Here is a take on NAND vs. HDs:
needhamco.com
[EDIT: It's easier to read in the original PDF, but I put it here in case the original "goes missing," as they say. Warning: it is very long (20 pp., including boilerplate stuff), but it covers a lot of material. More favorable to HDs than to NAND.]
CHARLIE GLAVIN • CGLAVIN@NEEDHAMCO.COM • 415/262-4874 October 12, 2005
RICHARD KUGELE • RKUGELE@NEEDHAMCO.COM • 617/457-0908
NAND vs. Hard Disk Drives: Hype, Myth and Reality
Recent comments by Samsung and Hynix have hyped speculation that NAND Flash will soon displace
hard disk drives in all mobile devices. While the NAND market is expected to exhibit high unit/bit
growth due to the advantages of solid state memory in handhelds, to paraphrase Mark Twain, “the
rumors of HDD death are greatly exaggerated.” We believe investors can take advantage of such
misperceptions by buying “the arms dealer” stocks that would benefit from such heated competition,
while purchasing the much-maligned HDD stocks and their chip suppliers that have been oversold.
This report examines and compares the advantages of each media type, along with a supply/demand
analysis of the NAND market. Our conclusions are as follows:
• We believe the hype surrounding the battle between Flash and Hard Disk Drives (HDDs) has been
overblown and has clouded the true market opportunities for both industries. Each technology has
certain advantages/disadvantages that make it more suited for certain markets, and both should
experience significant unit growth in the coming years.
• Contrary to the myths surrounding the Flash vs. HDDs debate, our analysis suggests that both
technologies are on similar trajectories regarding cost and areal density improvements (due to
their common semiconductor manufacturing processes), making a “crossover” point unlikely over the
foreseeable future.
• We believe that the key market growth opportunities for Flash include segments where footprint,
ruggedness and power are paramount such as MP3 players (sub-10GB), digital cameras and
camcorders, and mainstream handsets (90-95% of the market).
• Hard drives likely to remain dominant in markets where cost/GB and throughput are essential,
such as PVRs, large capacity MP3 and video applications, high-end handsets/converged devices, and
mainstream laptops (with the possible exception of the niche UltraPortable market).
• From a market perspective, we believe that the transition at Apple on the Nano from a 1-inch drive to
Flash has created only a modest headwind for the HDD industry. Overall unit growth is still expected
to remain 15-20 over the next 2-3 years, with the only casualty from Flash’s increasing presence
being the nascent 0.85-inch drive (<1% of industry revenue), whose form factor prevents areal
densities that can effectively compete with Flash.
• Additionally, rather than look at the NAND vs. hard disk drive debate as an “either/or” scenario, we
believe “Hybrid” solutions are likely to emerge that that use increasing amounts of NAND memory
as a buffer for hard drives to increase performance and life.
• Overall, we believe that all of this competition between NAND and HDDs could be boon for digital
media chip makers that are fairly agnostic to the memory media type, so long as cheaper memory
enables more consumer-friendly applications (such as Silicon Motion).
• Semiconductor equipment stocks (covered by Needham Analyst Robert Maire) could also benefit as
our supply/demand model suggests the Flash industry will need to build just under two (1.8) new 300 mm
fabs next year to satisfy demand (and migrate quickly to 60 nm in 2006 and then 50 nm in 2007 to meet
current market commitment).
• Within the drive space, we believe Seagate, with the broadest product line, is best positioned for the rapid
adoption of hard drives in the home. Component suppliers Komag (with significant exposure to highcapacity,
high-platter count drives) and Hutchinson (suspensions) also benefit from the industry growth
rate.
2
Introduction
It’s Déjà vu All Over Again: Beware the Hype
While NAND unit growth is expected to outpace the overall market in the near term, we would caution against
accepting the hype that NAND will take over disk drive market. Similar prognostications were made in the
1990s when RISC processors would suppose to kill CISC processors such as the Pentium, or, in 1999 when
George Gilder said that optical fibre would kill cable and satellite options. Ironically, in 1992-94, similar hype
was made about NOR Flash displacing DRAM and disk drives in PCs, yet died down when the Apple Newton
(using Intel’s 12-volt, 8Mb Flash chips) failed to take off and IBM introduced the first microdrives.
The following quotes highlight how NAND is being (over-) heralded as the next “can’t miss” market:
• “There are so many new applications for NAND,” said Reza Faramarzi, marketing manager for
Hynix’ U.S. subsidiary. “There is an infinite demand for flash at the current price points.” EE
Times, 10/6/05.
• “The high margin of NAND will push capital spending on NAND.” IDC report, “Worldwide NAND
Flash Memory Demand and Supply 1Q05-4Q06 Forecast and Analysis,” #34037, Sept. 2005.
• "NAND flash will eventually replace other storage mediums, especially those used in mobile
products, creating a Flash Rush, as NAND continues to register an unprecedented surge in
demand as the backbone of the mobile electronics era." Dr. Chang Gyu Hwang, CEO of Samsung
Semiconductor.
• “Hard drives are living on borrowed time and will be replaced with solid-state Flash memory. From
Dr. Chang Gyu Hwang, CEO Samsung Semiconductor. Chris Mellor, Techworld, 9/13/05.
Such recent comments have created a press storm about how dominant Flash is and why the hard disk drive
market is doomed. While we agree that NAND flash is a great option and will be one of the dominant
form factors, we believe that the marcom (marketing communications) hype is way ahead of reality.
Each product has its own advantages, and we believe the market is more likely to see more co-existence and
hybrid designs over the remainder of this decade than one factor killing the other.
Our Key Conclusions:
1. NAND is not expected to achieve cost parity or density levels comparable with hard disk drive, limiting
some of the markets that it can penetrate (laptops).
2. However, current NAND roadmap will mean more fabs need to be built (roughly two per year) to keep
pace with demand, but more importantly, NAND fabs will need to be upgraded (to 60 and 50nm
processes) to achieve higher density cards.
3. NAND and Disk Drives likely to Co-exist in Hybrid designs due to the advantages of both media
types.
4. We believe that the only casualty of flash’s greater presence in consumer applications will be the
0.85-inch drive; while drive industry gets a boost from new smaller form factor 1-inch “Mikey” drives
from Hitachi and Seagate.
This report is split into two parts. In the first section, we examine and compare hard disk drives versus NAND
Flash drive alternatives. In the second half, we present a supply/demand outlook for the NAND Flash Market.
3
PART I: A View From the Front Lines of the Storage Wars—Flash vs. Drives
Summary
Investor focus in recent months on the broader penetration of flash-based storage into consumer electronics
applications has been steadily increasing. Apple’s recent announcement for the iPod Nano brought this issue
to the forefront, as the previous product (the 1-inch drive based iPod Mini) was replaced, at least temporarily.
