics Corp. (UMC) have a broader base of customers.
Reflecting the future, in the same period, Chartered's R&D expenses increased by US$3.9 million from the year-ago quarter as the company stepped up investments in next-generation technologies and modules in support of its strategy to provide a full suite of processes necessary for enabling system-level integration.
 Figure 3. A UMC fab worker inspects a 200mm wafer. |
Barry Waite, president and CEO of Chartered Semiconductor, said that "there is growing evidence that the semiconductor industry is facing one of its most challenging years. Weakening economic conditions are prolonging the time required for our customers to bring inventory levels back into balance, particularly in the communications segment. However, at times like this, it is possible to lose sight of the exciting long-term outlook for the semiconductor industry, and more particularly the foundry services industry. There is nothing that has happened in the last year that would make us waver from our belief that the foundry industry can grow at a compounded annual rate of 30% or more during the current decade, driven by the compelling economics of outsourcing and the rapid growth rate of fabless semiconductor companies."
At Chartered, the company's 300mm Fab 7 project (see Table 2 "Solid State Technology's 300mm transition and fab watch") entered the equipment selection phase in mid-2001 and the company is maintaining its plans to increase capital spending by 10% in 2001. Chartered has invested US$3.5 billion in the new fab and by 2004 hopes to reach 30,000 wafers/month.
Kevin Meyer, VP of world business development at Chartered told Solid State Technology [2] that the company is focusing on system-level technology, noting that as the PC becomes a web appliance, the focus will be less on megahertz and more on connectivity. "Technology is well focused on opportunities," he said, "and we are developing working relationships with customers on process differentiation. It's not one size fits all."
Meyer said that one of Chartered's major focuses is "reducing cycle times for prototypes." The company can produce a prototype in 30 days, including maskmaking. He noted that, as a foundry, Chartered is very flexible. The company takes only days to turn around a masking layer, making a six-layer wafer in approximately 12 days. Fab 7 will begin production with SRAMs and 180nm technology. Chartered hopes to have its first qualified 300mm wafer from the fab by mid-2002.
Meyer acknowledged that building a fab is a huge financial challenge. He commented that while the Singapore government has assisted with some of the cost, a major benefit to Chartered's financial wellbeing during the construction of Fab 7 was that its other fabs were already in production. Chartered produced nearly one million 200mm wafers in 2000. Its plan is to double production to two million by 2003. "Each of the Chartered fabs runs three technology nodes to maintain versatility. One challenge facing Chartered, according to Meyer, is expanding the company's customer base. He said that as the industry is moving toward 130nm technology, which is now a manufacturing reality, and with 300mm on the horizon, it is key to demonstrate to customers that they can spend money on IP rather than maintaining spending on process technology.
 |
The company has moved to joint development relationships with "partners who want to research with core technology." Those companies include Agere and Ericsson. The joint development agreement with Agere, Meyer said, was created in order to "take advantage of the best technologies." Meyer said that Agere has an understanding of advanced manufacturing and low-k dielectrics, which is a benefit to Chartered. Meyer said that acquiring market share is not part of Chartered's goals. Rather, the company feels that its investment, its roadmap, and its worldwide service plan are among the keys to its future.
In Taiwan
TSMC and UMC have also suffered in the six-month-old downturn. In April 2001, TSMC, which claims to be the world's largest silicon foundry supplier, reported that its fab utilization rate had fallen close to 50%, and it warned that the level could drop slightly lower in subsequent months. That represents a sequential drop from 70% in the March quarter and 100% in the fourth quarter of last year.
It was much the same story at UMC. The company posted net sales for 1Q01 of NT$23.59 billion, representing a 22.2% improvement from NT$19.31 billion for 1Q00. Quarter-over-quarter, however, net sales decreased by 25.9% from NT$31.85 billion in 4Q00. In 1Q01, UMC shipped 443,000 200mm equivalent wafers compared with 629,000 in 4Q2000 and 628,000 in 3Q2000, not including shipments at Nippon Foundry Inc., a dedicated foundry company in Japan that is owned by UMC.
