Semiconductors
Q4 Earnings Preview
January 8, 2002 SUMMARY
* We believe most of our broadline and personal computer
Jonathan Joseph component companies will meet, or slightly exceed our CQ4
estimates, and provide guidance for revenues to be flat
to slightly down in Q1. This is in-line with analyst
Clark Westmont expectations and the seasonal norm.
* PC components remain the strongest segment, led by
microprocessors and memory. That component market should
Ramesh Misra be flat or slightly down in Q1 as well, in-line with the
seasonal norm.
* Communications ICs will be more of a mixed bag. Given
the limited visibility (i.e. backlog) at this time,
managements will likely lean toward conservatism, which
is already reflected in most analyst models.
* Historically, P/S and P/E valuations are not cheap (as
would be expected at a cyclical bottom), though P/B is
more reasonable. With renewed earnings momentum underway,
we believe "the tribe has spoken" with regard to lows.
SSB SEMICONDUCTOR COMPANIES' QUARTERLY EARNINGS PREVIEW SUMMARY
Source: Salomon Smith Barney
SSB SEMICONDUCTOR COMPANIES' REPORTING SCHEDULE
Source: Salomon Smith Barney and Company Reports
OVERVIEW: CAN THE UPTURN DEVELOP SYMMETRY?
Though the notion was controversial a couple of quarters ago, debate over
whether the semiconductor sector hit a bottom in Q3 has pretty well dried up.
Since bottoming in July at -51%, an all-time record, semiconductor orders
have reaccelerated to -32%, a roughly 5percentage points per month change
that is generally symmetrical with the slope of the downturn. Since bottoming
in September at --45% yoy, an all-time record, semiconductor shipments have
grown for two consecutive months by about 2% mom in October and November,
which if it continues, would describe a roughly 25% annual growth rate within
12 months. If the upturn is aided by firmer pricing---as we have already seen
in the microprocessor and DRAM segments---the recovery may show a high degree
of symmetry to the downturn. But first, lets look at expectations.
As we mentioned in our Semiconductor Beat this week, we analysts and
investors live in a world of "second derivatives". Far more important than
actual results are results relative to expectations. That is what makes Q1 so
important. We had earlier believed that analysts Q1 outlook for revenues
might be somewhat optimistic given the general habit of modeling in
sequential revenue gains without regard to seasonal trends (the "hope springs
eternal" school of analysis that caused some volatility in National's stock a
month ago). However, when we checked FirstCall revenue forecasts for the
companies we follow, we found that aggregate revenues estimates called for a
1.4% decline from Q4 to Q1, compared to an average 0.2% increase over the
last 11 years for these companies. The Street estimates are, however, pretty
much in-line with the seasonal norm of a 1% qoq decline in overall semi
shipments, according to SIA data. This is pretty good news, because we
believe most companies will essentially guide to flat revenues in Q1, based
on backlogs that grew in the quarter, inventories that are low, and customer
indications for a seasonally in-line outlook.
Q1 is not an easy quarter to predict, and historically it can be very
volatile. In 1986 and 1995 (a recovery year and a peak year, respectively),
revenues grew 10% and 9% qoq. By contrast, in 2001 (a downturn year, if you
need reminding), Q1 revenues were down a record 19% qoq. Historically, that
is in contrast to Q4, which is widely believed to be a seasonally strong
quarter. We recently revised upward our outlook for Q4 from 0% growth to 5%
growth. Recall that SIA figures released a week ago showed November described
the second month of sequential increase in shipments, which grew 2% mom and
down 42% yoy (compared to down 45% yoy in September).
FIGURE 1. SEQUENTIAL CHANGE IN REVENUES IN MARCH QUARTER VS. DECEMBER QUARTER
Q1 revenues vs.
Q4 revenues
SSB Industry Forecast 2 %
SIA Industry Forecast* ~ 3 %
SSB Company Estimates** 2.0 %
Street Company Estimate -1.4 %
Company Historical Performance ^ 0.2 %
Overall Semi Shipments S -1.0 %
* Our estimate of SIA projections, based on their full year forecast of
+6.3%
** SSB estimates for 16 of our primary broad-line semi companies
^ Average historical performance for these companies from 1991 through 2001
S Average total semiconductor shipments Q1/Q4 since 1985
Source: Salomon Smith Barney, The Semiconductor Industry Association, and
FirstCall
Now, back to pricing. Do not underestimate the power of pricing to impact
revenue changes. Powerful upward price trends have been most notable in the
DRAM segment (currently 8% of total shipments), where spot market prices have
more than tripled over the last two months (see below). Significantly, DRAM
bit growth has also been quite strong, growing about 8% mom and 110% yoy in
November. In addition to DRAM, we have seen positive pricing trends in
microprocessors. Advanced Micro pre-announced to the upside based on an
estimated 20% qoq increase in prices, a trend that could carry over into Q1.
