Intel Corp (INTC)

[burn2learn]

UBS on "What if AMD beats" thesis. Warning very long

Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 1 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005
UBS 1 July 2005
www.ubs.com/investmentresearch

Thomas A. Thornhill III, CFA ([phone]) [e-mail]
Analyst
Steven Eliscu ([phone]) [e-mail]
Associate Analyst

US Semiconductors, Americas
Advanced Micro Devices
(AMD.N) (AMD US)
RATING:Reduce 2 PRIOR: Unchanged

Price Price Target 52-week range Mkt Cap
-------- --------------- ----------------- ----------
US$17.37 US$11.00 US$24.85-10.86 US$6.83bn
-------- --------------- ----------------- ----------


What If AMD Beats Our Current Market Share Projections?




* AMD thesis: limited MPU share gain, tough Spansion IPO:
We believe AMD's stock is overvalued as our analysis of AMD's businesses
supports an $11.00 target price and Reduce 2 rating. We believe built into
AMD's stock price are x86 MPU share gains that are unlikely to be achieved
and overly optimistic expectations for a successful IPO for its Flash memory
subsidiary.

* Assumptions for maintaining the Reduce 2 rating for AMD::
1) Intel's NB lead widens, 2) AMD's share growth in DT is limited, 3) Intel
is successful with a new class of DTs, 4) Intel is price aggressive in
low-end servers, 5) IT mgrs will be slow to deploy Opteron, 6) Intel's 9-12
month process lead is maintained, 7) Uncertainty regarding AMD's capital
reuse strategy, 8) Intel's business model runs optimally when it has about
80-85% market share.

* "What if we are wrong?" - Baseline and Optimistic Scenarios:
We test our expectations against more optimistic share gains scenarios.
Even these scenarios only support current valuations where we believe other
stocks under our coverage are priced similarly (ALTR, ADI, MXIM) or less
(INTC, TXN) and have lower risk. In the note we detail the risks associated
with each of our underlying assumptions.

* Valuation: Maintain $11 Target Price, Reduce 2 Rating:
Our 12-month target price of $11 is based on applying our 18x target PE
multiple to our revised 2006 EPS estimate of $0.61. We maintain our Reduce 2
rating.

This report has been prepared by UBS Securities LLC
UBS does and seeks to do business with companies covered in its research
reports. As a result, investors should be aware that the firm may have a
conflict of interest that could affect the objectivity of this report.
Investors should consider this report as only a single factor in making their
investment decision. Customers of UBS in the United States can receive
independent, third-party research on the company or companies covered in this
report, at no cost to them, where such research is available. Customers can
access this independent research at www.ubs.com/independentresearch or may
call [phone] to request a copy of this research.


Scenario Analysis: What If AMD Beats Our Current Market Share Assumptions?

AMD Investment Thesis

We believe AMD's stock is overvalued with significant downside risk.

We believe AMD's stock is overvalued, as our analysis of AMD's businesses
supports an $11.00 target price and Reduce 2 rating. We believe that at
current share price levels, there is significant downside risk.

We believe built into AMD's current valuation are x86 processor share gains
that are unlikely to be achieved.

We believe AMD's current valuation is based on x86 processor share gains that
are unlikely to be achieved - we show an extensive analysis in the sections
below.

As AMD's Flash memory subsidiary, Spansion, has filed an S-1 (and subsequent
S1-A) in preparation for an IPO, we believe some investors may be excluding
Spansion, and its related drag on AMD results, from AMD's consolidated
earnings for valuation purposes. As Spansion is currently losing money and in
our opinion is unlikely to show profitability soon, we believe there is
material risk that the IPO will at a minimum be delayed (see discussion
below). If this were to occur, and investors realize that AMD will continue
to be burdened with Spansion, AMD's stock could react negatively.

AMD's antitrust complaint / anti-Intel marketing campaign does not change our
thesis

As outlined in our AMD First Read note on June 27th, we do not expect AMD's
antitrust complaint filed against Intel to have a material effect on near term
business trends for either Intel or AMD. While a settlement or injunction is
possible, AMD management expressed to us that they believe this is highly
unlikely and would expect the complaint to go to trial in late 2006.

We believe the only near term impact is likely to be an increase in legal and
advertising expense, which will have a larger negative impact on AMD than
Intel. We expect AMD to provide guidance for its legal expenses related to
this complaint on it July 13th 2Q05 earnings conference call. Our first guess
is that the legal and related anti-Intel marketing campaign expenses will
burden SG&A in the $10-15 million range per quarter.

It is exactly because AMD has failed to appreciably gain share against Intel
during the last two years that so poignantly reinforces our AMD investment
thesis.

Underlying AMD's complaint is management's belief that AMD "should" have or
"ought" to have achieved greater market share than it has, especially as it
has released its architecturally competitive Opteron and Athlon 64 processors
over the past two years. Stepping aside from the moral imperative of the word
"ought", what is made clear by AMD's filing of the complaint and its related
anti-Intel marketing campaign is that AMD has not achieved the market share
position that it had expected or planned to achieve - or the market share
position for which it has planned capacity. It is exactly because AMD has
failed to appreciably gain share against Intel during this time that so
poignantly reinforces our thesis.

Having watched the processor market over the last couple of decades, it is our
observation that the highest performance processor does not necessarily "win".
Performance is a "necessary but not sufficient" condition to market share
success. Other factors include ease of adoption, integration with other
systems elements, field support, ability to deliver volume, other support
features that impact total systems performance and customer's total cost of
ownership " etc, etc. just to name a few. The RISC vs. CISC architecture
battles of the early 1990s clearly showed that processor performance was not
the determinant of market success.

AMD's airing of alleged Intel business practices bolsters our belief that
Intel is an aggressive tough competitor. AMD's objective in the complaint and
the related anti-Intel marketing campaign is to vilify Intel and paint what
may be aggressive marketing as "illegal". As this issue is not likely to be
resolved outside the judicial process and any resolution to AMD's complaint is
likely to occur beyond our investment horizon, it would be premature and
inappropriate to pontificate upon possible outcomes.

