Listened to the CC playback. 1-800-633-8284 code 1035568.
It is about only fifteen minutes into the call, then it was cut off!
Is this because AMD can not afford the bill? :-)
Goldman, Sachs & Co. Investment Research
Advanced Micro Devices, Inc.
* * EPS lower despite strong demand; ests unch.; MP * *
******************************************************************
* AMD reported EPS of $0.15, below our expectations despite substantial *
* unit upside. Units of 5.5 million units vs. our above consensus *
* estimate of 4.9 million, but average selling prices of $88 below long *
* term $100 target and $8 below our estimate on weaker mix. Company *
* targets flat units and rising ASP's in Q1, which will likely be a *
* challenge on both counts given the environment. No change in 99 *
* estimates of $0.70 (which have been well below consensus), continue to *
* rate the stock a hold. *
*******************************************************************
* The shortfall was driven by problems with the 400 MHz process that were
actually well known. The surprising issue is that product which was not
qualified at 400 MHz had to be downbinned all the way down to 300 or 333
MHz, essentially a liquidation product by quarter end. The company had
instituted a late-stage mask fix, expected to come out of the fab in
February, that should address the issue. The company made 400 thousand
400 MHz parts in the quarter, actually slightly above our expectations,
but only 34% of units were at 350 MHz or higher vs. mgmt projection of
50%.
* As far as the business environment for microprocessors was concerned, Q4
was the best quarter since Q4 of 1996, with all major microprocessor
vendors supply-constrained and aggregate microprocessor unit demand up
over 20% YOY. Nothing will get any easier in Q1, given a clearly
adequate supply level (with Intel clearly building a small amount of
Celeron finished goods inventory due to their Celeron shortfall) and a
competitor actively focused on bringing prices down in this segment. We
project that any ASP increase will be very small in Q1 and will be more
than offset by operating expense increases.
* Key going forward will be rate of introduction of K6-3 (reportedly 50%
of K6 output by Q4 99), K7 (expected in June), and 0.18 micron (expected
in Q3). As always in this business, very strong execution will be
required on all fronts, but the opportunity for significant leverage is
still there if all milestones are hit.
We give AMD credit for improved execution, given that the K6-2 met their
mid-year unit objectives for 98 and the company saw its highest penetration
of consumer PC's since at least the heyday of the 486. By making 400
thousand 400 MHz parts, the company positioned itself near the heart of
Intel's price/performance, and they actually were early to introduce
floating point and 3D instruction set extensions, achieving strong API
support for those new innovations.
The fact that in the strongest quarter for microprocessors in two years
earnings were slightly below expectations and returns on capital are
clearly lower then is desirable on a long term basis is a testament to the
challenges of competing in this market. As has often been the case, small
hiccups in clock speed mix depressed overall results despite 20% unit
upside.
In Q1 it is not likely to get materially easier, though with improvements
in mix the company could see an uptick in ASP. But achieving flat units in
combination with rising ASP's could still be a challenge. Flat units for a
retail driven processor company requires outperforming the company's
primary served market (the retail-based consumer business) by 15% or so, in
an environment where Intel has already outlined plans to bring its own
average selling prices down towards the longer term level specifically in
an effort to make life difficult for AMD. One of the initial reasons cited
for this outperformance was increased penetration of Europe, but that
penetration occurred in Q4 as the company cited over 1.4 million units
shipped to Europe and increasing penetration substantially in Q1 would be
challenging. The outperformance would have to come from a combination of
increased penetration of existing OEMs, the early rampup in the K6 notebook
business, and, potentially, penetration of one or more direct PC vendors.
We are budgeting for flat units and ASP's of $91, but in our view this
could be a challenge without some penetration of the low end of Gateway's
business. We continue to hear expressions of interest from Gateway
management (as we have since Q2 98) but so far nothing has been confirmed
as Intel has aggressively worked to keep Gateway 100% Intel.