Our view, which is outlined below, is that the truth lies somewhere in the middle, with each technology having
a distinct place within the storage requirements of consumers.
We believe that our analysis suggests the following conclusions:
• Rather than look at NAND and hard disk drives as an “either/or” scenario, it is our belief that they
will both co-exist and prosper as OEMs leverage the advantages of each technology. Look for
“Hybrid” solutions that will use increasing amounts of NAND memory as a buffer for hard drives to
increase performance and extend product life.
• Both hard drives and Flash have pros and cons that are unique and solve distinct problems for
consumer application designers.
• We believe the real battleground between drives and Flash appears to be confined to applications
where space limitations, power requirements, or ruggedness are less attractive for HDD-based
solutions. Even so, we believe that the 1-inch and similar small form factor drives (SFF) will be a
relative low percentage of total drive shipments over the next three years, and should be viewed as
“icing” to a good growth market.
• Flash will likely only challenge the drive space at low capacity points (under 20GB) over the next 2-
3 years, that larger capacity environments requiring higher performance will remain the domain of
HDDs, and that current fears of a more substantial threat are overblown.
Flash vs. Drives: Advantages of Each Technology
We believe that at least part of the confusion in the flash vs. drives discussion is the difficulty in finding an
honest comparison of the advantages of each technology. While not an exhaustive list, we believe our
findings in Figure 1 include some of the most critical features. Note that pros for drives should be considered
cons for flash and vice versa. Additionally, where possible, we have used Hitachi’s new “Mikey” microdrive
design as a proxy for HDDs, as it is the closest comparable to flash in targeted end markets.
Figure 1: Technological Comparison of Flash and HDD storage
- Robustness (i.e. temperature, shock) - Significantly More Capacity Per $
- Smaller Footprint - Faster Write Speeds
- Lower Power Consumption - Better Bit Stability
- Building block architecture/cost - Multistreaming Capability
Source: Needham & Company
Flash Advantage Drive Advantage
Our extensive industry discussions give flash an advantage in four key areas: robustness, footprint, power
consumption and cost scalability. In terms of robustness, flash has very high tolerances with regards to heat (-
40 to 85 degrees C vs. drives at 0-65 C) and shock resistance. Operating shock tolerances of 2,000G
surpasses drives of 400G, although non-operating shock is comparable at 2,000G. It is also worth noting
these statistics are based on the storage on a standalone basis. The ultimate packing will play a key role in
shock performance as many consumer electronics manufacturers are designing their solutions with added
cushioning of the HDD in mind. The footprint is also smaller for flash vs. the 1-inch, although improvements
continue to close that gap. Typical flash form factors are 12.5x20x1.4 mm vs. “Mikey” at 30x40x5 mm.
Additionally, power can be at least 50% lower in flash compared to even the significant advancements in
“Mikey.” Finally, unlike drives, flash appears to have a better ability to scale its costs. For example, drives can
reduce COGS only to the point when the costs of a head, platter, etc. become equal to the price. Traditionally,
the industry migrates buyers to the next capacity at that point to reset the cost curve.
Key advantages for hard drives include significantly more capacity per dollar, faster data transfer speeds
(especially at higher capacities), better bit stability, and multi-streaming capability. As Figure 2 indicates, the
slope of cost reductions, while simultaneously increasing capacity, has enabled the drive industry to continue
to offer a compelling value proposition for customers. Industry checks also suggest that while read speeds
4
may be comparable between flash and drives, write speeds can be up to 10x faster for HDDs. Additionally,
with data transfer rates upwards of 130 MB/sec, we believe drives are well positioned for applications such as
video, where even small errors or lags are particularly evident to consumers (compared to music).
Discussions also indicate that drives have longer bit stability, as magnetic media is designed for higher levels
of writes/erases. While this is not an issue with MP3 players, it does become more of a factor in high I/O
applications such as laptops. Finally, drives are well-suited to multi-streaming data, such as handling an
incoming data flow while simultaneously managing two additional outgoing streams (which is helpful in
applications such as video).
Improvements in both technologies continue to be made. For example, Hitachi’s recent “Mikey” design
improves power consumption by 40%, weight by 10% and overall design dimensions by 20% all with a 100%
increase in density (from 4GB to 8GB). Our message is that neither industry should assume the other is
standing still just waiting to be eliminated.
Overall, we believe that in applications where space, power and ruggedness are most critical, flash will be the
storage medium of choice. Conversely, in situations where speed and capacity at a reasonable price are
essential, HDDs appear to have a clear advantage.
Areal Density/Pricing Analysis
One of the arguments made by flash players is that the flash industry will be able to catch up to or surpass
HDD technology. We disagree. The drive industry has always found new and ingenious ways to move
technology forward while improving overall reliability. Since the disk drive was first invented by IBM in 1957,
capacities have increased by a factor of 35 million. Prior to the release of GMR (Giant Magneto Resistive)
heads in 1997, areal densities were increasing 60% per year. From 1997 until roughly 2003, densities
increased to more than 100% per year. Looking forward, we anticipate a growth rate more at 40% as the
industry is limited by, among other factors, the rate of improvement in the semiconductor manufacturing tools
used to produce the read/write heads. However, this growth rate puts the industry in parallel with the rate of
increase in flash (as it is largely the same equipment set). The price per GB is on a similar trajectory.
Consequently, we believe that for the foreseeable future, there is no crossover point between capacity points
on disk drives and flash.
To provide some further perspective, our checks suggest that the drive industry is nearing significant new
announcements for 1-inch capacity. Utilizing perpendicular recording, we believe that players such as Hitachi
and Seagate are tracking to a 10GB 1-inch solution by the end of C2005, 12GB in 1H06 and 15GB by the end
of C2006. A 20GB 1-inch solution is on the horizon for sometime in C2007. These capacity points are also
made possible by the larger amount of “real estate” that a 1-inch provides over a 0.85-inch drive. In fact, we
believe that the only casualty of flash’s greater presence in consumer applications will be the 0.85-
inch drive, which will likely be unable maintain the traditional $/GB advantage of drives over flash, and may
not survive its infancy as a product.
The following three figures outline the price and capacity curves for each technology.