We believe it is likely that the need to fill Fab 36 factored into AMD's
decision as to the current timing to launch its complaint.

From AMD's point of view, its inability to gain share is more than academic.
We believe AMD has committed its capacity plans for its 300mm Fab 36 (and its
related fixed cost) based on a targeted market share that it is now unlikely
to achieve. AMD is currently in the process of ramping Fab 36, which is
capable of 50% more 8-inch equivalent wafer-outs vs, its current 200mm Fab 30.
In addition as Intel accelerates its dual core 65nm processor roadmap AMD
will be highly motivated to move production to Fab 36 which will be AMD's only
300mm 65nm production. The current Fab 30 is 200mm and is not capable of
running AMD's 65nm SOI process. AMD will of course have some flexibility
regarding the level to which it equips Fab 36, but there is some production
level below which the fixed cost burden of the new Fab 36 will be a drag on
margins.

Given this looming pronounced step-up in capacity, we believe AMD has a
heightened sense of urgency to gain share. We believe it is likely that the
need to fill Fab 36 factored into AMD's decision as to the current timing to
launch its complaint.

Our AMD investment thesis has been grounded in our belief that 1) AMD's market
share gains in servers and desktops are likely to be modest in '0506 and 2)
AMD is likely to continue to lose share in the rapidly growing notebook
segment, where Intel holds a commanding lead.

Our expectations for AMD going forward

Our AMD investment thesis is grounded in our belief that:

AMD's market share gains in servers and desktop processors are likely to be
modest in 2005-06

AMD is likely to continue to lose share in the rapidly growing notebook
processor segment, where Intel holds a commanding technology and branding lead
with its Centrino line.

Base assumptions

The base assumptions underlying our thesis, which we review in more detail in
a section below, include:

Intel's lead in notebooks widens with the upcoming Centrino/ Napa and Santa
Rosa platforms to be released successively over the next 18 months.

We do not believe AMD's relative success in the corporate market with its
Opteron-based server platform will easily carry over to corporate desktops.
First Call Corporation, a Thomson Financial company.
All rights reserved. 800.347.7822


Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 2 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005


AMD will not be able to materially increase its corporate desktop market
share. We do not believe AMD's relative success in the corporate market with
its Opteron-based server platform will easily carry over to corporate
desktops. We base this on our expectation that IT managers will stay with
Intel given its focus on enabling platform technologies (iAMT, security, etc.)
that help reduce IT total cost of ownership. Additionally, based on
discussions with AMD management, we do not expect AMD to attempt to grow its
consumer desktop market share much more as the company focuses an increasing
amount of attention on gaining share in the more profitable corporate market.

Intel is successful in establishing a new class of slim desktops based on
notebook platforms (low power processor and chipset). Our expectation for
Apple's use of the Napa notebook platform in its mid06 Mac mini is a near-term
validation that this approach could achieve at least modest success. AMD is
also beginning to promote its similar "blade PC" concept, but the value
proposition is less clear given AMD's lack of a widely accepted notebook
platform from which to leverage this strategy.

Intel's price aggressiveness at the low-end of the server market maintains
Intel's low-end server market share.

Characteristically conservative corporate IT managers will be slow to deploy
Opteron based servers, making any share gains modest in 2005 and minimal in
2006 when Intel's introduces its own dual-core server processors. Based on
Intel delivering increasingly competitive performance, we expect AMD's gains
to reverse in 2H06.

Intel's 9-12 month wafer fab process technology lead is maintained.

A continued high-level of uncertainty regarding AMD's (wafer fab)
capacity/capital reuse strategy. We believe a capacity reuse strategy is
vital to maximize return on invested capital (ROIC) and help recoup AMD's
significant investments in leading-edge manufacturing technology necessary to
compete against Intel. The key issue here is the uncertainty surrounding
AMD's plans for Fab 30 when it ceases to be competitive for leading edge
processor production in 2006 when Intel ramps its dual core 65nm road map.

The foundation of Intel's business model is based on maintaining MPU share in
the 80-85% range.

The foundation of Intel's business model is based on maintaining its MPU share
- We believe that when Intel commits to its capacity plans, it bases its
assumptions on a processor unit market share in the 80-85% range, a level it
has historically maintained. The importance to Intel's of its near constant
market share was highlighted at its May annual stockholders' meeting, where
Intel showed a chart of its relatively flat ~80% share since 1998.

Architectural Advantages are Not Sustainable; Manufacturing Advantages Are

AMD's current advantage in processors is based on its architecture, which we
believe is not defensible in the long run.

This analysis is not to diminish AMD's capabilities, as no other company has
been able to sustain a material position in the x86 processor market against
Intel. Additionally, AMD has capitalized very well on Intel's weaknesses in
x86 server processors, where it had held an unchallenged position for a long
time. AMD's success with Opteron, which has pushed AMD's market share from
3.9% in 4Q03 to 6.5% in 1Q05 (flat with 4Q04), has been the first real
challenge to Intel's position in industry standard servers.

However, AMD's current advantage is based on its architecture, which we
believe is not defensible in the long run. Architectural techniques (on-chip
memory controller, etc.) and tradeoffs are well known, and we believe over the
next 18-24 months, Intel will produce an architecturally capable equivalent to
Opteron.

This belief in architectural parity has been affirmed by recent reports we
highlighted in a recent SemiBytes piece that Intel's 2007 high-end server
processors will have an on-chip memory controller with a common serial
interconnect (CSI) interface between the processor and memory as well as other
processors - thus neutralizing key architectural advantages of AMD's Opteron.

This on-chip memory controller addresses a key Intel shortcoming by removing
the chipset as a bandwidth bottleneck. Based on Intel achieving architectural
parity, coupled with its platform strategy and manufacturing advantage, we
believe projections for AMD gaining greater than 10-15% unit share of x86
server market or greater than 20% of overall x86 processor market on a
sustained basis are untenable.