Jim Ballingall, VP of worldwide marketing at UMC, told Solid State Technology that the company's revenues, which grew at roughly 20%/quarter sequentially from early 1999 through the end of 2000, rapidly subsided with a steep decline in semiconductor demand at the end of last year. "Our 1Q2001 revenues dropped 26% from 4Q2000," he said. We anticipate that 2Q2001 will drop by 30% relative to 1Q. Our expectation is that 2Q2001 will be the weakest quarter this year, though timing for a recovery to the demand levels of last year is not clear."
 |
Accordingly, UMC has cut back on 200mm expansion plans (Fig. 3), but the company remains committed to 300mm-expansion projects in Japan and Taiwan. "In fact, these factories are the first and second 300mm fabs in the world," claims Ballingall. "The first, in Japan, is operating now, and the second, in Taiwan, is scheduled for pilot production [mid-2001]. Both of these have capacities of 40,000 300mm wafers/month. We have broken ground on a third 300mm fab in Singapore (Fig. 4)."
According to Ballingall, UMC continues to take a long-term view of the semiconductor industry cycles. "We are committed to have the capacity for leading-edge technology our customers will require," he said.
Across the street at TSMC
Ed Ross, president of TSMC North America, told Solid State Technology, "Clearly, the semiconductor foundry business is suffering from substantially lower utilization levels compared to the same period a year ago." But, like its staunch competitor UMC, TSMC executives are optimistic that the IC industry cycle will begin to improve in the latter half of the year. "While overall growth may never return to the same overheated levels we witnessed in the latter half of 2000, TSMC will continue to be the industry leader in capacity, technology and services. By the end of this decade, the foundry industry as a whole could fulfill as much as half of the worldwide demand for IC wafers," said Ross.
 Figure 4. UMC's future 300mm fab in Singapore. (Source: UMC) |
Ross noted that at TSMC, engineers and product designers are increasingly targeting deep-submicron geometries to achieve the best performance and feature sets, and therefore the best competitive positions. "In 2001, more than half of our sales will involve products built on geometries at or below 0.25µm. TSMC leads the industry in terms of number of tapeouts to our 130nm process technology."
Again like UMC, TSMC remains committed to capacity expansions that would lead the foundry industry and perhaps gain a number-two ranking in the overall semiconductor industry. This growth is consistent with historic trends. For example, TSMC's capacity increased from 1.8 million 200mm equivalent wafers in 1999 to 3.4 million 200mm equivalent wafers in 2000. "In 2001, our capacity will increase 32%, to 4.4 million 200mm equivalent wafers. By 2005 we expect to have capacity for 11.4 million 200mm equivalent wafers," said Ross.
Ross added, "Beginning this year, all of our new production capacity will be in the form of 300mm manufacturing. By 2005, half of our manufacturing capacity will derive from 300mm lines. TSMC was the first dedicated foundry to deliver customer product on 300mm wafers and we continue to deliver the highest number of customer products from our 300mm lines. Through next year, we expect to deliver the most 300mm foundry wafers to dedicated foundry customers."
By the end of 2001, the 300mm pilot line in TSMC Fab 6 in Tainan, Taiwan, will produce about 6000 300mm wafers/month; this fab line is running process technologies down to 130nm. "TSMC Fab 12 is our first dedicated 300mm facility. It will begin producing its first customer wafers in 4Q2001, graduating to 30,000 wafers/month when fully ramped," Ross said. "TSMC Fab 14 is currently scheduled for equipment move-in later this year. However, this schedule may change if economic conditions in the global IC market do not appear to improve. We will make this determination at a later date."
References
- P. Burggraaf, "Silicon foundries in Asia: New capacity with 130nm technology," Asia Pacific: A supplement to Solid State Technology, November 2000, pp. S15-S22.
- As reported in WaferNews, Vol. 8-24, June 18, 2001.
Pieter "Pete" Burggraaf has more than 25 years of experience in the semiconductor industry, including work at Motorola, Siemens, and ASM. He can be reached at 875 S. Yucca Dr., Wickenburg, AZ 85390; ph 928/684-1265, fax 928/441-3139, [email protected].
___________________________________
Silterra Malaysia
Silterra Malaysia found itself ramping up just as the industry's downturn intensified, but was still able to generate significant revenue by the end of 1Q01, less than three months after completing installation of manufacturing equipment. Cy Hannon, president and CEO of Silterra, said, "Throughout the past year our operational success has been very impressive. We have met all major milestones as scheduled, and have produced and shipped revenue wafers one month ahead of schedule."
Hannon added, "As we continue to ramp up production volume, we will further demonstrate our ability to deliver quality products on time. Our customers have been giving us tremendous support. A recent purchase order of approximately $80 million for delivery in 2001 shows that there is a lot of confidence in our capabilities. Our current operational capacity is 12,000 wafers/month and we are continuing our ramp to 30,000/month on schedule."