5
Figure 2: HDD $/GB Declines vs. Areal Density Growth
1.46 2.23
36.97
7.70
17.22
78.71
3.90
1.04 0.82 0.68 0.55 0.44
320
240
30
15
40
60
80
120
160
$0
$20
$40
$60
$80
$100
199
7A
199
8A
199
9A
2000A
2001A
2002A
2003A
2004A
2005E
2006E
2
00
7E
200
8E
0
50
100
150
200
250
300
350
400
450
500
Price
Per
GB
GB
per
Platter
2000-2008
$/GB CAGR: -30%
2000-2008
AD CAGR: +47%
Source: Komag, Inc., IDC
Figure 3: NAND Pricing vs. Capacity
120 169
1,524
483
813
2,858
352
52 30 19 11 8 12
378
32
20
54
86
137
187
266
$0
$500
$1,000
$1,500
$2,000
$2,500
$3,000
1998A
1999
A
2000
A
2001
A
2002
A
2003
A
2004A
2005E
2006
E
2007
E
2008
E
0
100
200
300
400
500
600
Price
Per
GB
Avg
Density
MB
(MB/Unit)
2000-2008
$/GB CAGR: -42%
2000-2008
AD CAGR: +54%
Source: IDC, Needham Estimates
6
Figure 4: HDD vs. NAND Cost Comparison
HDD vs NAND Cost/GB Comparison
$-
$20
$40
$60
$80
$100
$120
$140
$160
$180
2003 2004 2005 2006 2007 2008
Cost/GB ($)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Cost / GB of HDD as % of NAND
NAND
1" HDD
Cost/GB of 1" HDD as % of NAND
Sources: WSTS, IDC, Needham & Co estimates
Conclusion: NAND is unlikely to even approach cost parity with disk drives this decade. Since both
NAND and disk drive components are made on the same semiconductor process (even the head on disk
drives are semiconductors), there is no price per GB crossover expected this decade with the difference being
fairly constant: disk drives having roughly one-fourth the cost of NAND.
NAND Needs Faster Process Adoption Curve to Meet Higher Densities Goals
Recent comments by Korea’s two largest NAND producers that Flash is ready to displace all disk drives seem
to overlook one major factor: they don’t even have the required chip densities yet and it will take years .
Dr. Chang Gyu Hwang, head of Samsung’s semiconductor group, recently claimed his company would have
a 32Gb chip out next year enabling 64 GB cards by the end of 2006. He further went on during recent media
events to predict the ramp of NAND-based solid state drives would be the death knoll for the hard disk drive
market. While this made for very impressive announcements, the reality is that Samsung only just announced
a new 16Gb SLC chip that needs a 50 nm process. The problem with this roadmap scenario is that Samsung
hasn’t even qualified its 60nm process, and ASML sources say 50nm equipment isn’t even been shipped to
Samsung yet. Even assuming a new process in just one year, instead of normal 18 months, market won’t see
the 16Gb in mass production until beginning of 2007. Even Hynix, who is pushing NAND, is just beginning to
sample a 8Gbit solution this month, but it is not a monolithic chip, according to DigiTimes, but four 2Gb chips
stacked together. Samsung is doing both an 8Gb chip and a two 4Gb stacked solution.
To give some idea of why this roadmap is so critical, the current 4 GB iPod Nano is using four 8Gb chips
(actually, two 16Gb “stacked chips”). According to various suppliers to Apple, doing an 8GB Nano would
require either a series of stacked 16Gb chips or a slightly larger chasis. Given that Samsung’s16Gb chips are
not due out until late 2006 or early 2007, Apple may face a pause before it can refresh the Nano line.
Penetration by Flash Into the PC Market: A Niche Opportunity
Emboldened by their pyrrhic victory at Apple, certain flash players are talking again about a 32GB flash
product for the laptop market. Pricing comments suggest that this 32GB product would be available for the
low discount price of $300 (in 2007 when it might see the light of day). For comparison, a 40GB mobile drive
today can be purchased (retail) for a mere $62. However, we do believe a select portion of the laptop market
(such as ultra-light or tablet PCs) may adopt a flash-based solid state solution despite significant trade-offs
such as capacity and cost.
One argument made by flash manufacturers is that disk drives consume a tremendous amount of the power
available for laptop applications. Again our checks suggest otherwise, bolstering our view that the opportunity
in laptops will be a niche. According to Intel (see Figure 5) the hard drive represents only 8% of the power
drain of a laptop, well below that of the screen (estimated at 33%). With even the clock drawing 5%, we
believe the drive is not a significant problem for laptop designers. The screen and the keyboard represent the
greatest drivers for the shape of a laptop, in our view. In desktop applications, the power argument is even
less relevant. Similarly, space requirements in desktops are not an issue as the drive requires only a small
space within a PC, with much of the balance for cooling and future expansion.
7
Figure 5: Power Consumption by Laptop Components (% of Total)
Graphics Card,
8%
Miscellaneous,
8%
Intel® Memory
Controller Hub,
9%
Power Supply,
10%
Display, 33%
CPU, 10%
Clock, 5%
Intel® I/O
Controller Hub,
3%
LAN, 2%
Fan, 2%
DVD, 2%
Hard Drive, 8%
Source: Intel
A Look at the Demand Side of the Equation
With power as an insufficient rationale in laptops for a large shift away from HDDs, we believe flash will find it
difficult to compete with the growing capacity requirements. As Figure 6 illustrates, capacity points for most
applications are moving higher. For example, within the mobile segment, approximately 86% of drives
consumed by laptops are between 40 and 80GB today. By 2009, 83% are expected to be between 80GB and
160GB. In our view, these capacity points are unattainable by flash at a cost effective level by that timeframe.
Consequently, we believe that over the foreseeable future, the encroachment by flash into laptop applications
will be relegated to the R&D lab and ultra-rugged environments such as the military market.
Figure 6: Percentage of Drives by Capacity Point into Notebook Applications
2005E 2006E 2007E 2008E 2009E
Notebooks 40-80GB (86%) 40-80GB (91%) 40-80GB (83%) 60-120GB (76%) 80-160GB (83%)
Total HDDs 40-80GB (63%) 40-80GB (57%) 80-160GB (55%) 80-160GB (53%) 80-160GB (44%)
Source: IDC, 2005
The Real Battleground
As our analysis shows, the opportunity for flash-based drives in the near term is largely confined within the
parameters of the consumer electronics market (and we would argue a subset of that market). We have noted
for some time that the consumer opportunity for drives was far larger than just the 1-inch MP3 market. For
drives, the growing recognition by the average consumer of the need for more storage in applications such as
PVRs and backup devices has created a market opportunity that is uniquely addressable by hard drive
suppliers. For example, IDC forecasts that consumer electronics and external storage will represent 40% of
all drive shipments in 2008 (30% and 10%, respectively). In the near term (2006), we expect consumer
electronics applications to be driven by set-top boxes and digital cameras/MP3 players in roughly equal
percentages (see Figure 7). It should be noted that although on a unit volume basis PVRs are equal to
cameras/MP3s, the profitability is significantly slanted towards PVRs. We believe that certain high-capacity
drives (such as 250GB and above) can garner gross margins in excess of 25-30% or more, compared with
the 1-inch space at 15-20%.
