Will growth bear profits?
Semiconductor buyers interested in knowing which suppliers will be successful in the future should take the advice of Deep Throat of Watergate notoriety and 'follow the money.'
By Jim Carbone -- Purchasing, 6/19/2003 2:00:00 AM
It's often said that it takes money to make money. That adage is true in the semiconductor industry where the companies that invest the most in new process technology, equipment and fabs are often the leading-edge companies with the highest sales revenues.
Chip companies that have the highest capital expenditures (capex) often are the most successful. They have the technology and the parts ability that electronics buyers need.
Case in point: Intel. The company is perennially the top semiconductor company in revenue as well as the top capital spender among semiconductor suppliers. It is able to develop newer faster microprocessors (MPU) and manufacture them efficiently, dominating the MPU market.
Intel had sales of $24 billion in 2002 and this year will invest $3.7 billion in new capacity, process technology and equipment and fabs. The company has about 80% of the microprocessor market and is the top flash memory supplier as well.
"Sales revenue and capex go hand in hand," says Brian Matas, an analyst with market researcher IC Insights. "The companies that have the greatest sales typically will invest the most as a percentage of revenue. "It's often about 18-23%. The ones leading the revenue list will lead the capital spending list as well," says Matas.
One company that has moved up the rankings of the top 25 semiconductor companies is Samsung. In 1998 Samsung was the seventh largest semiconductor company in revenue. Today it is number two with revenues of $8.7 billion.
"Samsung moved up from three in 2001 to two last year with a revenue increase of 33%," says Matas. "Growth was due to DRAMs. In the last half of the year they pushed the movement to double data rate (DDR) DRAMs, which at the time were in limited supply," he says. Samsung pushed DDR DRAM and at the time could command a 20% premium over standard synchronous DRAMs.
"That helped boost their sales. They also moved down to the smaller process geometries there and made some extra margin," says Matas.
Nand flash sales also contributed to Samsung's growth. Samsung boosted its nand flash business 267% to $1.1 billion in 2002. Nand flash is used for data storage and found in digital cameras and MP3 players among other consumer electronics equipment.
Good to be king
"Samsung is the king of the nand flash memory business now," says Matas. "They are following the Japanese model of the '80s. They are vertically integrated, they have product in demand from outside sources as well as internal business."
Matas says the success of Samsung was due in part to its aggressive investment in facilities and equipment. It was able to manufacture leading-edge products to meet market demand.
"Samsung is continuing to invest in its DRAM and flash memory business," says Matas. "The company is putting pressure on its competitors. It is using a lot of its money to outfit 300 mm fabs. Samsung had one [300 mm fab] come up late in 2002," he says.
"We do invest heavily and consistently," says Tom Quinn, vice president of U.S. sales and marketing for memory at Samsung. "Semiconductor manufacturing is capital intensive. Memory is getting the bulk of the investment. We are investing in capacity in the factory, in equipment and in technology. We are constantly shrinking the process and that takes a fair amount of capex."
For 2003, Samsung plans to invest a total of $3.4 billion in capital according to IC Insights. Over half will be for its memory IC operation including DRAM, SRAM and flash memory. The rest will be invested it its thin-film transistor liquid-crystal display operation and its system LSI chip business.
"We are migrating to 300 mm wafers," says Quinn. "It is all integrated because to go to 300 mm you have to have different equipment. You have to have a whole new factory," says Quinn.
Quinn says Samsung's aggressive capital spending has helped Samsung boost sales revenue and profitability during the downturn. The company was ahead of the curve on DDR DRAMs when computer manufacturers decided to transition to DDR from synchronous. Samsung was also able to lead the way with shipments of 256 megabit chips which are now the most in demand DRAM density.
"We are taking the profits and putting the money back into technology during a down cycle," says Quinn. He says today's investment will bear fruit in 18 months. By investing in 300 mm fabs and 90-nanometer process technology during a down cycle, Samsung will be able to maximize profits during the up cycle which many analysts forecast will happen in second half 2004 and in 2005.