In addition, we estimate Intel recorded a 3-4% qoq gain in prices, and should
see further gains in Q1 as product mix continues to favor the P4. Pricing is
firming across a broad range of product, with foundry wafers, packaging
services, and discrete semiconductors all seeing better than forecast
pricing. All told, ASPs rose by about 1% in November (following a 1% increase
in October), according to the SIA, a recovery trend we believe is just
beginning.
FIGURE 2. DRAM ASPS RECOVERING STRONGLY AFTER DECLINING MOST OF THE YEAR
Source: Salomon Smith Barney and Converge
FIGURE 3. PROCESSOR ASPS TRENDING UPWARDS TOO
Source: Salomon Smith Barney Estimates
PC AND BROADLINE IC OVERVIEW: SEASONALLY STRONG Q4 EXPECTED
The most dramatic tale in the quarter has been the very sharp turn-around in
DRAM pricing. After being down about 79% from January to October, spot market
prices then more than tripled. They were up 31% last week alone. What the
heck is going on here? Though we recognize there is a fair amount of broker
speculation underway, we also believe there are some sound fundamental
reasons for the move. One reason is the strongly growing demand for DRAM,
aided by Windows XP and new processors. Production cutbacks---both temporary
and long-term---have been announced by a number of firms, making nervous
buyers begin to rebuild inventory. Added to this were the already low
inventory levels in the channel, and consolidation discussions among some of
the largest manufacturers, and we had a recipe for an explosive snap-back in
DRAM pricing.
In the microprocessor segment, our channel checks have suggested that Q4
processor shipments should come in slightly ahead of the expectations for
both Intel and AMD. Additionally as we mentioned, the companies have been
aided by a strong pricing rebound.
There has been plenty of macro data in recent weeks supporting our outlook
for a stronger-than expected Q4 and a further improvement going into 2002.
For example, the PMI (Purchasing Manager's Index, which has replaced the
National Assoc. of Purchasing Manager's Index) reported that the
manufacturing sector of the U.S. economy has entered its 17th month of
decline in December. However, one of the few segments that showed a rebound
was Electronic Components and Equipment. In addition, Electronic components
was the only product area that was listed as "short in supply".
Sector data has at the margin been a little more positive than most of the
large U.S. PC OEMs anticipated, according to Richard Gardner, our SSB PC
analyst. US retail PC sales in Q4 rose 30% sequentially, compared to a 15-20%
norm, thanks largely to a strong holiday selling season and a sub-par
September. On the other hand, 2H01 shipments rose only 7% over 1H01, about
one half the normal half on half rise. Europe was also shaping up a little
better than analysts' pessimistic post-September 11 forecasts, and that
region should post a 10% qoq rise PC shipments in Q4. Though this is only
about a third the typical Q4 increase, guidance was for more moderate growth.
Finally, US corporate feels "flattish" sequentially in Q4.
We believe the modest upside surprises in Q4 presage an accelerating
turnaround in the group as the year progresses.
COMM ICS MIXED BAG STILL
The communications IC category is pretty much a mixed bag at this point.
Overall, expectations seem to have been reigned-in sufficiently since there
were no negative pre-announcements made within our comm IC stock coverage
during the last three months. There are signs of stabilization or recovery
in chips targeted at enterprise and cable end-markets. Telecom, in contrast,
appears to have seen no improvement in December with limited prospects for
sequential growth in March 2002.
As usual, the outlook for March and the rest of 2002 will be the primary
investor focus. Given the limited visibility (i.e., backlog) at this time,
most companies are likely to lean toward conservatism as they presumably did
three months ago. In general, we believe our models and Street expectations
already reflect that scenario. See the individual company write-ups for more
details.
In aggregate, we expect that sales for the communications IC sector will
decline 5-10% in 2002 compared to a 35% decline in 2001. This is roughly
consistent with our sales estimates for the nine broadband communications IC
stocks that we follow (combined sales down 30% in 2001 and estimated down 10%
in 2002). Several factors account for this phenomena. First, carrier
spending is expected to decline further during 2002. Second, the industry
exited 2001 at or near the lowest quarterly run rate of the year. We will
publish a detailed comm IC industry model shortly.
Paradoxically, while we do not anticipate a pickup in overall dollar terms
during 2002, some of the quarterly sequential sales comparisons could appear
high on a percentage basis. In large part this is because most Internet IC
companies' Q4-01 sales are 50-80% (!!) below peak levels, and it will take
only modest dollar increments to drive seemingly large percentage changes.