Silterra produced working silicon in the first complete wafer lot it started and has delivered working products based on its 0.25µm and 0.22µm process technologies.
"We have made significant progress with some of the top semiconductor companies in the world," said Steve Della Rocchetta, executive VP of sales and marketing at Silterra. "We recognized that customers want additional options for their foundry services, and Silterra, with solid financial backing and demonstrated ability to execute, is looking forward to making an impact on the semiconductor industry."
Silterra is offering its customers process technologies down to 180nm. Volume production has begun in its facility, which will eventually be 130nm-capable. Plans are to reach full capacity with 30,000 200mm wafers/month by 2002.
___________________________________
Silicon wafer shipments in early 2001
Statistics from Semiconductor Equipment and Materials International (Semi, San Jose, CA) show that worldwide silicon wafer-area shipments declined by 14% during 1Q01 compared to 4Q00 data (see table). In its analysis of the silicon wafer industry, Semi determined that silicon wafer area shipments totaled 1.25 billion in2 in 4Q01, down from the 1.45 billion in2 shipped during 4Q00 and down 2% from 1Q00.
 |
Issuing the summary, Stanley T. Myers, president and CEO of Semi, said, "Wafer manufacturers are experiencing the effects of the chip inventory overhang impacting their customers and the general slowing in the overall macroeconomic conditions. 2001 will prove to be a challenging year for the electronic material suppliers, particularly if there is a reduction in semiconductor unit output."
___________________________________
Korean foundry Dongbu — 300mm Fab 2
With the April announcement that it had begun volume production of 200mm CMOS wafers, Dongbu Electronics Co., Korea's first pure-play foundry, took a leap forward on an ambitious business plan. Dongbu looks to ramp to 20,000 wafers/month this year and then — hopefully adding $310 million in second-round funding to the $440 million already raised — expects to jump to 45,000/month in 2002. Plans are to be profitable by late next year, with an IPO on the NASDAQ eyed for the 2002/2003 timeframe.
Key to the success so far has been a partnership with Japanese giant Toshiba, which, in addition to providing $50 million in funding, has given Dongbu access to process technology for standard logic and mixed signal (0.25µm, 180nm and 130nm), access to its digital cell library of analog, logic and I/O, technical assistance and training, and a TSMC lookalike product derived from TSMC processes.
"We like to think this is a science, but there is still a lot of art to it," said Peter Hillen, Dongbu's executive VP of worldwide sales. "Having somebody who has already developed a process and has it running is invaluable to speeding our time to marketplace."
Toshiba, for its part, gets access to a foundry that works on its own processes, with technicians it helped train. However, in an effort to keep Dongbu truly open to the global marketplace, Toshiba is capped at being able to occupy only 25% of Dongbu's production efforts. Part of what will lead to Dongbu's success is simple geography, suggested Hillen. "The fact that we're just in a different spot, both seismically and politically, will cause a lot of people to consider Dongbu as an alternative source to Taiwan fabs," he said.
While Dongbu's first three customers include Toshiba, an American fabless chip-maker, and a Korean fabless firm, an 80-plus-strong contingent of Korean fabless companies will help Dongbu along, said COO Wesley Min. "In Korea, there is no pure foundry company. Hynix is trying to do some foundry work, but its business strategy is totally different from ours. Hynix wants to leverage the fab of memory, and they can do that at 0.35µm," said Min. "Our strategy is going to start at 0.25µm and will go to 180nm and then 130nm. We're trying for the most advanced technology. We can help each other. Customers can come to Korea for one-stop shopping; it's only a 15 minute drive to Hynix."
Speaking of Hynix, Min believes the Korean government, which has drawn international heat for its support of the foundering company, is stuck between a rock and a hard place. "If Hynix goes bankrupt, it will be a really big impact on the economy," said Min. "Hynix is a hot potato for the government. I don't think the government will help a lot; it can try to influence banks." In addition to addressing current need, Dongbu is also set to address future capacity demands. The 622,000 square-foot Class-1 facility sits on a 250-acre site, with room to support a total of seven plants.
Ground clearing has already started for Fab 2, which will be a 300mm plant, possibly opening in 2003 or early 2004. While that timeframe may seem a bit extended for 300mm, with several pilot lines already running around the world, Min said he still sees existing problems with the new diameter such as wafer warpage and stepper alignment issues. That problem, he said, should be resolved within a year or so of 300mm becoming more standard. "I don't want to be the guinea pig," he said.