More companies spend
Of course Samsung isn't the only chip company that is aggressive with capital spending. Nanya, Toshiba, Infineon and Grace will boost capital spending in 2003, according to IC Insights.
Capital spending among the top 25 semiconductor suppliers will increase 15% in 2003 to $31.9 billion. Some companies will add new chip making equipment. Some will build 300mm fabs and transition to 90-nanometer process technology. Ninety-nanometer technology is the next generation chip making process, which will provide manufacturers with more usable chips per wafer. Other chip companies will move from 0.25 micron to 0.18 or 0.13. By upgrading process technology companies can produce more chips per wafer, reducing cost.
But not all companies need to move to 300 mm or 90-nanometer technology, notes Matas. "The biggest companies, whether they are foundries like TSMC (Taiwan Semiconductor Manufacturing Co.) or UMC or some of the leading suppliers like Intel, Samsung or Infineon, are the ones spending on 300 mm and 90-nanometer or 0.13 micron," he says.
Others companies are not moving to 300 mm, but are upgrading their existing main line operations. There are companies that can't get up to that 90-nanometer level, and don't need to. They are content increasing the number of wafer starts on a 200 mm line or migrating process technology to 0.18 micron from 0.25 micron.
"For advanced processing they will farm it out to TSMC or UMC or IBM to have them take care of it," says Matas.
One company that is investing heavily is Toshiba. Toshiba will increase capital spending by 51% to $840 million. "Flash memory is Toshiba's driver. It has been one of their strengths over the last year," says Matas. "They started upgrading, making sure they are in place and in position to respond to any additional demand for flash this year. By most accounts, there will be significant demand for flash this year," says Matas.
Toshiba is building a 300 mm fab in Oita, Japan. Construction was to start in June and chip production is scheduled for the second half of 2004. The new fab will have capacity to produce 12,500 300 mm wafers per month. The new fab will make microprocessors for broadband networking operations.
Nanya, which makes memory ICs, will boost capital spending by 168% in 2003. "Nanya is putting in more advanced equipment," says Matas. "It is trying to take advantage of the sales it had and convert that into some investment to be prepared for the next upturn. Nanya has a joint 300 mm fab with Infineon," says Matas.
Infineon will boost spending 16% to $990 million and Grace will increase spending 25% to $750 million, according to IC Insights.
Spending is good
If total capital spending increases at 15% it will be good news for buyers. "[A] fifteen percent increase on capital spending is a healthy level although you'd like to see it at an even higher level if you are a buyer," says Matas. A 15% increase in capital spending will put pressure on prices.
The building of new fabs and migration to smaller process technologies puts pressure on pricing because a greater number of chips can be produced for the same process procedures. "Fifteen percent keeps the pressure on prices to stay low," says Matas.
In fact some semiconductor companies will not be increasing capital spending for that reason. They are afraid additional spending will cause further price erosion. Case-in-point: ST Microelectronics, which makes a variety of semiconductors including flash memory. ST Microelectronics' capital spending will be flat in 2003, according to IC Insights.
"Business may be increasing, but not enough to warrant new fabs, says Jean-Philippe Dauvin, chief economist for ST Microelectronics. "Many companies claim their fabs are 100% saturated with 0.13 or 0.15 micron production lines. But many of the fabs are half full. If there is any demand increase, which I believe will happen in the second quarter and continue for the rest of the year, it won't be sufficiently strong to warrant a new fab," says Dauvin.
He doesn't agree with forecasts that call for a 5-15% increase in capital spending by chip companies. In fact he says capital spending will fall 5%.
Dauvin says many companies don't want to increase capital spending because prices are weak. Any additional capacity will further weaken tags.
"We could have another two quarters of a very low capex," says Dauvin. That may be good for semiconductor companies, but not good for buyers.
