Last year was a miserable year for the Top 25 semiconductor suppliers. On average, revenue declined 33% as the industry was plagued by high inventory levels caused by declining demand for end equipment including networking gear, computers and cell phones.

Some chip companies were hurt more than others, according to market researcher In-Stat MDR, Scottsdale, Ariz. Memory IC makers took it on the chin as DRAM sales fell from about $29 billion in 2000 to just $11 billion in 2001. Samsung, the leading DRAM manufacturer, saw its sales drop 46%. Hynix and Micron's revenue fell 62% and Toshiba's sales fell 41%.

The steep sales declines affected the rankings of the Top 25 and the ability for semiconductor companies to invest in new fabs and new equipment, which has many electronics supply managers worried about shortages for next year.

Intel and Toshiba held on to their respective one and two positions, but NEC, which was third last year, fell to sixth. Samsung, which was fourth in 2000, dropped to fifth. Hynix fell from 10th to 18th and Micron fell from 11th to 19th.

While all semiconductor companies in the Top 25 suffered declining sales in 2001, some moved up in the ranking. STMicroelectronics moved from sixth to third despite having a 19.5% sales decline. Texas Instruments (TI) moved from fifth to fourth although sales dropped 33.7%. Philips improved from 12th to 10th despite a 27.4% sales decline.

STMicroelectronics has a diverse product mix, making parts such as microcontrollers, linear devices, discretes and flash memory for various industries. The same is true with Philips while TI has a dominant position in digital signal processors (DSPs) and supplies analog ICs and logic chips which are used across many industries.

"STMicroelectronics had the right product mix," says Steve Cullen, director of semiconductor research for In-Stat MDR. "A lot of other suppliers were heavy into communications which got hammered last year," he says. "One of the brighter spots last year was consumer and ST was nicely positioned. They probably have a better balanced mix than most of their competitors and they have quite a bit in consumer," says Cullen.

He says TI's revenue was down because the company supplies a lot of chips to the wireless industry, which was hit hard last year. "But digital signal processors and analog ICs are used in consumer electronics equipment and that helped TI. What also helped them is that they are not in DRAMS," says Cullen.

The decline of the DRAM market has affected the top semiconductor company rankings. "If you go back to the mid 1990s, DRAM suppliers were a lot higher in the rankings and so were Japanese semiconductor companies," says Cullen. Many Japanese companies such as Toshiba, NEC and Hitachi concentrate on DRAM and the DRAM market grew to $40 billion in 1995. In 1995, seven of the top 10 companies manufactured DRAM. In 2001 only three made DRAM. Both TI and Motorola got out of DRAM business while Hitachi and NEC spun off their DRAM businesses to form Elpida.

In the next several years, there will likely be more changes in the top 10 semiconductor company rankings. For one thing, while the DRAM market will grow in the next several years, it probably will not reach $40 billion again, so it's not likely that more DRAM manufacturers will make the top 10. In fact, the computer industry, which uses a lot of DRAM, may not be as important a driver to the semiconductor industry as in the past. Consider: In 1995, the computer industry consumed 56% of all semiconductors produced. In 2001, the figure slipped to 48% and a further decline is likely in coming years.

"We will see a shift in the rankings," says Bill McClean, president of market researcher IC Insights. " TSMC (Taiwan Semiconductor Manufacturing Co.) will break into the top 10 in the next few years. The company is signing so many agreements. It has big agreements with TI as it outsources more products," he says. TSMC is a semiconductor foundry.

"IBM will likely be in the top 10 because of its involvement in the communications areas. IBM is producing more system on a chip semiconductors than DRAMS," says McClean. He adds while communications equipment manufacturers are struggling, the market is expected to come back in the next year or two and so will its need for semiconductors and other components.

"It doesn't change rapidly. We will see an evolution. Foundries, system-on-a-chip companies will do well. DSP and high-end analog ICs, mixed signal devices will all be in demand. DRAM will represent less of the total IC market," says McClean.

While producing the right product mix will be important for companies to crack the Top 25, so too will capital investment. Some analysts argue that the semiconductor companies that will do the best in the next upturn will be those that have invested the most during the current downturn.

McClean says this year chip companies will invest about $30 billion in capital equipment. Unfortunately that will be about 20% less than the previous year and may not be enough to prevent shortages from occurring next year if demand for end equipment picks up as it is expected to do.

"Capital spending is focused on upgrading existing facilities to the finer feature sizes and increasing 300 mm wafer plans," says McClean. "The spending is very focused."

He says the amount of spending is justified based on last year's sales and anticipated demand for this year "but may not meet the business levels for 2003."

"The killer in this industry is the lag from the time you start putting the shovel to the ground for a new fab to the time when wafers come out. It is usually between 18 months and two years," says McClean. "We are at the bottom of the cycle and just heading upward. Now is when you need to spend for 2003 and 2004. However, sales revenues for chip companies are still weak, so few are making investments in new fabs. This means when business does come back in 2003 and 2004, there is not likely to be enough capacity.

"Semiconductor companies will wait until they are sure demand is for real," says McClean. In the first quarter, sales increased 6% from the fourth quarter, according to the Semiconductor Industry Association. "But if you take out DRAMS, the IC market declined 1% in the first quarter compared to the fourth," says McClean.

"We have an uneven recovery at this point. Not all of the products are taking off or showing signs of recovery," he says. That means semiconductor companies will be slow to build new fabs and add capacity. "It takes a lot of guts for an IC maker to say we are going to increase capital spending based on what we expect for next year," says McClean.

McClean says the lack of investment means shortages are likely next year when demand picks up. "IC manufacturers will be in catch-up mode, but we may not have long-term shortages of any part. We will have a lot of shortages here and there and they may last only a quarter or two. Manufacturers will see the opportunity and invest, but they will be acting not in an anticipation mode, but in a reaction mode. They will catch up in time for the next downturn," he quips.

