Vendor
Genesis
Hewlett-Packard
(HP) was formed in 1939 by Bill Hewlett and David Packard, starting out
in the famous garage behind Mr. Packard's house in Palo Alto, CA. The
first product was a resistance capacity audio oscillator, an instrument
used to test sound equipment. HP continued to grow through design and
production of instruments and test equipment.
HP's
first entry into a computer-related business was in 1958, when it acquired
a graphics-recorder business. This grew into HP's printer business, one
of the world's largest. HP continued to expand its product portfolio,
including medical devices and analytical instrumentation, primarily through
acquisition.
In 1999, Hewlett-Packard decided to split the company into two separate
companies. The computing-products retained the name Hewlett-Packard; the
new company, called "Agilent Technologies", encompasses the test and Measurement
(T&M), chemical analysis, medical, and semiconductor product lines. Because
of this split, the revenues and earnings figures for 1999 do not "map"
(vis--vis trends) to 1998. In
1966,
HP developed its first computer, the HP 2116A. This product was originally
designed as a test and measurement equipment controller. This eventually
led to the development of the business-focused HP 3000 minicomputer in
1972. In the early 1980's, Hewlett-Packard produced the HP-85, their first
PC, and in 1986 HP introduced its first Reduced Instruction Set Computing
(RISC) machine.
Hewlett-Packard
now provides personal computers, printers, workstations, notebook computers,
and servers. There are three different server product sets, using three
different architectures:
- The HP 3000 series, which uses the MPE operating system
- The HP 9000 series, HP's Unix product line, running HP-UX (HP's Unix
version) on PA-RISC architecture
- The HP NetServer series, their Intel-based, Windows NT-focused servers
Although
the first two products are interesting, and the HP 9000 is a high-performing
Unix system, we will be discussing only the NetServer in this Note.
Hewlett-Packard
derives approximately 85% of its revenues from product sales, with the
remaining 15% coming from services. The markets in which HP competes include
desktop PCs, notebook PCs, workstations, printers, storage, and servers.
HP
is the world leader in printers, and is a key player in all the other
markets mentioned. HP's revenues for fiscal 1999 were approximately $42.4
Billion, with income of approximately $3.5 Billion. Growth rates for the
1998-1999 period are not applicable due to the HP/Agilent split. However,
in years prior to the split HP's revenue growth was approximately 11%
per year, with income growth approximately 12% per year. Income grew faster
than revenue, a trend we expect to see continue post-split.
Vendor
Strategy and Trajectory
Hewlett-Packard's
corporate computing philosophy is to provide reliable solutions to whatever
computing needs customers might have across the spectrum of architectures.
These architectures include the Intel x86 architecture, Unix (their HP-UX
servers), and their proprietary PA-RISC architecture. With the exception
of IBM and possibly Compaq, HP has arguably the broadest set of computing
solutions available in the market today.
HP
has been a leader in some market areas, including workstations and Unix
servers, but has been relegated to fourth place in market share (both
US and worldwide) for Intel-based servers. Although fourth place is usually
not a bad showing, this is a relatively poor result. HP formerly had third
place, and the "Big Four" (Compaq, Dell, IBM, HP) control approximately
75% of the Intel-server market, which means that fifth place is effectively
"nowhere".
HP's
product strategy is to provide high-availability, high-reliability solutions.
This is highlighted by their "Assured Availability" program, which is
a joint effort between HP and Marathon Technologies and is discussed later
in this Note. In addition, HP has provided redundant systems (power, cooling)
in their NetServer line for close to five years, highlighting their commitment
to reliability. HP is trying to increase market share by combining their
reliability focus with their ability to deliver multiple OS/architecture
solutions throughout the enterprise .
Vendor
Strengths
Availability/Reliability/Manageability
HP provides reliable systems and servers, enhanced by their "Assured Availability"
program, which uses the "Endurance" configuration developed by Marathon
Technologies. For the appropriate price, the program will guarantee availability
up to 99.999% annually - in other words, five minutes of downtime per
year. We get nervous when we see the phrase "up to XX%" (because it does
not state the minimum guarantee), but we appreciate HP's willingness to
provide the service for NT servers, especially considering NT's reputation
for lack of robustness. HP has also focused on designing in hardware reliability
through redundancy. As mentioned above, their servers were among the earliest
adopters of redundant power and cooling for small servers (i.e. those
running Windows NT), features that have become standard in the industry.
HP's
management tools (OpenView and Top Tools) are well known through the industry.
In addition, other vendors (including Dell) have at times included HP's
products in their own offerings. It should be noted, however, that OpenView
is not universally loved - some feel it is difficult to use.
Packaging
HP
has earned a reputation for designing servers which make it easy for the
customer (or service tech) to upgrade or replace components. HP's design
team(s) generally pay attention to ergonomic issues, such as judicious
use of color coding, good labeling and "mapping" of functionality into
an effective means of presentation.
Corporate
Viability
HP
went through some tough times in the late 1990s, including a number of
fiscal quarters where they underperformed market expectations. As recently
as mid-1999, when new CEO Carleton "Carly" Fiorina was hit with some significant
earnings problems, HP looked to be thrashing about and its stock was pummeled.
However, Ms. Fiorina has managed to restore Wall Street's confidence in
HP, and earnings have rebounded as well.