Intel's most defensible advantages are its tremendous manufacturing scale and
leading process technology.

Intel's most defensible advantages are its tremendous manufacturing scale and
leading process technology. To highlight this, we juxtapose AMD's and Intel's
respective road maps, focusing on the respective process nodes for each
processor. Compared with Intel's road map, AMD is behind nearly a half
process node (912 months) against Intel with respect to volume production.

Compelling Analogy - Pentium M/Centrino Platform

To reinforce our case that AMD's ability to gain share beyond our estimates is
not likely, we use the compelling analogy of the Pentium M notebook processor
backed by the Centrino platform brand. Intel has been able to leverage its
segment leading architecture and process technology advantage to gain market
share in notebooks (shown below in the next section) since the introduction of
the platform in 1Q03.

We believe Intel's Centrino share gain success should be the benchmark to
judge if AMD can achieve sustainable share gains in desktops and servers.

We believe Intel's Centrino share gain success should be the benchmark to
judge if AMD can achieve sustainable share gains in desktops and servers going
forward. Intel is in the process of executing a platform strategy for
desktops and servers that we expect to mirror Centrino with: 1) a processor
architecture based on the performance leading Pentium M, 2) systems level or
platform features that are enabled by proprietary IP embedded in the
supporting chipset and 3) an overarching branding and marketing campaign. As
a result, we expect AMD's desktop and server share gains to stop in 1H06 and
actually reverse beginning in 2H06 (although our estimates show AMD gains a
slight amount of desktop and server market share for the full year 2006 over
2005).

AMD Market Share Scenario Analysis

We are very aware that there is a vocal investor group, which advocates that
AMD's advantages are sustainable and that it is "different this time". Given
this chorus, we ask the question, "What if AMD exceeds our market share
assumptions?"

Sensitivity Analysis - Methodology

In order to perform a sensitivity analysis and test market share assumptions
greater than our estimates, we have built a model that:

Uses our latest AMD and Intel processor forecasts as the baseline.

Revises our AMD estimates based on its unit share gains across its product
line. We assume a fixed market size, so gains for AMD are reflected as Intel
losses in a comparable category, which together are reflected in our market
share calculations. We make the assumption that ASPs do not change given the
relatively small changes in market share. We believe this is a valid
assumption based on recent past history for stable processor price trends.

In general, we believe pricing will be stable based on: 1) Expectations for a
stable macro environment through 2006 driving sustained end market demand and
2) AMD's continued focus on driving higher-end, higher margin products (e.g.
dual-core Opteron/Athlon 64 and Turion).

Adjusts gross margin based on product mix.

Baseline Model - Updates to AMD and Intel

AMD

We have updated our AMD model with the following key changes:

More aggressive ramp of dual-core products, with more favorable ASP
assumptions, especially in servers, where dual-core processors are priced
about 2X of their single core counterparts based on the AMD price list.
Specifically in servers, AMD commented at its recent June analyst meeting that
customers generally gravitate toward the higher end SKUs as they become
available, which would infer a rapid migration to dual-core Opteron
processors.

Declining gross margin trend in 2006 based on our expectation Intel will
leverage its 65nm manufacturing advantage to more aggressively price its
leading-edge products. However, the net effect for the year on gross margin
is small vs our prior model, and the trend is likely to flatten into 2007 as
AMD ramps its 65nm products in late '06.

A lower gross margin for the Flash business (of 15% vs the prior estimate of
15-20% in '06) and operating margin (of 5% vs the prior estimate of 5%0% in
'06). We base these lower estimates on the continued lack of evidence that
Spansion 1) is gaining broad acceptance for its two-bit per cell, MirrorBit
technology 2) can transition MirrorBit products to 90nm process technology in
2H05 and ramp volume. As we detail below, we believe there is material risk
that the Spansion IPO can be executed and thus continue to include Spansion
results consolidated under AMD through 2006.

Intel

We have updated our processor estimates for Intel based on Intel's 1Q05 10-Q,
which separates desktop/server processors from mobile. These unit changes are
minor and do not change any of our revenue or EPS estimates. However, our
current Intel estimates may prove to be too conservative, as we are estimating
sequential rates of growth in 3Q and 4Q05 that are below the last 5-year
average.

Baseline Model - AMD Market Share Expectations

Based on these changes, our expectations for AMD's unit market share for
servers, desktops and notebooks are as follows:

Table 1: AMD Baseline Scenario

Source: Mercury Research, UBS estimates; market share estimates include Intel
and AMD only
First Call Corporation, a Thomson Financial company.
All rights reserved. 800.347.7822


Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 3 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005


Although AMD publicly has stated an x86 server processor market share goal of
30% by the end of 2006, we believe this could happen only if Intel were to
fail to execute on its dual-core server processors. We view this as highly
unlikely.

More Optimistic Scenarios - AMD beats our current market share projections

We now test our baseline estimates against two more optimistic scenarios:

"Sunny Day" Scenario

Our first test is based on a "sunny day" scenario, i.e. one that could happen
under highly favorable market conditions. The biggest difference between this
and our baseline scenario is an assumption that AMD's server processor market
share grows to 15% in 2006 vs our baseline estimate of 10%. Although AMD has
publicly stated a market share goal of 30% by the end of 2006, we believe this
could happen only if Intel were to fail to deliver its dual-core server
processors. We view this as highly unlikely, as Intel is already shipping
dual-core server platform prototypes to OEMs and developers.

Sunny Day Scenario - Key Assumptions:

Server: AMD continues to grow its server market share, leveraging its current
dual-core advantage. Hewlett Packard, which is among AMD's largest Opteron
customers, views an Opteron-based server customer to be less likely to move to
archrival Dell according to industry sources; this creates a powerful
incentive for HP to promote its Opteron-based offering. Given HP's sales
channels and breadth of coverage, we view HP to be the most valuable server
OEM for AMD, and we may be underestimating HP's potential success with its
Opteron-based offering.