As of early second quarter 2002, only a handful of semiconductor companies were increasing capital spending over 2001. Some memory IC companies were buoyed by first quarter sales and are considering increasing capital spending. Case in point: Samsung Semiconductor, the leading DRAM and SRAM manufacturer, says it is considering boosting spending another 50%. Earlier, it said it would invest about $2.2 billion. Any additional spending would be for memory chips and liquid crystal display production lines.

"AMD is talking about going to $800 million from $680 million in 2001," says McClean. While that sounds like a lot, Intel, AMD's rival, is spending $5.5 billion, down from $7.3 billion in 2001, he says.

Another company boosting capital spending is foundry TSMC. TSMC will increase spending from $2.2 billion to $2.6 billion. "If you look back at 1998, in that downturn, the first companies to turn on spending were the foundries," says McClean. " TSMC, UMC and Charter have increased their capital spending from what they originally planned. So we have the three main foundries that have bumped up their capital spending."

JP Dauvin, STMicroelectronics' chief economist, estimates that capital spending will be about $28 billion. "It is not sufficient to create capacity. You need 10% of your total sales if you want to maintain existing operations," he says.

He says the problem will be for leading edge chips building on 0.18, 0.15 and 0.13 micron process technology. Only about 40% of forward capacity will be for these leading-edge process technologies, but demand for chips on those technologies is expected to rise significantly, according to Dauvin.

"If we are decreasing capital expenditures and 40% of capacity is for the line width that is most in demand, we should expect something to happen next year in terms of capacity tightness," says Dauvin.

With the lack of investment, capacity utilization rates will be a problem next year. "Today there is sufficient capacity everywhere. Capacity utilization rates for discrete and lineardevices are below 70%. Next year they will be 85% and discretes and linears will be tight in the second part of the year," says Dauvin.

"My expectation is the memory business could be at a 90-95% capacity utilization rate next year. Fabs are considered to be running at full capacity in the 85-87% range. When you are at 90-95% percent you are in an overheating situation and then you are short capacity," he says.

Dauvin says despite the lack of investment this year, buyers can expect there will be overcapacity through the first half. In the second half, capacity will start coming in line with demand and in the first half of 2003 supply will tighten and some allocations are likely. New capacity won't come on line until the second half of 2003, he says.

That has many purchasing and supply management executives worried. In fact, at a recent meeting of PURCHASING Magazine's editorial advisory board, some purchasing leaders expressed concerns over the lack of investment by electronics companies during the downturn. While purchasing managers aren't pushing the panic button, many figure that leadtimes will lengthen at the very least and, at worst, there could be some serious shortages next year.

John Paterson, chief procurement officer for IBM, echoes Dauvin's observations and says he is concerned about leading edge parts.

"I think the limited investment will have an impact on new advanced technologies. I think one or two companies will keep pushing forward, but overall there will be inadequate supply. The problem will be felt throughout the supply chain," he says.

Paterson adds that there may not be a solution to the problem, but there are steps that can be taken to minimize the damage.

"The actions that need to be taken in this type of environment are improved forecasting by OEMs and tighter collaboration with suppliers," he says. He adds that OEMs will also have to design to the supply chain instead of to new technologies whenever possible. He says it is a good idea for OEMs to have as many qualified suppliers for key commodities as is practical.

The lack of investment by semiconductor companies won't just affect OEMs. Electronics manufacturing services (EMS) providers are on the front lines of components shortages when they occur.

However, after the long buyers' market, it's hard for some EMS buyers to see a shortage coming anytime soon.

Andrew Gort, corporate vice president global supply chain management for EMS company Celestica, says he doesn't foresee any shortages in the near future even though some suppliers have closed some factories. "We assume when demand comes back, we will see lengthening of leadtimes, but we haven't see any shortages just yet," he says.

He notes that suppliers have been able to shrink die smaller and therefore get higher yields from wafers. "Die get smaller all the time which means out of a wafer you get exponentially larger numbers of chips," says Gort.

He says if demand comes back gradually, suppliers will have the capability to incrementally put capacity into place.

"If demand picks up quickly, it will put pressure in the supply base. It depends on how gradual the increase in demand is and how far technology has moved to allow more chips to be produced from the same wafer," says Gort.

He says in 2000 there was unexpected demand for cell phones, which caused flash memory and tantalum capacitor shortages. "All of a sudden the market just exploded with hundreds of millions of cell phone shipments and PDAs (personal digital assistants) being shipped. I'm not sure if we'll see that exponential explosion of wireless devices again," says Gort.

Theresa Metty, vice president of supply management for Motorola's Personal Communications Sector says even with new capital spending and die shrinks by chip companies some buyers will get caught short.

"The odd thing about the semiconductor industry is as they phase out the old fabs, anyone that has product life cycles that are long can get caught in a trap where they need the old technology. It is a big problem for our infrastructure business where equipment is around a long time and you need to maintain it," she says.

She says ASICS and flash memory are major components for the cell phone business. "We aren't sensing any flash memory problems at this point.

Metty notes the current lack of investment by chip companies is part of the semiconductor cycle. "When prices are up and supply is down chip suppliers go out and invest. Then all that capacity comes on line in a time span of 12 months," says Metty. "There is a surplus in capacity at a time when chip companies need tight conditions and higher prices to get returns on their huge investments, but then prices go down because of the surplus capacity.

"It is wildly cyclical and completely predictable, yet we can't overcome it," Metty adds. "[Making semiconductors] is massively capital intensive. There is a fleeting moment when we are in perfect supply/demand balance and it passes in a nanosecond."