8
Figure 7: Estimated Consumer Device Breakdown for HDDs (Shipments in Thousands)
% of % of % of
2004 Total CE 2005 Total CE 2006 Total CE
Set-top box 15,649 37% 23,883 36% 31,188 38%
(PVR/DVR)
Information Appliance 5,511 13% 5,692 8% 12,544 15%
(i.e. game consoles)
Digital Camera (1in/1.8) 12,213 15,000
Audio Players (1in/1.8) 21,000 17,000
Total Camera/MP3* 17,001 40% 33,213 50% 32,000 39%
Commerical 2,639 6% 1,920 3% 1,849 2%
Applications
GPS & 1,379 3% 2,388 4% 3,634 4%
Telematics
Total Consumer 42,179 67,096 81,215
Total Drive Shipments 306,055 377,000 433,000
% of Total Drives 14% 18% 19%
Sources: IDC, June 2005; IDC, September 2005; Needham Estimates
*IDC only provides a combined total historically.
In our view, the real opportunity for flash in the consumer segment is MP3 applications (sub 10GB), and all
but high-end handsets/converged devices (which combined represent 5-10% of the handset market). Game
consoles could also prove a potential point of conflict between drives and flash, but to date it has remained a
low-volume, low-profitability segment for both. Looking at possible penetration into the market opportunity for
drives in 2006, we estimate that the maximum ‘at risk’ TAM (total market opportunity) is less than 7% of drive
shipments (39% of CE related to cameras/MP3s X 19% CE percentage of total drives). Revenue contribution
from the roughly 7% unit battleground is less than 6%. We suspect that the most likely near-term outcome is a
modest headwind to the industry’s overall 15-20% unit growth (for example, IDC’s recent downward revision
of 1-inch shipments in C2005 by 1%). The main casualty is likely to be the sub 1-inch market. Specifically, we
do not believe the 0.85-inch segment (<1% of industry revenue and a niche product supplied in limited
volume by Toshiba and Samsung today) will get off the ground as its form factor does not allow sufficient
areal density to compete with Flash. Overall, we believe flash will only challenge the drive space at low
capacity points (under 20GB) over the next 2-3 years, that larger capacity environments requiring high
performance will remain the domain of HDDs and that current fears of a more substantial threat are
overblown.
Figure 8: Estimated HDD Unit Shipment Percentages,
Breakdown by Form Factor
Form Factor 2005E 2006E
Mobile 0.85" 0.2% 0.6%
Mobile 1.0" 4.5% 5.1%
Mobile 1.8" 4.0% 4.1%
Mobile 2.5" 19.9% 20.8%
Desktop 3.5" 52.0% 49.9%
Enterprise 2.5", 3.5" 19.4% 19.5%
Total 100% 100%
Source: IDC
9
Flash and HDD Likely to Co-Exist: The “Hybrid” Approach
In the near future, we believe that NAND and HDD will co-exist, often within the same device. This is because
there is no panacea or “one-size-fits-all” device. Many consumer devices today use some form of non-volatile
memory (usually single cell NAND Flash) as a buffer. For examples, disk drives have embedded memory of
typically 2, 8 or 16 MB used in buffering. Intel suggests that increasing this amount could also provide vast
improvements in overall drive efficiency as well as power consumption by reducing the number of accesses.
Company data indicates that using a 16 MB buffer in a disk drive is only 44% efficient, but increasing that
level to 128 MB yields 86% efficiency. With Microsoft’s rollout of Vista (previously known as Longhorn) on the
horizon, the requests are coming in from OEMs to improve drive performance to accommodate the significant
increase in desired boot time. We estimate this could increase the cost of a drive by $3-5, which of course
OEMs are reluctant to pay. Nevertheless, we expect hybrid drives to begin to appear during 1H06. Overall,
we believe this blending of technologies is more likely to occur as the strengths of each technology are
harnessed rather than the wholesale displacement of drives.
For users/applications that need to do both code storage as well as high density storage (e.g., cell phones)
the industry trend has been to use multi-chip packaging that combines NOR, NAND and even other types of
memory (DRAM, PSRAM) into a single package. We believe this model will also occur in the HDD vs. NAND
market. Digital camera makers have already begun to ship products that includes controllers which use 2-32
MB of embedded NAND memory as a buffer for capturing and reviewing photos or short videos, before
“burning them” to a more permanent storage media such as an external Flash card.
Why All the Rhetoric? Is NAND Flash Facing a Multi-Front Battle?
One possible explanation for NAND makers so aggressively marketing the technology is that the NAND
market may be facing a multi-front battle.
1. High End: At the 8GB and higher segments, NAND cards/SDD are trying to pushing up the density
curve, while the hard disk drive makers trying to use small form factor drives to get into the consumer
market. That battle is likely to intensify in 2006 when the new 1-inch “Mikey” drives hit the market. As
an example, we believe the high capacity NAND flash and the 1-inch drives will continue to compete
in the portable music player market even after Apple’s launch of iPod Nano, as new applications for
these devices such as photo or video would require more storage capacity. Figure 9 illustrated our
projections of the NAND vs. HDD in the portable music player market.
Figure 9: Estimated Number of Portable Music Player Shipped Worldwide by Media Type.
Portable Music Player Shipments
0
10
20
30
40
50
60
2004
2005
2006
2007
2008
2009
2
010
2011
2012
2013
2014
Calendar Year
High Capacity Flash
Low Capacity Flash
1.8" Hard Drive
1" Hard Drive
CD
Source: IDC, Needham & Co estimates.
10
2. Low End: One of the growing markets for NAND Flash has been the lower density flash cards and
“disposable” camera markets (such as Lexar relationship with Kodak). However, there are three
alternative non-volatile execution memory options that could hit the mainstream market within 2-4
years: Ovonics, FeRAM and MRAM. There is another option, being development by Matrix that could
pose the biggest threat: it is a 3-D (stacked) memory architecture that would essential be a ROM
(read-only) type memory that could be anywhere from a third to one-fifth the cost of NAND. This could
be a great option for people who use the memory cards similar to camcorder tapes: record on them
once and then buy a new one, instead re-recording over them. Again, this would most likely involve a
hybrid solution
(Agnostic) Consumer Media Makers May Be Among the Best Positioned Stocks
No one benefits more from competition between NAND and disk drives than consumers such as chip
suppliers that enjoy cheaper memory costs. In our case, most of the stocks covered by Needham’s Digital
Media analyst, Quinn Bolton, would be mostly agnostic as to which media type become more dominant in the
consumer arena. For example, Flash Translation Layer (FTL) software makes flash look like a disk drive to
the operating system, enabling digital media chip makers to design to any interface. For these companies, the
current market trends have accelerated consumer adoption patterns that could increase use of their chips.