Of course, sequential rates carry less meaning when coming off a base that is
a fifth of peak levels. Nevertheless, it should be recognized that some
wireline IC companies might be able to achieve strong (even double-digit)
quarter-by-quarter percentage growth in 2002 as the inventory correction
clears and end markets stabilize (albeit at substantially lower levels).
In our opinion, the stock environment should be more favorable in 2002
because comm IC stocks are usually driven by sequential sales changes. In
addition, consensus expectations are generally bleak for the group. For
those that are interested in year-over-year comparisons, we believe that Q4-
01 versus Q4-00 represents the hardest comparison for most Internet IC
companies.
Ironically, despite widespread near-term skepticism about the group,
valuations are roughly 100% higher than the 'non-comm' semiconductor industry
average -- i.e., 13.0x versus 6.4x estimated 2002 sales as of this writing.
We believe this phenomenon reflects two factors. First, the communications
IC group's high beta has made it attractive during the market rally from
October through November. That effect can easily be undone by a market
pullback. Second, despite the difficult near-term environment, it seems that
most investors still believe that the long-term growth potential in the
communications market is relatively better than other semiconductor end-
markets. We agree with this second proposition.
Since fundamentals in the broadband cable IC segment (e.g., ICs for digital
set-top boxes and cable modems) appear to be better than other segments,
stocks such as Broadcom and Microtune represent relatively defensible ways to
play the group, in our opinion. Telecommunications IC suppliers such as
AMCC, PMC-Sierra and Vitesse represent higher risk/reward prospects.
There are stocks in our communications IC coverage that are attractively
valued such as Agere Systems (4.0x estimated 2002 sales as of this writing),
Centillium (also 4.0x) and Multilink (4.7x).
Generally, the prevailing wisdom is that a recovery will follow a 'First In,
First Out', so the sectors that were fundamentally first into the downturn
will be the first out. For example, enterprise-oriented communications ICs
were among the first into the downturn and the worst now seems behind us.
Telecom ICs are likely to have the last recovery in fundamentals.
For better or worse, the communications IC stocks do not track the FIFO logic
very closely. For example, Broadcom's business is better than PMC's, but
both stocks are up 160% from the 52-week lows. With few exceptions,
comparing a two-year chart of the SOX with a similar chart for individual
comm IC stocks shows a similar pattern: a NASDAQ-related peak in March 2000,
a steep decline in Q2-00, a recovery close to peak levels in the middle of
the year, followed by a plunging decline from September 2000 into April 2001.
We conclude that the sector decision (e.g., overweight semis) overshadows the
sub-sector choice.
PLD COMPANIES HITTING GUIDANCE TARGETS
The programmable logic device (PLD) category seems to be exhibiting many
signs of a cyclical bottom and turn at this point in time. First, the
December quarter updates given by each of the three largest PLD companies
(Xilinx, Altera, and Lattice) were basically a reiteration of prior sales
guidance of down 5-10% in December -- in fact, Xilinx stated that its sales
are tracking toward the high end of that range.
In addition, distributors suggest that PLDs are one of the best segments in
the electronics component market. Avnet (AVT-$27, NR), for example,
indicates that the book-to-bill ratio for its PLD business in North America
and (non-Japan) Asia are above unity and has been for months. Our North
American channel checks also suggest that the book-to-bill ratios have
improved.
We believe that PLD sales in Europe (about 20-25% of the mix) are still in
sequential decline in Q4, but probably not as bad as originally feared.
Japan (representing 15-20% of total sales for Altera and Xilinx) continues to
be the most troubled geographic region for PLDs.
Overall, we estimate that the overall book-to-bill ratios for Altera and
Xilinx during December were slightly above 1.0, with Xilinx tracking slightly
better than Altera. Based on the September sales mix, Xilinx's geographic
exposure is more favorable with 53% of September's sales derived from North
America versus 44% for Altera. Xilinx also has a slightly better product mix
-- for example, more FPGAs vs. CPLDs and the solid Virtex family.
Altera indicated on January 7th that it expects modest growth in March. In
our view, both companies are likely to guide to 1-5% sequential sales growth
for March, which is in-line with our models but should be taken positively.
Typical 'first half' seasonal strength and the cyclical recovery in chips
will push the actual results to the high end of that range, in our view.
'Turns'-dependency remains high, however, on the order of 50-60%. In other
words, neither company is likely to start the March quarter with more than
50% of the target in backlog. This is not an unusual state of affairs for
the PLD sector, but it does imply limited visibility. We expect that the
margin and inventory results will be in line with expectations, and the
revenue picture will be the most closely watch part of the picture.
# Within the past three years, Salomon Smith Barney, including its parent,
subsidiaries, and/or affiliates, has acted as manager or co-manager of a
public offering of the securities of this company. |