From a product standpoint, in our view the most significant issue in Q1
will be Intel's introduction of the socketed Celeron product. One of the
major reasons for AMD's unit success in 1998 has been Intel's insistence
upon moving the world to slot one, when slot one implementations are
actually quite expensive. Above and beyond the cost of the cartridge,
which is not unsubstantial, the cost of managing heat and power in a
cartridge system is as much as $15 higher than in a socket system, which
became material in the low end consumer market in 1998. Beyond a doubt,
Intel would have held stronger short term market share in 98 by increasing
speeds of socket 7 Pentiums instead of moving the market to low end Pentium
II's (Celeron's), but felt it was strategically important to move the world
to a proprietary bus standard. When Intel moved the SRAM cache out of the
cartridge and onto the die (with the Celeron version code-named
''Mendocino'), the cartridge became irrelevant but the high cost of
implementation remained.
By introducing a version of Celeron that is in a proprietary socket, Intel
initially created some OEM backlash (from those who wished Intel had stayed
with socket 7 all along) but has clearly now seen a strong 370 pin socket
support. By introducing the socketed product, Celeron implementations will
be less expensive than previously. Further, Intel has clearly shown a
willingness to bring Celeron ASP's down in Q1 and, potentially, beyond,
given that Q4 ASP's were about $10 higher than where we believe Intel would
like to keep them ($210-$215 longer term vs. Q4's $224). AMD will respond
in Q1 by accelerating clock speeds as the 400 MHz fix comes out of the fab
(February) and pushing into the low end of the notebook space, where Intel
doesn't really have a product currently. We project flat units and 2%
rises in ASP's, which does imply down EPS given ramping operating expenses.
In Q2 the playing field will shift a bit again as Intel starts to ramp the
Pentium 3 (Pentium 2 with Katmai instruction set extension) and AMD will
ramp K6-3 (K6-2 with integrated L2 cache). In our view Katmai will be an
effective transition for Intel, effectively differentiating their
mainstream line from Celeron and allowing even more aggressive Celeron
pricing without disruption of the mainstream (and offering a more
compelling 'trade-up' for Intel). AMD, of course, has already introduced
its own instruction set extension, called '3D-Now', and they are focusing
on the fact that they have support in Microsoft's DirectX 6 while Katmai
will not and that the current installed base of 3D Now machines dwarfs the
installed base of Katmai.
This is important in the near term but it is a non-issue for the longer
term: consumers don't buy PC's just to run the software available the day
of purchase, they buy PC's to run software for the next several years.
Very few ISV's in today's game market use Direct3D anyway, choosing instead
to write directly to the hardware, and Katmai will likely have more staying
power than 3D-Now (and we would be surprised if AMD wasn't readying Katmai
instruction set compatible processors down the road). Of course, AMD's
strong retail momentum guarantees that there will be a strong installed
base for game designers today. But if AMD eventually does move to Katmai
instruction set capability there will be no incentive for game developers
to include 3D-Now support two years from now, only about halfway through
the life of the average consumer PC.
Q3 will tell us how strong AMD will perform in 2000, as they start to ramp
both 0.18 micron wafers and the K-7. Clearly, this is where execution will
be critical for the stock price performance, though 0.25 micron K6's will
drive earnings, since AMD's success in ramping both 0.18 and K-7 will
determine 2000 results. With K-7 (which looks, at least on paper, to be a
very powerful chip), AMD is trying to create its own alternative Slot A
infrastructure, and any delays will make such infrastructure creation more
difficult. AMD management says that they will have a higher percentage of
wafers on 0.18 micron in Q4 than Intel (and told us 'write that down').
Non microprocessor businesses continue to be difficult, with Q4 revenues
flat from Q3 as expected and projected to be flat in Q1. The company has
seen some turbulence in its Japanese distribution channel for its PLD
products during Q4, and flash pricing continues to be problematic. We do
expect the flash market to recovery through 1999.
In general, despite the tough pricing in Q4 the company continues to have
opportunities with strong execution. We are watching the progress of 0.18
micron and K-7 carefully and would become more upbeat on the stock for 2000
with stronger visibility on those transitions. In the meantime we continue
to be cautious about the overall environment in the first half of 1999,
when aggressive ramps in wafer starts from all PC component vendors are
likely to allow supply to catch up to demand. We therefore continue to
rate the stock a market performer. |