Desktop: AMD grows desktop market share more than our assumptions, as it ramps
its Athlon 64 through 2005 and dual-core Athlon 64 X2 into 2006. AMD desktop
processors continue to perform well in independent benchmark tests against
comparable Intel processors, and this could translate to larger share gains
than we estimate.

Notebook: AMD holds its notebook market share at about 10% through 2006. We
believe this could happen if 1) notebook growth rates continue to accelerate
beyond Intel's ability to keep up, 2) Intel chip shortages continue through
2005 and 3) Intel slows or delays the ramp of its 65nm dual-core Yonah
notebook processor at the end of 2005.

Sunny Day Scenario - Results

Table 2: AMD Sunny Day Scenario

Source: Mercury Research, UBS estimates; market share estimates include Intel
and AMD only

We believe stocks other than AMD in our coverage universe that are currently
trading at about 20x our 2006 EPS are more attractive with lower risk; these
include Altera (ALTR, Buy 2), Analog Devices (ADI, Buy 2 RRD) and Maxim (MXIM,
Buy 2). Even more attractive are Intel (INTC, Buy 2) at 17x and Texas
Instruments (TXN, Buy 2) at 18x.

In our sunny-day scenario, a 20x PE multiple applied to our 2006 EPS is
required to support AMD's current share price. We believe other stocks in our
coverage universe that are currently trading at about 20x our 2006 EPS
estimates are more attractive with lower risk: these include Altera (ALTR, Buy
2), Analog Devices (ADI, Buy 2 RRD) and Maxim (MXIM, Buy 2). Even more
attractive are Intel (INTC, Buy 2) at 17x and Texas Instruments (TXN, Buy 2)
at 18x.

An Even More Optimistic Scenario - Making a Buck

We can also view this analysis from a different perspective - For a given EPS
expectation, what market share level would AMD need to achieve? Specifically,
we are interested in what it takes to generate $1.00 EPS estimate in 2006. We
use $1.00 since given our recent target PE multiple of 18x, it is this EPS
level that would justify the stock to be upgraded from a Reduce 2 to a Neutral
2 rating under the UBS rating system.

Although we can generate an infinite number of scenarios to generate $1.00
EPS, we simply multiply the added AMD unit gains used in the sunny-day
scenario by a common factor (approximately 1.6x) that results in a $1.00 EPS
estimate for AMD in 2006.

Table 3: AMD "$1.00 EPS in 2006" Scenario

Source: Mercury Research, UBS estimates; market share estimates include Intel
and AMD only

While we do not believe these market share numbers are achievable, we believe
this analysis sheds light on what AMD would have to achieve to justify our
fair value.

We believe these optimistic AMD scenarios are not achievable.

Why We Believe our Optimistic AMD Scenarios are Not Achievable

At this point, we dig deeper into our baseline model assumptions summarized
above to better understand why we believe our optimistic scenarios are not
achievable:

Intel's notebook lead is likely to increase

In the rapidly growing notebook segment, we cannot see how AMD can leapfrog
Intel.

We believe Intel's lead in notebooks is commanding and likely to increase.
Intel is currently sampling its third generation Centrino platform called
Napa, which is based on the 65nm dual-core Yonah processor. Unless AMD
delivers a better platform than Napa in the next 6 months, we expect Intel to
gain more notebook market share in 2006.

We expect Intel's lead to increase further with Intel's follow-on Santa Rosa
platform, which is based on the 64-bit Merom processor. Santa Rosa, expected
to launch 1Q07, will likely have increased integrated graphics performance
(shipping in about two-thirds of Centrino systems today) and add
second-generation Active Management Technology (iAMT2), LaGrande technology
(LT) for hardware-based security, WiMAX, UWB and seamless synchronization with
advanced cell phones and handhelds. Based on Intel combining all of these
complex technologies with the powerful Centrino brand, we cannot see how AMD
would be able to leapfrog Intel. While AMD has belatedly responded to
Centrino recently with its own Turion brand, tepid benchmarks and only limited
OEM support for the first Turion notebooks are not encouraging.

We do not expect AMD to be able to materially grow its share in corporate
desktops

AMD expects to be able to translate its relative success in the corporate
market with its Opteron-based server platforms into increased desktop sales.
We believe Intel's integrated platform technologies, including Active
Management Technology (iAMT), virtualization technology (VT) and its road map
for future versions and additions (security, networking I/O acceleration), are
compelling for IT managers. We believe a demonstrated ability for these Intel
technologies to lower costs while increasing manageability and security is
likely to help Intel hold its dominant position in the corporate market.

Based on a recent meeting with AMD Chairman, President and CEO, Hector Ruiz,
we learned AMD is not likely to aggressively pursue more consumer desktop
market share where it has the strongest position against Intel. A primary
focus for AMD is increased ASPs and levels of profitability. This is
supported by AMD's recent track record, as the company has shown a trend
towards increased blended processor ASPs since mid-2002.

We expect a new Intel based "mini" desktop class to become material

Among the categories we believe are most likely to grow the dollar value of
Intel silicon and promote growth in the desktop market is a new class of
desktops based on notebook platforms. These desktop "mini" platforms (similar
to AAPL's Mac mini) use the low power characteristics of the notebook
processor/chipset to create a small form factor fanless desktop.

Beyond the Mac mini design win in the consumer market, we believe there is a
more lucrative corporate market opportunity based on the potential for OEMs to
introduce a line of corporate desktop "minis" based on the notebook platform.

This class of PC would enable IT managers to lower ownership costs by:

Consuming less power - we estimate $300-$350 in electricity savings over the
4-year life of a desktop PC

Simplifying software maintenance by being able to standardize on a common
software image for both desktop and notebook machines.