Low capex will lessen price pressure and improve margins for semiconductor companies. "We need two or three quarters of zero price erosion or 5% increases in order to have some capex increase," says Dauvin. He says as of the end of March the 12-month rolling average for prices was minus 6%.
"We ended 2002 at minus 14% so there is improvement," says Dauvin. "I am sure at the end of the year, the trend for average selling prices will be positive, maybe in the 5-6 % range. At that point semiconductor companies will start to think about investing," he says.
However, IC Insights says that if the economy gets rolling and demand for electronics equipment rises, capital spending will flow right away and will be invested in new chip making equipment, fabs and next generation technology.
Ranking changes
Based on capital investment plans and on business and technology trends buyers can expect some change in semiconductor company rankings next year, although the top chip companies, Intel, Samsung, Texas Instruments and ST Microelectronics will likely maintain their respective positions.
TSMC could move a place or two. The foundry company moved up from 10th in 2001 to 6th in 2002. "Foundry business was going great last year," says Matas. "In 2002, there were few companies who had the capability to run small process geometries in volume. A lot of companies were struggling to cut cost or overhead by shutting fabs. They would outsource that to TSMC or UMC." The same scenario is likely for this year as well.
IBM Microelectronics, which moved down from number eight to 11 in the semiconductor sales rankings could bounce back on the strength of its foundry business.
"IBM's application specific integrated circuit (ASIC) business cooled off last year and they shifted their chip business to foundry," says Matas. "They opened up a new 300 mm fab. Rather than fill that with their own production, IBM is looking at capturing the leading edge foundry business."
AMD dropped in rankings from ninth in 2001 to 15th in 2002 and may drop again in 2003. "AMD has good production, a good portfolio, a wide variety of flash memory built in different packages, voltages and configurations," says Matas. "AMD wasn't able to generate as much in the flash business as some of the other leading manufacturers," he says.
As for processors, AMD's business is in desktop PCs, but that's not where the most lucrative part of the microprocessors business is right now. "It is more the laptop, the portable-type environment or the workstation and server markets. These are the markets they are trying to address this year," says Matas.
Toshiba may move up a spot or two on the strength of its nand flash business, despite a nand flash price war.
With its investment in new 300 mm fabs and 90-nanometer process technology Samsung will be in the catbird seat in 2003. It will likely distance itself from ST Microelectronics and Texas Instruments, but won't be able to approach Intel in sales.
"We feel we are pulling away and have a strong second position in the industry and we are looking for more growth this year," says Quinn of Samsung.
"The growth potential for Samsung is very good. We think the emerging applications that are consuming silicon are nicely tuned to memory technology," he says. "Logic technology used to be the core element of the PC. We think if you look at all the mobile products, it is memory technology that has more play leverage growth. We are bullish," says Quinn.
Samsung's advantage is that it can run flash technology on advanced DRAM fab lines, which means the chip maker has flexibility. "If the price of flash is favorable it can build more flash. If too much product is put into the market, softening prices and it looks like revenue per wafer is not as attractive as DRAM, we can quickly use that capacity to build a different technology," he says.
