Intel Seen Scaling Back on Capex

The area of capital investments is a peculiar Catch-22 for the chip sector and Deutsche Bank’s latest report on Intel
illustrates this point clearly.

When it announced its second-quarter earnings on July 15th, the Santa
Clara, Calif.-based company said
that capital spending was expected to be between $3.5 billion to $3.9
billion for 2003.

In his newest report, Deutsche Bank analyst Ben Lynch said Intel will
likely push out this year’s capital expenditures into 2004. The 2003 capex
estimates will now likely fall to be between $3.3 billion to $3.4 billion.

Capital expenditure is a necessary evil for chip fabricators because
while the added investment affects manufacturing costs (and hence gross
margins), companies like Intel and Texas Instruments need
to keep up spending to insure manufacturing efficiencies down the road. In
fact, the trends toward larger fab wafer sizes of 300-millimeters (to make
more in one batch) and 90-nanometers manufacturing processes (to yield
denser, more efficient chips) has led to a shakeout in the industry.
Chipmakers like AMD and Motorola are being left by the wayside while
deep-pocketed competitors like TI, Taiwan Semiconductor Manufacturing Co.
and United
Microelectronics chase after Intel.

But during the previous tech recession, Intel itself saw the dangers
associated with capex. The company began building a 300mm fabrication
facility in Ireland in June 2000 but halted construction at the plant when
the economy soured — only to restart the work in April 2002. The facility
is currently is scheduled to begin operations in the first half of 2004.

This past February, Intel decided
to spend $2 billion to upgrade a fabrication facility located in Arizona
from 200-mm wafer to 300-mm wafer sizes. When finished, the Chandler, Ariz.,
plant will be Intel’s fifth 300-mm wafer facility. The company currently has
a 300-mm fabs operating in Rio Rancho, N.M. and another one in Hillsboro,
Ore.

A second 300-mm facility in Oregon was under construction and was
supposed to begin operations later this year. But in May, Intel decided to
push out the ramp-up of Oregon Fab 12 until 2004, DB’s Lynch wrote in his
report. (Compared to the estimated $1.5 billion it takes to build a 200mm
fab plant, 300mm fabs are estimated
to cost between $2.5 billion to $3.5 billion to build.)

As a result of the lower capex for 2003, “the difference may be shifted
into [2004] making [2004] spending flat to slightly up, versus our current
expectation of down,” Lynch said.

Part of the reason for the deferment may be due to Intel’s emphasis to
shift manufacturing to smaller and smaller densities, which allows the chips
to do more with less energy. After shifting from 0.18 microns (or
180-nanometers) to 0.13 microns (130nm), Intel has now stepped up its
efforts to get its Pentium line onto 0.09 micron (90nm) technology. Intel
president and chief operating officer Paul Otellini recently reiterated the
company’s promise to ship 90nm-based Pentium chips (code-named “Prescott”)
by the end of the year saying that testing is underway.

And in order to
facilitate the shift (as well as invest further on the future) Otellini said
Intel will spend $4.2 billion in R&D funds to developing transistors using
the 65-nm process, now that 90nm chips are in the pipeline. Chips produced
at the 65nm level are expected to appear in 2005.

“However, our checks confirm that current equipment order patterns are
not consistent with broad spending allocation to 90nm,” Lynch said.

Yet, while Intel’s move may translate to a slower-than-anticipated
migration for the overall industry, it does help the company in the
near-term. Because of the lower capex and associated decline in
depreciation, Intel’s gross margins may “see a major bounce from [the third
quarter of 2003] onwards,” the analyst wrote.

To be sure, Lynch still rates Intel’s stock with a “hold” recommendation.