While AMD recently stated its desire to enter this market with its "blade PC"
concept at its June analyst meeting, we believe that Intel's ability to
leverage its leading notebook platforms provides an important advantage.

We expect Intel to be very price aggressive at the low-end of the server
market and to maintain market share

Our discussions with Intel management confirm a more challenging pricing
environment recently in the server processor market. We believe Intel is
being the most price aggressive in the higher volume 1-socket and 2-socket
systems.

Evidence of this price aggressiveness shows in $/tpmC (price per number of
transactions per minute using the latest TPC-C benchmark) data on the tpcc.org
web site, an independent benchmark organization that measure on-line
transaction processing (OLTP) performance. (OLTP is a standard performance
metric for how fast database applications run on servers.) A lower $/tpmC
number indicates more processing power for the buck.

To summarize the results:

The latest 1/2-socket Xeon servers have $/tpmC numbers under 2.0, which are
the lowest tested with the latest revision of the TPC-C benchmark.

The AMD Opteron shines in the $/tpmC data is in HP's higher-end 4socket
systems, where AMD's Opteron has $/tpmC in the range of 2.04 based on the
latest dual-core Opteron up to 2.94 for single-core Opterons.
First Call Corporation, a Thomson Financial company.
All rights reserved. 800.347.7822


Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 4 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005


While Dell and HP have recently introduced systems based on Intel's latest
4-socket Xeon MP-based Truland platform, there are no TPC-C benchmarks
published yet for these new systems. IBM was among the first to market with
its new 4-way+ xSeries MP server based on the new Xeon MP processor and IBM's
proprietary (and expensive) chipset. The TPC-C benchmarks are impressive but
the $/tpmC values are in the range of $6.

As a result of Intel's price aggressiveness, we expect Intel to largely
maintain its market share in the volume portion of the business. Given this
assumption, we expect AMD's server market share to top at about 10-11% in 2005
going into early 2006.

Slow rate of IT change = Modest Opteron-based server gains

We believe that beyond the scientific community and specialized departments
within enterprises, corporate IT managers based on their conservative nature
will be slow to broadly deploy Opteron-based servers. We believe this issue
makes it likely that AMD's server share gains will be modest in 2005 and
minimal in 2006, when Intel's introduces its dual-core server processors.

We expect AMD's share trend to stall later in 2006 as Intel ramps its
dual-core 65nm server processors (Dempsey, Tulsa) and to reverse in 2007 when
we expect Intel to deliver superior performance based on:

Common Serial Interconnect (CSI) interface - CSI enables a direct connection
(i.e. no intervening chipset) from the processor to memory, other processors
and peripherals. We believe that among Intel's first processors with CSI are
its 4core 65nm Whitefield and the follow-on Dunnington. However, AMD is
likely to deliver similar capabilities as CSI with faster versions of its
current on-chip memory controller and Direct Connect Architecture.

Large (i.e. 16MB+) shared caches - Large caches increase performance by
keeping more of the likely-to-be-used data from memory within the processor
and thus reducing the need to go off-chip. Because of Intel's ability to get
consistently good yields on big chips in the most advanced processes, it can
economically manufacture processors with large caches. A 16-megabyte cache is
roughly a billion transistors, which Intel has shown it can already make in
90nm technology; Intel's 90nm Montecito Itanium 2 is 1.7B transistors. The
ability to share the cache among multiple processor cores is an optimization
that enables the cache to be dynamically partitioned among the processor cores
based on asymmetrical workloads, enabling even better overall performance.

We believe Intel's ability to retake market share in servers to be sustained
beyond 2007, when we expect Intel to ship its first 45nm processors based on
its Penryn architecture.

Tightly integrated chipsets - Intel's platform strategy should enable it to
deliver value-added features in its chipsets that better optimize performance
at a system-level. Through AMD's chipset partnerships, most notably Broadcom
for servers, we expect AMD to deliver similar chipset-based features as Intel.
However, because Intel designs both the processor and chipset silicon within
the same organization, we believe Intel's execution risk is lower. In the
longer-run, we would expect Intel to gain a time-to-market advantage, as we
believe Intel will increasingly drive the market in setting platform-level
standards.

We believe Intel's ability to retake share in servers to be sustained beyond
2007, when we expect Intel to ship its first 45nm processors based on its
Penryn architecture.

AMD's window for 90nm process parity is closing

Based on our belief that Intel will continue to hold a 9-12 month wafer fab
process technology lead, we believe it will be difficult for AMD to sustain
market share gains and to increase gross margins.

Based on our belief that Intel will continue to hold a 9-12 month wafer fab
process technology lead, we believe it will be difficult for AMD to sustain
market share gains or to increase gross margins. We expect Intel to begin
shipping 65nm product in late 4Q05 with a near complete conversion to 65nm by
4Q06. For most of 2006, we expect Intel to be shipping an increasing
percentage of 65nm processors, while we expect AMD to only begin ramping 65nm
volume in late 2006.

Because 65nm technology enables more performance for a given amount of silicon
vs 90nm, Intel gains a distinct advantage to flexibly drive either higher
gross margins or market share gains through more aggressive pricing. We
expect more of the latter and are currently modeling for Intel to gain only 30
bpts of gross margin in 2006 vs 2005.

What happens to AMD's Fab 30 after 2007?

AMD's Fab 30 not having a post-2007 defined role raises the level of risk and
uncertainty surrounding AMD's manufacturing model.

As AMD plans to transition to its 300mm Fab 36 concurrently with its
transition to 65nm process technology, it is not clear how AMD is going to
utilize its 200mm 90nm Fab 30 during this transition in 2006 or after 2007.
We believe not having a post-2007 defined role for Fab 30 raises the level of
risk and uncertainty surrounding AMD's manufacturing model. In this respect,
AMD is at a disadvantage to Intel, which has a well-refined capacity reuse
strategy that supports the manufacturing of chipsets, communications chips and
Flash memory. Intel's ability to continue the productive utilization of fab
capacity, even as it moves its processor production to newer leading edge
capacity, adds significantly to Intel's long-term ROIC. Ultimately, AMD's
inability to do the same will make it harder for AMD to realize competitive
returns on invested capital.