Top 25 semiconductor companies
($ millions worldwide)
| Rank | Company | Region | 2002 ($M) |
| Source: IC Insights |
|||
| 1 | Intel | U.S. | 24,084 |
| 2 | Samsung | South Korea | 8,720 |
| 3 | TI | U.S. | 6,650 |
| 4 | ST | Europe | 5,605 |
| 5 | Toshiba | Japan | 4,823 |
| 6 | TSMC | Taiwan | 4,655 |
| 7 | Infineon | Europe | 4,481 |
| 8 | NEC | Japan | 4,450 |
| 9 | Motorola | U.S. | 4,335 |
| 10 | Hitachi | Japan | 3,575 |
| 11 | IBM | U.S. | 3,500 |
| 12 | Philips | Europe | 3,482 |
| 13 | Fujitsu | Japan | 2,980 |
| 14 | Micron | U.S. | 2,845 |
| 15 | AMD | Japan | 2,697 |
| 16 | Mitsubishi | Japan | 2,640 |
| 17 | Hynix | South Korea | 2,544 |
| 18 | Sony | Japan | 2,520 |
| 19 | Matsushita | Japan | 2,320 |
| 20 | UMC | Taiwan | 1,950 |
| 21 | Qualcom | U.S. | 1,942 |
| 22 | Nvida | U.S. | 1,915 |
| 23 | ADI | U.S. | 1,797 |
| 24 | Sanyo | Japan | 1,775 |
| 25 | Agere | U.S. | 1,700 |
Top 25 capital spenders
(2002 spending, $ millions, worldwide)
| Rank | Company | Region | 2002 ($M) |
| Source: IC Insights |
|||
| 1 | Intel | U.S. | 4,700 |
| 2 | Samsung | South Korea | 1,870 |
| 3 | TSMC | Taiwan | 1,574 |
| 4 | IBM | U.S. | 1,050 |
| 5 | ST | Europe | 996 |
| 6 | Infineon | Europe | 850 |
| 7 | Micron | U.S. | 850 |
| 8 | UMC Group | Taiwan | 807 |
| 9 | TI | U.S. | 802 |
| 10 | SMIC | China | 800 |
| 11 | AMD | U.S. | 750 |
| 12 | Elpida | Japan | 700 |
| 13 | Matsushita | Japan | 675 |
| 14 | Grace | China | 600 |
| 15 | Toshiba | Japan | 555 |
| 16 | Sony | Japan | 500 |
| 17 | Philips | Europe | 416 |
| 18 | Renasas* | Japan | 395 |
| 19 | Hynix | South Korea | 350 |
| 20 | Rohm | Japan | 350 |
| 21 | Fujitsu | Japan | 340 |
| 22 | Sharp | Japan | 340 |
| 23 | NEC | Japan | 325 |
| 24 | Nanya | Taiwan | 250 |
| 25 | Dongbu/Anam** | South Korea | 100 |
| Top 25 Total ($ Billions) | 20,945 | ||
| *Hitachi and Mitsubishi spending combined for 2002. | |||
| **Dongbu and Anam spending combined for 2002. | |||
Top microprocessor suppliers
($ millions)
| Rank | Company | 2001 sales | 2002 sales |
| Source IC Insights |
|||
| 1. | Intel | 18,575 | 19,000 |
| 2. | AMD | 2,419 | 1,705 |
| 3. | Motorola | 700 | 809 |
| 4. | IBM | 385 | 405 |
Top DSP suppliers
($ millions)
| Rank | Company | 2001 sales | 2002 sales |
| Source: IC Insights |
|||
| 1. | Texas Instruments | 1,750 | 2,150 |
| 2. | Motorola | 675 | 684 |
| 3. | Agere Systems | 685 | 650 |
| 4. | Analog Devices | 252 | 210 |
Top analog chip suppliers
($ millions)
| Rank | Company | 2001 sales | 2002 sales |
| Source: IC Insights |
|||
| 1. | ST Micro. | 3,212 | 3,360 |
| 2. | Texas Inst. | 3,100 | 3,100 |
| 3. | Infineon | 1,698 | 1,791 |
| 4. | Analog Devices | 1,567 | 1,587 |
| 5. | Philips | 1,572 | 1,438 |
Top fabless suppliers
($ millions)
| Rank | Company | 2001 | 2002 |
| Source: IC Insights |
|||
| 1. | Qualcom | 1,395 | 1,942 |
| 2. | Nvida | 1,275 | 1,915 |
| 3. | Xilinx | 1,149 | 1,125 |
| 4. | Broadcom | 962 | 1,083 |
| 5. | MediaTek | 447 | 854 |
IC industry poised for recovery in 2010
07/15/2009Why capital spending matters
07/13/2005IC companies cut capital spending
06/10/2009Chip growth reaches new heights
08/16/2006
Now Playing: 