One potential “arms dealer” that would benefit from the proliferation of Flash drives /cards would be
Silicon Motion, which makes the controllers for these devices.
11
Part II: NAND Flash Supply vs. Demand Analysis
NAND flash demand has been strong throughout 2005, fueled by the recent introduction of Apple’s iPod Nano
which uses NAND flash for storage. This strong demand has allowed the average prices of some high
capacity NAND to increase during the past quarter (spot prices for 2 and 4Gb chips are up 6-7%, while 8Gb
chips are up 20% since June 30). The key investment question is whether these trends can last. In order to
better understand the market, we constructed a supply/demand model comparing the bit production from
manufacturers to projected market demand for NAND memory by application type and density. Some key
conclusions:
1. While NAND bit/unit growth is “slowing” from 2005 to 2006, NAND is still expected to grow through
2008 at a rate of 19% per year.
2. Some supply constraints in 2H2005 in the high Gbit chip segments should be resolved by 1Q06;
however, demand is NOT expected to outpace supply during 2006.
3. We estimate that the industry will need to build 1.8 new 300 mm equivalent fabs in 2006 to meet
incremental demand.
4. The big wild card (no pun intended) to our demand model is usage patterns for Flash cards: how
consumer reuse, upgrade and store their external Flash cards.
Flash Market Continues to Grow
The flash market is expected to continue to grow in the next few years. NAND bit growth is expected to “slow”
from 262% in 2005 to 145% in 2006, while unit growth is forecasted to drop from 84% to 31% over that same
period as the market builds more 8Gb chips. According to the SIA, NAND revenue is expected to surpass
NOR this year and continue to grow through 2008 at 19% per year. This compares to a flat revenue projection
for NOR in the same time period (see Figure 10 for more details).
Figure 10: Flash Market Projections
2003A 2004A 2005E 2006E 2007 2008 CAGR
(04' - 08')
NAND Flash
Bits Total (Millions of 1GB Equiv.) 21 56 203 496
YoY Growth 165% 262% 145%
Units Shipment (Millions) 386 656 1,210 1,586
YoY Growth 70% 84% 31%
Revenue ($Millions) $4,503 $6,467 $8,794 $9,639 $10,931 $12,898 19%
YoY Growth 44% 36% 10% 13% 18%
NOR Flash
Revenue ($Millions) $7,237 $9,143 $8,004 $8,804 $9,069 $9,431 1%
YoY Growth 26% -12% 10% 3% 4%
Source: SIA (for Revenue Projections), IDC and Needham & Co. estimates.
Supply Tracking Demand Growth, Assuming 30% More Capacity Added in 2006
By overlaying demand projections onto our supply model, we believe we get a clearer picture of the NAND
flash market. As shown in Figure 11, bit supply and bit demand (in terms of 1GigaByte equivalents) are
tracking close to parity. Overall, we believe NAND bit demand will not outpace supply in 2006, and that the
industry will only need a 29% increase in capacity next year to meet demand. While our model shows an 18%
oversupply by 4Q2006, we believe the excess is a healthy buffer and not a sign of structural overcapacity.
Put a different way, that 18% delta would translate into less than one extra 300 mm fab (~0.8 fabs) next year
for the entire industry, or 35,000 wafer starts per month, easily absorbed over the seven major suppliers.
Such a buffer could also handle the non-consumer NAND business or any increase in unexpected, new flash
applications.
12
Figure 11: NAND Flash Supply vs. Demand
0
20
40
60
80
100
120
140
160
180
200
1Q05 2Q05 3Q05 4Q05 1Q06 2Q06 3Q06 4Q06
Millions of 1GB Equiv.
Total Flash Supply
Total Flash Demand
Source: Needham & Co. estimates
Manufacturers Need to Continue to Increase Supply
On the supply side, we expect to see strong bit growth continue through 2H06, as suppliers build more high
capacity (>2Gb) products while maintaining the volume of low capacity in their older generation fab. Using
data from published reports, third parties data and our own checks, we modeled the 2005-2006 quarterly total
supply by looking at the production targets of the top seven NAND manufacturers. We projected the
corresponding wafer starts (in 200 mm equivalents) and number of wafer fab necessary to meet the
production targets. Figure 12 shows the estimates number worldwide 200 mm equivalent wafers starts
needed to meet the production targets. Overall, our model projects unit supply should continue to grow rapidly
in 2H2005 (84.4% YoY), but at a slower pace in 2006 (31.1% YoY). Based on our finding, we believe the
number of 200 mm equivalent wafer fabs worldwide will increase from 10.8 fabs in 2Q05 to 13.7 by the end
2005, and up to 17.7 by the end of 2006. This data indicates that the number of chips output would slow in
2006 as the results of increases in density per chip. Since our projections are base on 200 mm equivalents,
the current plans to build 300 mm fabs by leading NAND manufacturers would future slow the need to build
more fabs. The four (200 mm) fab delta between the end of 2005 and 2006 would roughly translate into 1.8
300mm equivalent fabs.
Figure 12: Worldwide Wafer Supply for NAND Flash Productions
Total Worldwide Supply of Equiv. No. of 200mm Wafers per
Quarter
0
500
1000
1500
2000
2500
3000
1Q04 2Q04 3Q04 4Q04 1Q05 2Q05 3Q05 4Q05 1Q06 2Q06 3Q06 4Q06
Equiv. No. of 200mm Wafers (Thousands of
Wafers)
Source: Needham & Co. estimates
Delay of Process Technology Migration Would Impact Supply
With memory density per chip grew 56% and bit growth increased 163% in 2004, NAND flash manufacturers
had to constantly lower their cost to maintain profits. The simplicity of the NAND flash device architecture
13
(fewer metal layers than NOR or logic chips), combined with its low reliability requirements, makes it easier
for manufacturers to migrate NAND device process technology to lower manufacturing cost. As highlighted by
recent Samsung’s announcement of its 50 nm process just one year after its 60 nm (although we believe 50
nm will not be ready by 2007), most NAND manufacturers are following Samsung’s lead to reduce cost by
transitioning productions every four to five quarters to new technology nodes, versus more traditional six to
eight quarters for other semiconductor devices. Figure 13 demonstrates the increase utilization of newer
process technologies.
Figure 13: Unit Production of NAND Chips by Process Technology Node
Unit Production of NAND Chips, By Process
0
50
100
150
200
250
300
350
400
450
500
1Q04 2Q04 3Q04 4Q04 1Q05 2Q05 3Q05 4Q05 1Q06 2Q06 3Q06 4Q06
Millions of NAND Chips
60nm
70nm
90nm
120nm
150nm
Source: IDC, Needham & Co. estimates
In our model, we assume aggressive technology migration for NAND, especially for the high capacity devices.