Intel's model is based on maintaining its MPU share

We believe Intel makes capacity decisions based on its ability to maintain a
processor unit market share in the 80-85% range.

Why do we believe this?

One of the major determinants of the success of Intel's business model is its
ability to match the timing of its investments in process technology and fab
capacity with the market transitions to the products supported by these
investments. It is important that market acceptance of new processors and
chipsets roughly match the timing of Intel's investments in the underlying
process technology and fab capacity. Intel's difficulties in 2004 are
evidence of the consequences of misjudgments and lack of coordination between
product launch schedules (Prescott/Grantsdale) and process transitions
(300mm/90nm).

It is our opinion, having personally watched these two companies compete for
the last 25 years, Intel believes that it has the opportunity to
influence/lead the PC industry in these product transitions if it has roughly
an 80-85% market share. If Intel's market share were to get above 85%, we
believe Intel risks its ability to maintain or grow gross margins as it
attempts to win every piece of business. We believe Intel recognizes that a
strategy to push market share above 85% faces a diminishing return scenario.
Below 80% market share, we believe Intel risks undermining its leadership
market position and its ability to influence the direction and timing of the
PC/server market in critical product transitions, which could result in
underutilization of its factories, below-average industry growth and margin
compression.

In addition, given its dominant market position, Intel maintains a tremendous
responsibility to the entire IT industry. Hundreds of billions of dollars of
IT systems revenue are based on Intel's ability to ensure an adequate supply
of its chips. Conflicting with this responsibility are Intel's ROIC goals,
which Intel highlighted at its recent stockholders' meeting. Maximizing ROIC
implies minimizing volatility, even in a rapidly changing demand environment.
To help buffer against these demand shifts, Intel needs a viable competitor,
which can absorb some of this volatility, a role that we feel AMD fills well
in the current PC food chain.

Spansion IPO Risk is Material

Up to this point, we have focused on the risks surrounding AMD's processor
business. However, what compounds these risks is our belief there is material
risk that the initial public offering of AMD's 60%-owned Spansion subsidiary
will be delayed or postponed.

We feel any valuation of AMD that assumes a successful IPO of Spansion is
risky.

Spansion LLC, the 60% AMD/ 40% Fujitsu owned NOR Flash joint venture, has
filed an S1-A for a proposed IPO. AMD's intent is for Spansion to sell stock
to the public and raise equity capital to meet its own cash needs and in the
process dilute AMD's position below the 50% level that requires AMD to
consolidate Spansion's money losing results in AMD's financial statements.
While theoretically AMD would still have to report Spansion's results as a
minority interest line on its income statement, we would expect that if the
Spansion IPO is successful, AMD would quickly move to spin or sell its
remaining share as quickly as is reasonable or possible. As we believe that
AMD's tight cash flow position relative to the combined cash needs of the
processor and Flash businesses is the primary impetus for the Spansion IPO,
any delay or a postponement of the IPO into 2006 is a significant issue for
AMD. We feel any valuation of AMD that assumes a successful IPO of Spansion
is risky.

Factors contributing to this risk are:

Spansion's market opportunity is likely to remain clouded, as we do not expect
the current unfavorable competitive environment to markedly improve in the
near term

The market environment - Gartner forecasts the NOR Flash market to drop 6% in
2005 from 2004's $8.8B figure, rising slightly to $9.1B in 2006. We believe
these relatively flat numbers reflect the competitive NOR Flash environment
and slower growth of end market applications. The largest NOR Flash
application is the wireless handset market, which UBS estimates to grow 13% in
2005 on a unit sell-in basis vs 28% in 2004. Spansion's market opportunity is
likely to remain clouded, as we do not expect the current unfavorable
competitive environment to markedly improve in the near-term.

Intel - We believe Intel's recent execution and share gains are sustainable
given its 1) Aggressive Flash technology road map to 90nm, 65nm and beyond 2)
Redesigning its devices to meet unserved application needs, most notably
Qualcomm's CDMA handset chipsets 3) Expansion into embedded markets it
previously underserved by broadening its product portfolio 4) Expansion into
new market areas including low-density USB Flash drives where it believes it
can compete on cost against Samsung's low density NAND Flash devices.

Given what is likely to be a relentless market expansion effort by Intel,
which still has excess Flash fab capacity, we would be cautious to assume any
near-term turnaround scenario for Spansion.



Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 5 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005


Spansion's technology competitiveness - Given what we believe is a closing
window for Spansion to demonstrate that its technology is competitive with
Intel, we believe that by the September quarter, Spansion must show it can:

Gain material market acceptance for its two-bit-per-cell MirrorBit-based
devices against Intel's established StrataFlash equivalent. At the end of the
March quarter MirrorBit accounted for roughly 15% of Spansion shipments vs
StrataFlash, which is over 85% of Intel's Flash shipments, giving Intel a
significant cost/bit advantage. Evidence of MirrorBit acceptance should
exhibit itself in flat or increasing ASP trends, as a ramp of higher density
MirrorBit-based devices should compensate for anticipated per bit price
declines.

Spansion highlights in its S-1 filed in April and revised S-1A in June that
"market acceptance of products based on our MirrorBit technology is a critical
factor impacting our ability to increase revenues and market share." To date,
we do not see compelling evidence that AMD is substantially increasing its
MirrorBit shipments.

