Amid a weak

outlook for overall chip sales in the second half of the year, demand for the latest generation of semiconductors is expected to be a pillar of strength.

The new versions of silicon chips are capable of powering the faster, more whiz-bang PCs and gadgets that will have the greatest appeal when sales pick up.

The catch is, many cash-strapped chipmakers have been holding off investing in the technologies necessary to make the latest chips. Perhaps as soon as the end of the year, assuming demand accelerates, there could be shortages of so-called leading-edge chips, according to Auguste Richard, an analyst at First Albany. Made up of smaller-than-ever, tightly packed transistors, the newest generation of chips is carved from big, more cost-efficient silicon wafers.

The upshot: the handful of companies that have plowed money into figuring out how to manufacture the chips -- names like

Intel

(INTC) - Get Report

,

Texas Instruments

(TXN) - Get Report

,

IBM

(IBM) - Get Report

, and Asian foundries such as

Taiwan Semiconductor

(TSM) - Get Report

and

United Microelectronics

(UMC) - Get Report

-- are poised to see an additional revenue boost when IT spending increases, because the tight supply could help push up the cost of their products.

Consider the sharp decline in investment: This year's expected semiconductor capital spending dip of 10% follows on last year's 32% drop, the biggest one-year decrease ever. Tight corporate purse strings mean "there is likely to be a shortage of manufacturing capacity, particularly at the leading edge, as only a few companies invest," Richard explains.

He adds that any shortages could be made worse by the possible collapse later in the year of

Hynix

, the ailing Korean chipmaker, which accounted for about 16% of the memory chip market last year.

Against this backdrop, foundries already are nearing the limits of leading-edge chips they can produce, as they crank up production of the latest generation of chips. In the just-ended second quarter, capacity utilization for smaller transistor sizes stood at about 90%, up from 84% in the fourth quarter of last year. It could reach around 95% in the second half of '02, estimates Bill McClean, president of IC Insights, a semiconductor market research firm.

"That's where you start to see price increases, spot shortages here and there that can pop up for different components," McClean says.

Better Things in Smaller Packages

From the customer standpoint, the appeal of the new generation of cutting-edge chips is easy to understand. Consider the transition under way to smaller transistors.

"Customers want higher performance, and it takes the most advanced technology to deliver. You pretty much double the megahertz, the clock speed when you go from 0.18- to 0.13-micron" transistor sizes, explains Rob Willoner, a technology analyst at Intel. The resulting chip "is much faster; you can put more transistors on it and more cache, or memory close to the microprocessor. That translates into more performance."

Another change: Chipmakers have begun making chips from silicon wafers that are 300 millimeters in size rather than 200 millimeters. The bigger wafers yield almost 2 1/2 times as many chips.

"It gives us tremendous economies of scale and cost reductions," says Willoner.

Because customers have a preference for the more powerful, cost-effective semiconductors based on the new technologies, chipmakers that have invested in the area will have an edge when demand accelerates.

"Since the industry began, smart companies spend through the downturn so they have capacity when things turn up," says McClean.

Of course, that's easier said than done for companies facing empty coffers. The cost of a 300mm, 0.13-micron fab can reach as much as $2.5 billion, while the expense of developing a new process could cost between $500 million and $1.5 billion, estimates Richard.

Also, new factories take a long time to build. "One of the problems is the lag time between when you plan to put up new capacity and when it can come online," points out McClean. "By the time you put the first shovel in the ground for a new fabrication facility, it's 18 months to two years before it's up and running."

Yet companies that don't invest may pay a heavy price in the long term. "On the flip side, now

Motorola

(MOT)

is cutting back. It's only going to spend a couple hundred million, even though its semi business is a $5 billion business. Those companies when you look down the road have a difficult time catching up with manufacturing," says McClean. "If you decide to increase spending only when times are good, you are behind the curve."

Other companies that are underinvesting, in Richard's opinion, are

Philips

,

NEC

,

Toshiba

,

Hitachi

,

Matsushita

, and

Mitsubishi

.

In other words, the downturn has magnified the advantage enjoyed by the players with deep pockets. Foundries also are benefiting from the tough times, as chipmakers have begun outsourcing their manufacturing in a bid to cut costs.

"There are only a very few semiconductor companies that are making what are frankly some very significant advancements in this next generation of process technologies," says Ron Slaymaker, head of investor relations at TI. "What we have found historically is that when we come out of a downturn and demand starts to build, the newest, most advanced products are most in demand." By the end of the year, Texas Instruments plans to boost its 300-mm wafer, 0.13 micron capacity by 75%, to 10,000 wafers per month.

Meanwhile, Intel says it will build only facilities capable of producing transistors that are 0.13 microns and smaller. In the last year and a half, Intel has brought leading-edge capacity online at five fabs, or about a third of its total fab capacity. It will launch another leading-edge facility in New Mexico later this quarter.

Intel "has the financial muscle to outrun any company in the semiconductor industry on the technology treadmill," writes First Albany's Richard.

The giant chipmaker plans to introduce a 0.09-micron, Pentium 4-based chip by the third quarter of 2003 -- in other words, in a little more than a year it will roll out components 30% smaller than the current leading-age technology.

"Nobody's using these products

currently," acknowledges Willoner. But historical trends make clear they will, he says. "The world is using more and more computing resources. Demand for ever-more powerful microprocessors has been growing over time and we expect that to continue to grow."