We also modeled in a lower starting yield (60% in our model) for any new technology than mature technology
(80% in our model). Looking at the sensitivity of technology migration and yield improvements, we observed
these impacts on the number of wafer fabs needed to meet supply targets:
1. If technology migration would slow down by one quarter for all device sizes, our model shows that by
4Q06, this would increase the total 300 mm monthly wafer starts by 16,200 wafers, which translates
to approx 0.4 fabs. Based on this analysis, we believe the timing of technology migration has a mild
effect on NAND supply.
2. If we assume yield drops to 50% from each technology migration for all device sizes, the total 300
mm monthly wafer starts would increase by 50,000 wafers, or 1.26 300mm fabs, in 2006. This
shows that while technology migration would improve profit margins, yield impacted by newer
technology has a crucial effect on production.
Demand Growth: Strong Bits but Slowing Units
We believe the while bit demand growth for NAND flash continue at its torrid pace into 2006, the growth in
units and total revenue will slow. To be clear, we believe NAND demand would continue to see a health
growth and outpace the entire market. In fact, our model shows a 133% YoY bit growth in 2006; however, we
believe this it the result of continue price/bit decrease and the elasticity of the NAND market, and as the
result, the unit growth will be at a more moderate pace in 2006.
We note that some market participants assume a much higher NAND consumptions by digital camera via
flash memory card than our opinion. For example, a recent report by one third party research group implied
that all flash memory cards sold would be used in digital camera, and used this assumption with digital
camera sales volume projections to estimate the demand of flash memory cards. While we agree that one of
the main growth drivers behind flash memory cards is digital cameras sales, we believe such approach might
overestimate flash card demand. Flash memory cards are being used in other mobile devices such as smart
handhelds, mobile phones, handheld game console and even some mp3 players. We believe assuming all
cards to be used in digital cameras would overestimate the number of card used per camera that results in an
over-projection of future card consumptions.
14
We believe a better way to analyze flash consumptions is to separate the external flash memory cards
demand from the projected market for embedded memory per device. In our approach, we model our flash
memory card projections based on data collected by ourselves and third party reports. For example, our
sources at Micron indicate that they believe there is roughly 17MB of embedded memory per “point and
shoot” digital camera, and 376MB for each card sold into the DSC segment in 2005.
Combining our digital camera and flash memory cards data with demands of six other consumer products
groups, we form a complete picture of NAND bit demand per device in 2005-2007, as shown in Figure 14.
Figure 14: Flash Content per Device (MB)
2004 2005 2006 2007
Digital still camera (embedded) 27 38 78 113
Digital video camcorder 134 175 333 490
MP3 400 667 1294 1897
Mobile phone 8 20 75 147
Smart handheld device 15 34 80 125
USB flash drive 264 381 660 966
Game console 38 61 97 119
Flash Memory Card 166 197 222 255
Sources: IDC, Micron, SanDisk
While the average density per flash memory card may seem low over time, this is a blended average of all
types of cards: CompactFlash, MultiMedia Cards, SD, etc. We will elaborate more on Flash Cards /Drives in
the next section.
Flash Card Usage: The Wild Card in Elastic Market
We believe flash memory card market to be very price elastic, with the price per bit of NAND memory is
expected to continue to fall at a similar pace as new supply comes on line next year. However, the biggest
swing factor in our demand analysis is the usage model of Flash Memory Cards. If the price of cards
becomes low enough, consumer may adopt a use-it-once mentality (similar to how VCR or camcorder tapes
evolved), which could accelerate the growth curve. Furthermore, consumers might find new ways for using
flash memory cards if price falls. Similar to the explosive growth of digital photography, digital videography
could be the next popular media for the mass consumer market. As compression technology improve, digital
video files would become small enough that consumer could fit multiple files within a flash card. However, as
shown in Figure 15, a high capacity flash card could store quite a lot of video, pictures and song files.
Ultimately, we believe even free memory may find that the elastic curve flattens out, since there are only so
many video, pictures and song a person use in a year.
15
Figure 15: Estimates Capacity and Usage of Flash Card
Sources: SanDisk, April 2005, Canon, Needham estimates
MP3: 128Mb/sec, average song = 4.5 mins. (~ 4MB); DSC 2 Megapixel camera
DSC: 4 Megapixel camera (high quality JPEG) - Canon website
Video Clip & Broadcast Recording: VGA, 30 fps, MPEG 4 @ 384 Kbps
Average Card Capacity - Mobile
180 Songs
40 Songs
217 Pics
30" clips
120" video
-
200
400
600
800
1,000
1,200
1,400
1,600
1,800
2,000
2005 2008
MB / Card
Broadcast Video
Personal Video
Camera
MP3
What was somewhat surprising during our analysis was that the overall flash memory card segment is
expected to decline from 33% of the total flash bits shipped in 2005 to 21% of 2007 shipments. This decline,
however, is due to the mix of low and high density flash cards (mentioned in the previous section) and a shift
in the amount of embedded NAND used in MP3 players and mobile phones. The recent success of the Apple
Nano would suggest that consumers would accept and pay up for the embedded memory.
In Figure 16, we examined the market by breaking out three major types of external, high density Flash cards
or drives: CompactFlash, Secured Devices and USB drives. Three key points from the market trends and
their implications:
1. High density, external drives will stay at a fairly constant level percentage of the total amount of
NAND Flash shipped. This is important since it means that higher density cards/drives will need
higher density chips, pushing the need for OEMs to upgrade their equipment.
2. High density media devices (USB drives, CompactFlash and SD cards) comprise just 41-47% of total
unit shipments for cards/drives. This means that lower density cards (MMC, miniSD) with slower
density ramps are weighing down the blended average of memory per card.
3. The year-over-year growth in units is very high, even as the density per device more than doubles by
2007. For companies such as Silicon Motion, which makes Flash controllers for external devices, it
isn’t the amount of density, as much as the number of devices shipped that would be critical to the
business of these “arms dealers.”