According to an April Gartner report, "AMD has been struggling to gain
widespread acceptance for its two-bits-per-cell MirrorBit technology""

To reinforce this, according to a recent market research report on the NOR
market: "AMD has been struggling to gain widespread acceptance for its
two-bits-per-cell MirrorBit technology and to increase production levels to
high volumes. Whereas Intel's StrataFlash technology is seen as a robust,
proven technology, the market has doubts about nitride-based Flash memory
technologies like MirrorBit. In particular, there is the issue of
program/erase endurance limitations and the concern that the memory cell
structure may prove hard to scale. The proposed sale of Spansion must raise
the suspicion that AMD is unwilling to expose itself to the spiraling
investment costs that may be needed to attain the various milestones along the
MirrorBit road map." (Gartner, 4/05)

iSuppli further reinforces the urgency for AMD to execute on MirrorBit: ""As
2.5 and 3G mobile phones claim a larger portion of handset shipments,
(MirrorBit) two-bit-per-cell NOR will be the only viable option."

In another market research report, iSuppli further reinforces the urgency for
AMD to execute on MirrorBit: "Intel will make it tough on Spansion and
STMicroelectronics. But without the availability of two-bit-per-cell
(MirrorBit) NOR devices, the pain would become excruciating...As 2.5 and 3G
mobile phones claim a larger portion of handset shipments, two-bit-per-cell
NOR will be the only viable option." (iSuppli, 6/05)

Show the ability to successfully manufacture and ramp MirrorBit in 90nm.
Spansion recently reinforced its density leadership at 512Mb with an
announcement of a design win with French handset maker, SAGEM. However, this
win was based on a 110nm technology device into a high-end Windows Smartphone,
which we would not expect to ship in high volume. To date, we have seen no
evidence that Spansion is ready to ramp its 90nm MirrorBit technology. Intel
however, has recently stated that it is sampling its 90nm 512Mb StrataFlash
device and is on schedule to ramp high volumes in 2H05. Thus, even if
Spansion makes a transition to 90nm MirrorBit in 2H05 (with its 512Mb device),
it is likely to face stiff competition.

Spansion's big capex step-up in 2006 - We believe another key issue for
Spansion is the huge step-up in capital spending that expected beyond 2005.
In its S-1A, Spansion states that its planned capital expenditures for fiscal
2005 are approximately $500 million (raised from $450M in its original S-1 two
months prior). We expect the cost of a new 300mm Flash facility, which
Spansion expects to begin production in 2007 according to the S-1A, will cost
at least $2B, leading us to believe that capex would likely to at least double
in 2006. To remain competitive beyond 2006, we believe high capex levels
would need to be sustained, especially as the average age of Spansion's three
older fabs is about 10 years:

Table 4: Spansion Wafer Fabrication Facilities

* Capacity refers to the capacity of currently installed equipment

** Fab 25 was originally a logic fab and converted to a Flash fab in 2002

Source: Company reports, Gartner

Spansion's debt servicing requirements - Spansion has a total of $865M of
contractual cash obligations over the next 5 years, with $305M due in 2005 and
most of the balance due in 2006. In its S-1A Spansion outlines that it plans
to concurrently issue debt to repay a portion of its outstanding borrowings.
We anticipate a debt restructuring that effectively extends the terms of its
existing debt while providing additional capital to support its 300mm wafer
fab construction. Given what we believe will be a challenging environment to
generate positive free cash given the anticipated capex step-up from 2005,
higher levels of debt would raise Spansion's risk profile and thus its cost of
equity as well.

Spansion's renegotiation of its term loan agreement covenants as disclosed in
its recent 8-K filing only reinforces this risk. Most notably, Spansion
reduced or eliminated certain cash balance holding requirements. We believe
Spansion would only rewrite its covenants if it had to, especially as it is
preparing for an IPO, and this could be a potential leading indicator of more
difficulties given the competitive NOR Flash market. This 8-K is consistent
with our belief that Spansion and AMD are both facing issues of generating
sufficient cash from operations to support capital spending investments.

Spansion's limitations on corporate restructuring - According to Spansion's
S-1A, the company has limited operational flexibility regarding its Japanese
fabs and significant restrictions to:

Shut down operations at JV3 or any facilities of the non-manufacturing
organization of Spansion Japan before April 1, 2007

Shut down operations at JV1 or JV2 before April 1, 2007 without giving AMD and
Fujitsu six months advance notice

Reduce headcount at JV1, JV2, JV3 or any facilities of the non-manufacturing
organization of Spansion Japan prior to April 1, 2007 without giving AMD and
Fujitsu six months advance notice

While we believe that both AMD and Fujitsu would consent to corporate
restructuring necessary to ensure Spansion remaining a viable entity, we
believe that likely delays in gaining approvals raise the level of execution
risk for any potential restructuring required in the future.

In the extreme case, we believe Intel's need to satisfy its capital reuse
goals could indefinitely drive out profitability from the entire NOR market.
This scenario would be devastating for Spansion.

Spansion's manufacturing model must be competitive against Intel's
fundamentally different Flash business priorities - Spansion's manufacturing
model is based on building dedicated Flash memory fabs. This contrasts
sharply with Intel, which uses largely depreciated microprocessor fabs and
leverages the R&D used to develop the logic processes used to build
microprocessors. While we believe Intel's fundamental goal is make money in
Flash, we believe that its larger goal of maximizing ROIC (with Flash as piece
of Intel's capital reuse strategy) is the priority. Given that Intel
currently has excess Flash fab capacity, we expect Intel will continue to
focus on market share expansion in order to optimally use its fab resources.

In the extreme case, we believe Intel's need to satisfy its capital reuse
goals could indefinitely drive out profitability from the entire NOR market
(or at least until some major suppliers exit). While Intel may continue to
incur an operating loss in its Flash business in this case, we would expect
this to have a small effect on Intel as a whole. However, this scenario would
be devastating for Spansion, which must generate enough cash to finance its
capital spending plans just to remain competitive.

Given these risks, our current assumption of Spansion being a near-breakeven
business in 2006 may be too optimistic. Even Intel, which bases its Flash
manufacturing in largely depreciated microprocessors fabs, may be challenged
to earn an operating profit in Flash in 2006.

