Semiconductor buyers often scrutinize a semiconductor supplier's capital spending plans when evaluating a chip company. After all, a company that does not invest in capacity and new process technology may not be able to meet the technology and volume requirements of an OEM or electronics manufacturing services (EMS) provider.

However, as more semiconductor companies outsource production to chip foundries, buyers also need to take a close look at which foundries their semiconductor suppliers use and what the foundries' capital spending plans are.

For both foundries and semiconductor suppliers, capital spending plans can be a good indication of the success of semiconductor companies. Companies that invest the most tend to be the leaders in their respective product areas, says Bill McClean, president of researcher IC Insights in Scottsdale, Ariz.

"Semiconductor suppliers that have been successful such as Samsung, Intel and lately Hynix and Micron have consistently invested," McClean tells Purchasing. Intel and Samsung are perennially the two largest semiconductor suppliers. In 2007, Intel will make about $5.5 billion in capital expenditures and Samsung will spend $6.7 billion, according to IC Insights.

Semiconductor companies' capital spending can impact semiconductor prices. When chip companies boost capital spending, they often do so to transition to different process technology and to larger size wafers or to build new fabs. All of which creates more supply and often lower prices.

DRAM suppliers are often among the biggest capital spenders. Last year major DRAM makers boosted capital spending and added more capacity in anticipation of greater demand because of Microsoft's new Vista operating system, which requires about twice the amount of memory as Windows XP. However, Vista demand did not immediately kick in as strongly as suppliers anticipated. Too much capacity was added and the increase in supply led to lower prices.

"DRAM prices in the second half of June were about one-third what they were in January," says Kevin Lee, vice president of memory marketing at Samsung Semiconductor in San Jose, Calif.

Arun Kamat, vice president of marketing for Hynix Semiconductor, also in San Jose, says prices fell although unit demand was strong. Supply increased because some of the leading suppliers moved to 70 nanometer (nm) process technology and 300mm wafers.

"It's not that memory suppliers are building fabs right and left," Kamat says. "There are new factories coming up particularly in Taiwan and some Korean companies are building in China. More importantly, when you get to finer geometries, your total output in terms of net die per wafer increases and therefore the unit count increases."

McClean says leading companies such as Intel, Samsung and Hynix maintain their leadership positions in the industry by investing in leading-edge processes and technologies. "These companies need to be leading edge, whether it is design and functionality or process technology or both," he says. "If they don't, they won't grow."

Investing in leading-edge technologies and processes is important for the leading microprocessor, memory and application specific integrated circuit (ASIC) suppliers. Those products are leading edge and built in high volumes so 300mm wafers and 65nm process technology makes sense. But for more mature lower-volume parts, such as standard logic, investment in 200mm wafers is acceptable.

McClean says while many buyers are looking for leading-edge products made on 300mm wafer and using 65nm process technology, other buyers source more lagging-edge parts. If a supplier moves to 300mm wafers and 65nm process technology, the supplier may no longer support mature products that can be made on 200mm wafer and .13 micron technology.

"Depending on what you are buying, you need to look at what the supplier's support is for lagging-edge parts," he says.

"A good supplier will make sure it is investing on an ongoing basis in cost reductions, whether it is back-end packaging and test or front end and die," O'Donnell says.

He adds that some suppliers mistakenly use market growth of a certain product as an indicator of when to double up production for that part. "Suppliers need to make intelligent decisions about when the right time to add capacity is and how much capacity is needed," he says.

The semiconductor industry has a history of boosting capacity during good times, O'Donnell points out, which often results in oversupply and falling prices. Buyers may relish when that happens, but when prices fall, suppliers cut back on capital investment and supply gets tight and prices increase.

While most major chip companies see the advantages of making large capital investments in their business, some chip companies are moving to a "fab-lite" model and are investing less in new fabs and using foundries especially during an upturn, says McClean. One such company is Texas Instruments in Dallas. In 2007, TI's capital spending outlays will amount to $900 million, down 29% from 2006, he says.

The fab lite trend could make purchasing's job a little more complicated. In addition to checking out the capacity and capital spending plans of semiconductor suppliers, buyers also need to carefully evaluate the foundries their chip suppliers use.

During a downturn, a fab lite chip supplier can halt its outsourcing to foundries and keep its own fabs running at a relatively high rate because it had not invested in new fab capacity, says McClean. "Suppliers can endure the downturn better and invest money in research and development," says McClean.

"TI was early using the strategy and now AMD is talking about doing it more," says McClean. "In some ways it is being forced upon the companies by the very high cost of these fabrication facilities." Fabs can cost more than $3 billion.

"The foundry's goal has always been to convince the IC producer not to put in that next fab," says McClean. "They tell the producer they will do the $3 billion investment to support the IC producer. That idea is starting to catch hold."

From a processing technology point of view, most foundries use leading-edge technology. "They use to be about 1.5 years behind the technology curve but that has changed," say McClean. "If foundries can produce leading-edge technologies, then the advantages of owning your own fab become less and less apparent," he says.

Foundries are very judicious about capital spending. "If business is good they will increase capital spending. If it isn't, they will keep capital spending flat or even decrease it," says McClean.

However, while foundries may be careful about their capital expenditures, some of the biggest capital spenders among semiconductor companies are foundries. TSMC will make about $2.7 billion in capital spending outlays in 2007; Nanya, $2.4 billion and Powerchip, $2.2 billion.