16
Figure 16: Compare High Capacity External Flash Devices to Flash Market
2004 2005 2006 2007
Units (in millions)
USB Flash Drives 57.8 110.2 161.8 194.5
CompactFlash cards 31.3 32.3 34.0 34.2
Secured Digital (SD) cards 33.5 58.8 74.7 92.4
Sub-Total (high density) 122.5 201.3 270.5 321.0
y/y growth for high density cards/drives 64% 34% 19%
High density cards as % of total Flash Cards/drives 42% 47% 43% 41%
average memory density per card/drive (MB)
USB Flash Drives 264 381 660 966
CompactFlash cards 314 443 769 985
Secured Digital (SD) cards 136 199 259 317
Sub-Total (high density) 241 338 563 781
(Millions of GB shipped)
USB Flash Drives 15.2 42.0 106.8 187.8
CompactFlash cards 9.8 14.3 26.1 33.6
Secured Digital (SD) cards 4.5 11.7 19.4 29.3
Sub-Total (high density) 29.6 68.0 152.3 250.7
y/y growth for high density cards/drives 130% 124% 65%
% of total NAND bits sold
USB Flash Drives 19% 24% 26% 26%
CompactFlash cards 12% 8% 6% 5%
Secured Digital (SD) cards 6% 7% 5% 4%
Total High Density Cards/ Drives 37% 39% 37% 35%
Source: IDC, USB Alliance, Needham & Co. estimates
Overall Flash Demand Bit Growth Should Double Again
As shown in Figure 17, the flash bit demand is projected to continue to grow in a rapid pace at 134% YoY in
2006, while the unit shipment is growing in a much more moderate pace at 25.1% YoY. This is the results of
increase consumption of higher capacity NAND flash, such as Apple iPod Nano and higher capacity flash
memory cards. As a commodity product, NAND price/bit is expected to fall in the future. Therefore we believe
the revenue growth rate would track closer to the unit shipment estimate, and grow in a slower pace than
2005.
Figure 17: Flash Demand (Bits vs. Units)
0
100
200
300
400
500
600
700
1Q05 2Q05 3Q05 4Q05 1Q06 2Q06 3Q06 4Q06
Bit Demand in 1GB Equiv (Millions)
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
160.0
Unit Demand (Millions)
Unit Demand
Bit Demand
Source: IDC, Needham & Co. estimates
17
APPENDIX 1: A Quick Primer on Flash Memory
NOR vs. NAND
Flash memory is a type of nonvolatile memory that allows users to use electricity to program and erase
memory cell. The “non-volatile” aspect comes from a design where the memory is retained even when the
power is turned off; compared to DRAM memory which must be constantly “refreshed” while in use and
losses its content completely when the PC is shut-off and must be “rebooted”, when turned on again.
Technically, flash memory has these capabilities because it is designed with a thin oxide layer that is thin
enough to be reprogrammable, but effective enough to prevent electrons from leaking (no information lost)
when power is off. There are two major types of flash memory: NAND and NOR.
NOR Flash was initially developed and introduced by Intel in 1988 as a lower cost, high capacity replacement
for EEPROM. NOR is a random access device, which makes it appropriate for storing program codes where
access time (read speed) is important. NOR flash has a more robust memory design than NAND, which
means it seldom has bad block in its memory cell and does not require file management. However, NOR has
a much slower writing and erasing speed, and has a lower storage capacity per chip than NAND.
NAND Flash was initially developed by Toshiba in 1989 for mass storage purposes. It is a sequential access
device that has faster write and erase speed than NOR, but a slower read speed. NAND is also less reliable
than NOR and requires an on-chip write bad block management controller. However, it has the advantage of
much higher capacity per chip and lower $/MB than NOR because of its high density.
Figure 18: NOR vs. NAND
NOR Flash NAND Flash
Interface Random Access Sequential Access
Read Speed 55 MB/s 27 MB/s
Write Speed 0.15 MB/s 7.5 MB/s
Erase Speed 0.032 MB/s 85 MB/s
Max Density Shipping 1 Gbit 8 Gbit
Reliability No not need file management Requires bad blocks write controller
Applications Program Code Execution Mass Data Storage
Source: Toshiba and Needham & Co estimates
Gigabit (Gb) vs. GigaByte (GB)
There are eight (8) bits or memory in one byte; thus it take eight 1-Gigabit (Gb) chips to make one 1 Gigabyte
(GB) card. This may seem silly to mention, but there have been several articles and reports we’ve read that
have either had typos or have confused the difference between a Gb and a GB.
Monolithic vs. Stacked Chip
Samsung recently announced its first 16Gb NAND chip, but if you go to Samsung’s website, the company
already has a 16Gb product. Why? This is because Samsung (and other manufacturers) stacks two dies on
the single package to build higher capacity chip. Apple is using this two of these 16Gb (2GB) stacked chips
to fit the tiny package of their 4GB iPod Nano. Most manufacturers start a high capacity NAND with a 2-Stack
or 4-Stack package until they can reliable produce higher chip die, and then move to a monolithic package for
lower cost.
SLC vs. MLC
Most flash memory cells use a voltage level that distinguishes its value. In a single-level cell (SLC), there is
one memory bit per cell, with a value of 1 or 0 – if the voltage in that cell is above a reference threshold, the
value would be 1; if below, it would be 0 (or vice versa). However, in a multiple-level cell, the voltage is broken
up into additional states, like a measuring cup. In a 2-bit per cell MLC, there would now be four states; when
the voltage is between certain thresholds, it would hold a given set of values – one for the first bit, and one for
the second bit. Figure 19 compares a SLC with a MLC design. If this were a 3-bit per cell design, the cell
would be split into 8 difference states, for 4-bit per cells, sixteen states.
18
Figure 19: SLC vs. MLC
Source: Samsung
Advantages of SLC Over MLC:
• SLC uses less energy. Since SLC use only two electrical states instead of the 4+ by MLC, it can expend
less energy to program and manage it charges.
• Simpler architecture. MLC designs need more error correction (ECC or EDC) and have lower
performance than a SLC design in terms of slower access time (25ms vs 70ms). In addition, due to the
multiple energy states, this also leaves MLC’s with a lower margin of error to read the bit.
• SLC lasts 10x longer. Most SLC chips are qualified to endure 100,000 read/write cycles, while a MLC
chip is only spec’d for 10,000 cycles. To achieve over 10,000 cycles, MLC must go from 1-bit ECC to
multi-bit, increasing cost and lowering performance.
• MLC is a Process Generation Behind SLC. According to Samsung (which does not have an MLC design
yet), SLC NAND is 6-9 months ahead of MLC with the same technology and that difference is expected to
increase as geometries get smaller. Smaller die size (though less Gb per square mm of silicon).
• Cost advantage of MLC not as great as thought. While having 2-bits in one cell would seem to mean
that the cost of MLC is half that of SLC, the need to add more overhead circuits (increasing the die size),
lower die yields, and increased testing to handle the MLC, means that the final cost of MLC is just 80%
that of SLC. So the question is whether the 20% savings is worth giving up performance, time-to-market,
reliability and endurance.