AMD Cash Flow Concerns

We believe AMD's Statement of Cash Flows requires particular attention, as we
do not expect AMD to generate positive free cash in 2005, even when factoring
in the German subsidies. The company has recently indicated that its capital
spending requirements to build and fully facilitate its 300mm Fab 36 will
likely be less than the $2.5B originally forecasted based in part from more
favorable exchange rates. However, we believe this still only partially
offsets the shortfall from AMD's original +$100M free cash flow ($100M plus
$200M in subsidies/grants) the company forecasted at its November 2004 analyst
meeting.

In 2005, only in the December quarter we believe AMD can generate positive
free cash without subsidies.

Using our baseline forecast, we estimate that AMD will generate about $1.0B in
cash flow from operations in 2005 vs management's current plan to spend $1.5B
in capital expenditures. Even if we lower our capex expectations from $1.5B
to $1.4B, this still ends up as -$200M in free cash (including the $200M in
German subsidies/grants). While forecasting cash flow is difficult, we are
concerned AMD will have difficulty to even meet its revised net -$100M free
cash goal set in January. We base this on: 1) AMD's guidance for lower cash
flow from operations in the June quarter vs the March quarter and 2) the
required holiday season inventory build in the September quarter as a
significant use of cash. Only in the December quarter, we believe can AMD
generate positive free cash (without subsidies) as inventory is likely to be
depleted and net income is maximized.

Our cash flow concerns linger beyond 2005: As we cited above, we believe Intel
will leverage its 65nm manufacturing advantage in 2006 to more aggressively
price its leading edge products. As a result, AMD is likely to experience
increasing margin pressure through 2006, continuing to raise cash flow
concerns.

It is our opinion that the proposed IPO of Spansion is related to AMD's tight
cash flow position. Under the current structure, AMD is obligated to meet its
60% share of any Spansion cash and capex budget shortfall. Under the terms of
the proposed IPO (S1-A) following the IPO AMD would no longer have any
responsibility for Spansion's cash or capex shortfalls. Looking out into
2006, we believe AMD would be hard-pressed to cover the capex requirements of
both the processor and Flash businesses as the competitors push toward finer
line widths and larger wafers.

Footnote: Stock Option Expenses

Even if AMD's stock option expense were lowered to $15M from $31M net of
tax/quarter in 2006, this would still reduce our estimated 2006 net income of
$259M by $60M or 23%.




Advanced Micro Dev. "What If AMD Beats...?" (Reduce 2) Thor (Pt 6 of 8)
Source: UBS (US)
Author: Thomas A. Thornhill III, CFA
Date: 06/30/2005


Over the past 12 quarters, AMD's reported stock options expenses per FAS 123
have averaged $31M/quarter net of tax. In April, AMD accelerated vesting of
approximately 12 million options (with an exercise price above $14.51) out of
a total of 21.8M unexercisable options (as of the end of 1Q05). As a result
of this accelerated vesting, stock option expenses should be lower in 2006
when FAS 123R takes effect for AMD. However, even if AMD's stock option
expense were lowered to $15M net of tax/quarter in 2006, this would still
reduce our estimated 2006 net income of $259M by $60M or 23%.

Valuation: Maintain Target Price $11, Reduce 2 Rating

Our 12-month target price of $11 is based on applying our 18x target PE
multiple to our revised 2006 EPS estimate of $0.61. Given our belief that
AMD's challenges create significant risks not factored in its current stock
valuation, we maintain our Reduce 2 rating.

HIGHLIGHTS & PROFITABILITY/VALUATION
------------------------------------------------------------------------------
(US$Millions) 12/03 12/04 12/05E 12/06E 12/07E
------------------------------------------------------------------------------
Revenues 3,519 5,001 5,298 5,735 -
EBIT (239) 227 (59) 296 -
Net income (UBS) (280) 146 5 194 -
EPS (UBS, US$) (0.83) 0.39 0.02 0.61 -
Net DPS (UBS, US$) 0 0 0 0 -
------------------------------------------------------------------------------
5-yr hist.av. 12/04 12/05E 12/06E 12/07E
------------------------------------------------------------------------------
EBIT margin % -2.6 4.5 -1.1 5.2 -
ROIC (EBIT) % -3.1 5.1 -1.3 6.1 -
EV/EBITDA x 8.9 4.1 5.8 4.4 -
PE (UBS) x 38.6 40.2 100 28.6 -
Dividend yield % 0 0 0 0 -
------------------------------------------------------------------------------
Source: Company accounts, Thomson Financial, UBS estimates. UBS adjusted EPS
is stated before goodwill-related charges and other adjustments for abnormal
and economic items at the analysts' judgement.
Valuations: based on an average share price that year, (E): based on a share
price of US$17.37 on 29 Jun 2005 19:37 EDT
For full version of note with charts and tables, please refer to:
http://www.ubs.com/investmentresearch

EPS (UBS, US$)
--------------------------------------------------
12/05 12/04
----------------------
From To Cons Actual
--------------------------------------------------
Q1 (0.04) (0.04) (0.04) 0.12
Q2E (0.07) (0.07) (0.06) 0.09
Q3E 0.01 0.05 0.03 0.12
Q4E 0.03 0.09 0.13 0.06
--------------------------------------------------
12/05E (0.08) 0.02 0.06
12/06E 0.49 0.61 0.59
--------------------------------------------------

Trading data
----------------------------------------------------------
52-wk. range US$24.85-10.86
Market cap. US$6.83bn
Shares o/s 393m
Free float 100%
Avg. daily volume ('000) 5,402
Avg. daily value (US$m) 87.5

Balance sheet data 12/05E
----------------------------------------------------------
Shareholders' equity US$3.34bn
P/BV (UBS) 2.1x
Net cash (debt) (US$0.35bn)

Forecast returns
----------------------------------------------------------
Forecast price appreciation -36.7%
Forecast dividend yield 0.0%
Forecast stock return -36.7%
Market return assumption 8.7%
Forecast excess return -45.4%