<font color=red>yet another intel bug - a faulty graphic bus interface with the
Tehama chipset.
07:11am EDT 2-Oct-00 Credit Suisse First Boston (Glavin, Charles (415) 836-77
INTC: Scraps Timna, Delays P4-Launch; No Change to Model - Pt1 FBC
CREDIT SUISSE FIRST BOSTON CORPORATION
Equity Research
Americas
U.S./ Technology/Semiconductors
Charlie Glavin, CFA 1-415-836-7715 charlie.glavin@csfb.com
Regina Eberhart 1-415-836-7767 regina.eberhart@csfb.com
Caroline Moon 1-415-836-6321 caroline.moon@csfb.com
STRONG BUY
Intel Corporation (INTC)
Scraps Timna, Delays P4-Launch; No Change to Model
Summary
Intel announced that it is canceling its Timna chip, targeted for sub-$600
market. The move was driven by the high cost vs. Celeron/810E options and
the high cost of the Rambus interface, and was apparently prompted by two
large OEMs canceling plans for using Timna.
We also believe Intel has delayed the launch of its P4 (Willamette) chip from
October to late November, due to a faulty graphic bus interface with the
Tehama chipset. Tehama also supported only Rambus.
Finally, it appears that lower output of 1-GHz and other faster Pentium-3
chips was due to hastily implemented internal mask-sets-not manufacturing
yields or design problems-that caused slower-than-expected speed outputs. A
catch-up to plan should occur by January.
The Timna, Pentium-4, and 1-GHz products were not expected to contribute
material revenue for 6-9 months, so we are not changing our model. Meanwhile
, Intel could beat its revised guidance for 3Q (3-5% growth), given a modest
rebound in Europe.
Price Target Mkt.Value 52-Week
09/29/001 Price Div. Yield (MM) Price Range
USD 41.56 100 0.02 0.13% $291,127.8 76-32
Annual EPS Prev. Abs. Rel.
(Post-Split) EPS P/E P/E
12/01E $1.87 22.2X 95%
12/00E 1.67 24.9 98%
12/99A 1.16 35.8 125%
March June Sept. Dec. FY End
2001E $0.44 $0.44 $0.47 $0.52 Dec.
2000E 0.35 0.49 0.38E 0.45E
1999A 0.29 0.25 0.27 0.35
ROIC (12/98)
Total Debt (6/00) 1255
Book Value/Share (6/00) $5.23
WACC (12/98)
Debt/Total Capital (6/00) 3%
Common Shares 7005
EP Trend2
Est. 5-Yr EPS Growth 25%
Est. 5-Yr. Div. Growth 29%
1On 9/29/00 No text of specified style in document. DJIA closed at 10,650.9
and S&P 500 at 1436.5.
Intel is the world's largest semiconductor company and the largest supplier
of microprocessors (CPUs) with an estimated 80% unit share of the PC
processor market. The company is also the largest supplier of Flash memory
and PC chipsets and a leading supplier of low-end networking solutions.
Timna Cancellation, P4 Delays Add Salt to the Wound
Following up on its earnings pre-announcement, Intel also disclosed on Friday
that it plans to cancel its Timna product, which was originally targeted for
the sub-$600 market that appeared to be the next big segment for PC-makers.
However, two things happened: first, the market for that segment cooled off;
and second, the cost of an integrated processor such as Timna (including the
graphics, chipset, and CPU) was no longer advantageous relative to other
options. Since Timna essentially preset all of the configuration options for
the PC OEMs, it limited the ability of those OEMs to differentiate their
products. Feedback from OEMs also indicated that the cost of supporting
Rambus memory and the MTH chip needed to enable RDRAM was just too expensive (
ironic, given that Rambus originally made its name making memory for video
game machines with similar pricing). Meanwhile, Intel's current Celeron
processors, along with an integrated 810E chipset, essentially provide the
same solution without forcing OEMs to design a separate platform just for
Timna-that is, OEMs could use Celeron both for the sub-$600 and the sub-$1000
segments. We also believe that the cancellation of Timna could provide a
limited opening for competitive products from Transmeta and Via.