19
ANALYST CERTIFICATION
I, Charlie Glavin, hereby certify that the views expressed in this research report accurately reflect my personal
views about the subject company (ies) and its (their) securities. I also certify that I have not been, am not,
and will not be receiving direct or indirect compensation in exchange for expressing the specific
recommendation(s) in this report.
I, Richard Kugele, hereby certify that the views expressed in this research report accurately reflect my
personal views about the subject company (ies) and its (their) securities. I also certify that I have not been,
am not, and will not be receiving direct or indirect compensation in exchange for expressing the specific
recommendation(s) in this report.
Companies mentioned in this report under coverage by Needham & Company, LLC (Prices as of
October 11, 2005):
ASM Holdings (ASML, $16.29, Buy, Disclosures: B, G)
Apple (AAPL, $51.59, Buy, Disclosures: A, B, G)
Hutchinson (HTCH, $26.04, Buy, Disclosures: B, G)
Intel (INTC, $23.42, Buy, Disclosures: B, G)
Komag (KOMG, $28.91, Strong Buy, Disclosures: B, G)
Lexar (LEXR, $7.45, Hold, Disclosures: B, G)
Seagate (STX, $15.67, Strong Buy, Disclosures: B)
Silicon Motion (SIMO, $13.66, Buy, Disclosures: B, C, D, E, G, J)
445 Park Avenue, New York, NY 10022 (212) 371-8300
% of companies under coverage % for which investment banking services
with this rating have been provided for in the past 12 months
Strong Buy 7% 35%
Buy 59 20
Hold 29 3
Under perform 2 0
Rating Suspended 1 0
Restricted 2 71
Under Review <1 0
Needham & Company, LLC (the Firm) employs a rating system based on the following (Effective July 1, 2003):
Strong Buy: A security, which at the time the rating is instituted, indicates an expectation of a total return of at least 25% over the next 12 months.
Buy: A security, which at the time the rating is instituted, indicates an expectation of a total return between 10% and 25% over the next 12 months.
Hold: A security, which at the time the rating is instituted, indicates an expectation of a total return of +/- 10% over the next 12 months.
Underperform: A security, which at the time the rating is instituted, indicates an expectation that the price will depreciate by more than 10% over the next 12
months.
Under Review: Stocks may be placed UR by the analyst, indicating that the stock rating and/or price target are subject to possible change in the near term,
usually in response to an event that may effect the investment case or valuation.
Rating Suspended: Needham & Company, LLC has suspended the rating and/or price target, if any, for this stock, because there is not a sufficient
fundamental basis for determining a rating or price target. The previous rating and price target, if any, are no longer in effect and should not be relied upon.
Restricted: Needham & Company, LLC’s policy and/or applicable law and regulations preclude certain types of communications, including an investment
recommendation, during the course of Needham & Company, LLC’s engagement in an investment banking transaction and in certain other circumstances.
For disclosure purposes (in accordance with NASD requirements), we note that our Strong Buy and Buy ratings most closely correspond to a “Buy”
recommendation. When combined, 66% of companies under coverage would have a “Buy” rating and 21% have had investment banking services provided
within the past 12 months; Hold mostly correspond to a “Hold/Neutral” recommendation; while our Underperform rating closely corresponds to the Sell
recommendation required by the NASD.
Our rating system attempts to incorporate industry, company and/or overall market risk and volatility. Consequently, at any given point in time, our
investment rating on a stock and its implied price appreciation may not correspond to the stated 12-month price target. For valuation methods used to
determine our price targets and risks related to our price targets, please contact your Needham & Company, LLC salesperson for a copy of the most recent
research report on the company you are interested in.
To review our Rating system prior to July 1, 2003 please refer to the following link: needhamco.com
Stock price charts and rating histories for companies under coverage and discussed in this report are available at needhamco.com. You may also
request this information by writing to: Needham & Company, LLC, 445 Park Ave, 15th Floor (Attn: Compliance/Research), NY, NY 10022
ANALYST CERTIFICATION
By issuing this research report, each Needham & Company, LLC analyst and associate whose name appears within this report hereby certifies that (i) the
recommendations and opinions expressed in the research report accurately reflect the research analyst’s and associate’s personal views about any and all of the subject
securities or issuers discussed herein and (ii) no part of the research analyst's or associate’s compensation was, is or will be directly or indirectly related to the specific
recommendations or views expressed by the research analyst or associate in the research report.
The following disclosures (as listed by letter on the cover page) apply to the securities discussed in this research report:
“A” The research analyst and/or research associate (or household member) has a financial interest in the securities of the covered company (i.e., a long position
consisting of common stock).
“B” The research analyst and research associate have received compensation based upon various factors, including quality of research, investor client feedback and
the Firm’s overall revenues, which includes investment banking revenues.
“C” The Firm has managed or co-managed a public offering of securities for the subject company in the past 12 months.
“D” The Firm and/or its affiliate have received compensation for investment banking services from the subject company in the past 12 months.
“E” The Firm and/or its affiliate expect to receive or intends to seek compensation for investment banking services from the subject company in the next three
months.
“F” The analyst or a member of the analyst's household serves as officer, director or advisory board member of the covered company.
“G” The Firm, at the time of publication, makes a market in the subject company.
“H” The Firm and/or its affiliates beneficially own 1% or more of any class of common equity securities of the subject company.
“I” The analyst has received compensation from the subject company in the last 12 months.
“J” The subject company currently is or during the 12-month period preceding the date of distribution of this research report was a client of the Firm and received
investment banking services.
“J1” The subject company currently is or during the 12-month period preceding the date of distribution of this research report was a client of the Firm and received
non-investment banking securities related services.
“J2” The subject company currently is or during the 12-month period preceding the date of distribution of this research report was a client of the Firm and received
non-securities related services.
“K” Our affiliate has received compensation for products and services other than investment banking services from the subject company in the past 12 months.
This report is for informational purposes only and does not constitute a solicitation or an offer to buy or sell any securities mentioned herein. Information contained in
this report has been obtained from sources believed to be reliable, but Needham & Company, LLC makes no representation as to its accuracy or completeness, except
with respect to the Disclosure Section of the report. Any opinions expressed herein reflect our judgment as of the date of the materials and are subject to change without
notice. The securities discussed in this report may not be suitable for all investors and are not intended as recommendations of particular securities, financial instruments
or strategies to particular clients. Investors must make their own investment decisions based on their financial situations and investment objectives. The value of income
from your investment may vary because of changes in interest rates, changes in the financial and operational conditions of the companies and other factors. Investors
should be aware that the market price of securities discussed in this report may be volatile. Due to industry, company and overall market risk and volatility, at the
securities current price, our investment rating may not correspond to the stated price target. Additional information regarding the securities mentioned in this report is
available upon request. © Copyright 2005, Needham & Company, LLC, Member NASD, SIPC. |
