Next year will put to the test a key strategy for chipmakers: Invest as much as possible in technology during the downturn in order to become that much more profitable when semiconductor demand picks up.

The problem is that the recovery for semiconductors next year is expected to produce growth of only about 10% -- a pretty middling rebound. Big, rich chipmakers such as Intel ( INTC), Texas Instruments ( TXN) and IBM ( IBM) probably won't see the returns they would have expected a couple years ago when they started investing billions in state-of-the-art chip fabrication plants.

To be sure, leading semiconductor outfits may not have a choice about whether to invest, assuming they want to gain an edge in the long term. But the takeaway for investors isn't a cheery one: Chipmakers must spend tremendous amounts of capital to stay competitive, and as a result, offer increasingly humdrum growth prospects. That combination is bound to depress profitability amid a recovery -- and probably in the longer term as well.

A look at one measure of profits, return on capital, underscores the trend. A study of 15 chipmakers shows that returns fell to a 10-year low, at an average of 1.59%, in the third quarter, according to a recent report from Banc of America. That compares with an average of 17% annually over the past seven years.

Or consider Intel. In the fat days it boasted a return on capital as high as 30%, but as of the September quarter, its ROC was a mere 6.6%.

Chipmakers Can Expect Less Bang for Their Bucks
Biotech Looks Past Its Latest Annus Horribilis
Fund Firms Keep Customers by Expanding the Menu
Price Pressure Could Reduce ISPs to Tiers
Big Software Aims Small in 2003
Most Sectors Slide, but Gold and Silver Shine
Five Stocks Shine in Dreary Year
Networking: Next Year: Cisco, Juniper and Red Ink
Cisco Shift Gums Up the Works for Rivals

Granted, those numbers will improve when demand for chips picks up again. But analyst Mark FitzGerald still expects future returns for chipmakers in general to fall well below the historical averages, to the 10% range.

"In the '90s, it paid to be aggressive in investing when the market for PCs was high growth. PCs had 15% to 20% unit growth through most of the '90s. But now single-digit growth rates don't justify the same aggressive level of capital spending," he says.

Costs Grow but Returns Shrink

Consider the two parts of the equation: a diminishing growth outlook and the spiraling cost of new chip fabrication plants.

Just last month, the leading trade group acknowledged there's been a long-term shift towards a lower growth rate for semiconductors, given the saturation of markets from PCs to autos. Going forward, the industry should grow at only 8% to 10% annually, the Semiconductor Industry Association said, well below the historical rate of around 16% or 17%.

At the same time growth is flagging, manufacturing expenses have shot up dramatically as technology has grown more complex. A recent Merrill Lynch report noted that Moore's Law (the number of transistors that can be placed on a chip doubles roughly every 18 months) has a practical corollary: The minimum cost of an economically viable fab doubles every three years.

Hey, Big Spenders
Chipmakers spend billions on equipment and fabs for new semiconductor processes
Company Capital Expenditure Budget in $
Intel $4.7 billion
IBM (expenditure for total company, not just semiconductor division) 5 billion
TSMC 1.65 billion
Texas Instruments 800 million
Source: Company reports

But semiconductor revenue certainly isn't growing at the same pace.

With the cost of a cutting-edge factory estimated at $1.5 billion to $2 billion, all but a handful of chipmakers are being priced out of the manufacturing business. The rest are going "fabless," meaning they outsource the actual production of semiconductors to so-called foundries.

Meanwhile, a few select chipmakers have plowed ahead with investments in new fabs that can process 300-millimeter-sized wafers, which produce 2.4 times more chips than the more common 200-millimeter wafers. Semiconductor companies are also continuing to shrink the size of transistor linewidths to a mere 90 nanometers, a move that not only reduces costs but also enables more functions to be integrated on a piece of silicon.

IBM's leading-edge 300-millimeter fab in East Fishkill, N.Y., which started pilot production in August, will ramp up into volume production throughout 2003. Texas Instruments' Dallas fab, now at 5,700 wafers per month, will produce up to 10,000 a month by the end of 2002. Both companies plan to begin producing on 90 nanometer linewidths next year.

Intel has positioned itself at the forefront of the shift to larger 300-millimeter wafers. With two 300 mm fabs now in production, Intel will have another up and running by the close of next year. By the end of 2003, up to half of all Intel's processors may be produced at 90 nanometers, estimates UBS Warburg.

Yet with Intel's capital spending expected to reach $4.7 billion this year, equal to 18% of expected revenue, some analysts maintain the chipmaker is investing too heavily -- even though capex has dropped from 27% of sales in 2001. A ratio of 10% to 15% is a healthier ratio for chipmakers, given slowing growth rates, says Banc of America's FitzGerald.

"They can't afford it because the end markets aren't growing enough," he says. "If they throw more money at it and the end markets are seeing slowing growth, they're just digging a bigger hole for themselves."

Intel CFO Andy Bryant declined to comment on its future spending strategy, citing the quiet period leading up to the company's January earnings release.

Yet in light of the so-so demand environment and expensive manufacturing assets, some think Intel will find it tough to maintain its return on invested capital. "That's the main reason behind our sell recommendation on the stock," says Merrill Lynch analyst Joseph Osha.

Even so, wealthy Intel is much better equipped to deal with pressure than smaller fry that manufacture their own semiconductors. "We've said stocks like LSI Logic and Atmel are going to get squeezed from both directions," says Osha.

Unlike smaller companies, he adds, Intel could justify its strategy over the longer term, because it can afford to make big investments and because the resulting improvements in manufacturing will enable it to distance itself further from competitors. "If I were Intel, I'd do the same thing," says Osha. "If I were huge and rich, I'd probably use my biggest asset, which is being big and rich."

Indeed, some investors cast the aggressive-investing approach as the sensible, competitive response in a notoriously tough industry. "The ones taking the risk are the ones who are not spending," contends Graham Tanaka, head of Tanaka Capital Management and an Intel investor. "Look what happened to the Japanese DRAM-makers in the late '80s and early '90s. They stopped spending, and the Koreans took their business," he says. "Intel is doing the right thing" by investing heavily.

Take the transition from 200-millimeter to 300-millimeter wafers. "They allow you to pack on over two times as many chips per wafer, so your cost per chip is actually declining, even though the initial cost is higher," points out Tanaka. "You're getting much more chip production per line." That will be a boon to Intel's business when the economy accelerates, he says.

Meanwhile, smaller linewidths allow chipmakers to differentiate their processors, muscling out rivals. Intel reckons that at 90 nanometers it can crank out its next two processors at clock speeds of roughly double the 3 gigahertz of the Pentium 4.

Intel and other chipmakers that stay in the manufacturing game over the next few years should eventually face better prospects, even if profitability doesn't return to '90s levels. "The return on invested capital for guys who survive is going to improve," says Susan Crossley, an analyst at Wells Fargo. "If you play out all these doomsday scenarios about slowing growth rates and consolidation, what you end up with is a scenario a few years from now where there are a few real big winners."

But don't expect the interim to be painless. "I think next year will serve as a test," says Osha. "2003 doesn't look like a sharp recovery. But it will take a few years to shake out."