We believe three factors affected Intel's decision to pull the plug on Timna:
Delays forced Intel to miss this year's window. Back in June, Intel
announced that due to complications with the companion MTH (memory
translation hub) chip needed to support RDRAM, it was delaying the launch
from 3Q00 to 1Q01, and thereby missing the critical Christmas season.
Two major OEMs recently expressed no interest in pursuing Timna. As
mentioned above, we believe that OEMs no longer saw the sub-$600 market
taking off as once forecasted, and Intel was already offering an adequate
solution with Celeron. Rather than force OEMs to design a separate platform
and take up valuable internal wafer capacity for building Timna, Intel
decided to end it now before it went into full production.
We believe Timna also lack one critical feature: no integrated wireless
capability. While not cited as a reason, we believe that for any end product
that was to compete directly or indirectly with tablets, RIM pagers, Palm
Pilots, or Nokia 9000 devices, it had to have some sort of wireless capability
. StrongARM is certainly an embedded processor up to the task, but Timna did
not have it.
The timing of these events could not have occurred at a worst time, but it
was natural. Intel is currently going through its strategic review processes
, and had it not pre-announced lower earnings for 3Q, the cancellation of
Timna could have been perceived as a disappointing but prudent action for
Intel (kill it now before it goes to production). Unfortunately, it is yet
another straw undermining investors' confidence in Intel's execution.
P4 Delayed, Due to Interface Problems from Tehama chipset
While Intel continues to say that the Pentium-4 (Willamette) processor is on
track for a "Q4 launch," we believe OEMs were notified last week that the
official launch would be delayed until November 20 (from either an October 20
or October 30 launch, depending on to who you talk). While the processor
itself is fine, we believe that the Tehama chipset was having problems with
the graphics interface, failing at high speeds. Rather than have a repeat of
last year, when it had to recall many of its 820 chipsets, Intel decided to
do one more silicon spin to resolve the problem before it went out. Even a
minor design tweak like this results in about a one-month lag. The 820
problems were due to issues surrounding support of Rambus DRAM; however,
while Tehama is also dedicated to supporting RDRAM, we do not believe this
was the main culprit. A non-RDRAM version for the P4 (called Brookdale) should
be out in late 2Q01.
Speed Issue Problems Identified: Internal Mask Flaw
We also believe that Intel has resolved a nagging problem that has hurt its
pride and bragging rights against AMD. For the past two months, Intel has
not been getting high-speed yields on its Pentium-3 processors. We now
believe the problem stems from mask sets that were put into production too
quickly. Intel makes its own mask-sets internally. To be clear, we do not
believe that this is a problem with Intel's manufacturing yields, the basic
chip design, and not a problem from the equipment makers for mask sets.
One possible explanation is how masks are made. The basic physical design is
laid out, and essentially fit to a grid (think of it as very, very small
piece of graph paper). This way, the machines can work in a more precise
manor. Software checks the layout to make sure that when the layout is set,
the grid hasn't caused the lines to be too close together. Such a situation
could cause a short, which would probably render the chip useless, thus
lowering yields. Erring on having lines too wide is more likely to result in
a reduction in speed, but not an outright failure.
In its efforts to accelerate its roadmap, we believe Intel did not fully
simulate and test the new mask sets before they went into production, thus
resulting in adequate yields in terms of good die, but slower-than-expected
bin-outs. A bin-out is essentially a range of speeds that a set of
processors can pass. For example, one wafer may have half of its processors
pass speed tests for 933-Mhz, 30% for 866-MHz, but just 10% for 1-GHz. You
can sell a faster-speed chip as a low one (called down-binning), but you
obviously cannot do the reverse (sell a chip that only passed a lower speed as a
higher speed).
In this case, we believe that Intel is in the process of re-doing its mask
sets, with full production back on track in January. In addition, we believe
this issue had nothing to do with the Pentium-4 delay. Right now, there is
very little volume for the 1-GHz or higher speeds for either Intel or AMD,
but nevertheless, we believe it prevented Intel from accelerating its processor
roadmap.
EPS Estimates Unchanged; Maintain Strong Buy
The Timna, Pentium-4, and 1-GHz products were not expected to contribute
material revenue for 6-9 months, so we are not changing our model. Meanwhile
, Intel could beat its revised guidance for 3Q (3-5% growth), given a modest
rebound in Europe. |