Breadth
of Offering
As
mentioned earlier, HP is one of the two or three companies which can supply
computing hardware and software across the entire spectrum of needs. This
means it can provide a set of products for a heterogeneous environment,
rather than being limited to an Intel-only or a Unix-only offering. Although
IBM has a clear advantage vis--vis breadth, Dell's standard offerings
are more limited, and it is unclear whether Compaq can combine its various
acquisitions effectively enough.
Price/performance
HP
is paradoxical in this area. For some time, HP's price/performance story
was primarily confined to its Unix servers. NetServer performance, although
reasonable, had typically cost more - in some cases significantly more
- than offerings from Dell and Compaq. A typical example is the transaction-processing
benchmark of $/tpmC - HP's previous submission in January, 1999 (the most
recent one until the release of the LH 6000) provided 23143 transactions
per minute (tpmC) at a five-year cost of $26.87 each. This compares poorly
to a similar Compaq system, which delivered 22478 tpmC for $18.84 each.
This translates to a premium of nearly $200K. [Note: Although the TPC-C
benchmark is not the sole test of value, it is pervasive enough that Intel
and Unix vendors submit frequently.]
However,
with the recent release of the six-processor LH 6000, HP has beaten Dell
(the former leader) by almost $1/tpmC, approximately seven percent lower
than Dell's best mark. [Lower is better.] Although it is unwise to extrapolate
the performance of an entire product line from the results of a single
system, these latest results are a sign that HP is now more focused in
this area. This is a "good thing" for both HP and its customers, but customers
should monitor this closely over the coming months to see if the trend
continues.
Vendor
Challenges
Intel
Reliance
For a number of years, HP has relied on Intel to provide Main Logic Boards
for its servers. These boards are generally slightly modified versions
of Intel "vanilla" boards. Like Dell, HP relied heavily - although not
completely - on Intel for a lot of the "innards" for the eight-CPU servers
released last year. Although this path means HP does not need to expend
as much of R&D money as when they develop a product entirely internally,
we question the long-term viability of this strategy. HP might consider
weakening the bonds, as Dell claims it has.
Mixed
message on performance
As
with price/performance, HP is paradoxical. HP talks a lot about performance,
but the published results (prior to the release of the LH 6000) indicate
this is not one of their key strengths. In addition to a spotty submission
record to benchmark groups such as the TPCC (no Intel-based server submission
between January, 1999 and March, 2000), the performance figures have not
provided a compelling story. For example, HP touts the performance of
the new LH3000, but the only benchmark results are for the MAPI Messaging
Benchmark (MMB) test and are average at best. [A Pentium III/733 MHz system
giving an MMB of 15,000 vs. 14,600 for a January, 1999, Compaq ProLiant
7000 with Xeon/450MHz is hardly an earth-shattering result.]
There is nothing wrong with not being at the top of the heap in performance,
but making it appear that one is on top by "judicious" use of comparative
data is not a tactic with which we feel comfortable. If HP can continue
to produce results on a par with those of the LH 6000 (tpmC for a 4CPU
only 1% lower than Compaq's best result), we will modify our position.
Weakening
Market Position
Although
HP has not "fallen off the radar screen", its revenue market position
dropped from second to fourth in 1999, and its WW unit sales share decreased
from ~14% to 12.5%, and its US unit sales share lost even more[Source:
IDC]. Its Intel server market position was passed quite easily by Dell,
and it is now viewed by some almost as an also-ran. We believe part of
this is due to HP's reliance on selling through value-added resellers
(VARs). Dell has gone from nowhere to second place, primarily on the strength
of its direct-sales model. Compaq, seeing its former dominance dissipate,
is trying to emulate this. HP currently sells only three models (out of
nine manufactured) through direct online sales to general customers. (HP
has a special direct selling program for "named global accounts"; our
focus here is on the everyday customer, not special situations.)
Compaq
has estimated that a VAR/channel distribution can result in a cost differential
of almost 15% - either higher prices or reduced profits. With an already-high
price/performance position, HP cannot really afford the extra $$$. HP
must either reduce costs or increase functionality to overcome negative
perceptions.
Vendor
Predictions
Despite
the shrinking market share, HP will not disappear from the Intel server
space. The division has problems to overcome, but the products are generally
solid enough that the market will want to keep them around. We expect
HP to continue to be a serious competitor in the Intel space in three
years (75% probability).
HP has done well with its focus on reliability and availability. However,
it will find that the market is less willing to pay a high premium for
increased uptime. Dell has shown that high reliability and high customer
satisfaction can be delivered for a lower cost. This has generated (and
continues to generate) price pressures on the competition, including HP.
Dell does not offer the breadth of products that HP does, which translates
into some heterogeneous environments being better suited for HP than Dell.
However, Dell is extremely aggressive in the pursuit of sales, so this
may be a false sense of "security" for HP. We therefore continue to see
the need for HP to address pricing and distribution. (80% probability).
We
expect HP to continue to form alliances with companies having complementary
products, such as the deal with Procom for Network-Attached Storage (NAS)
servers (70% probability). HP might consider forming a relationship with
(or buying) a load-balancing software company, with the purpose of developing
a low-cost pre-packaged cluster solution, similar to Compaq's "Toucan"
unit (now called the ProLiant CL380). Although this can be considered
a "me too" product, we can see the benefits. HP should also consider a
relationship with one of the 1U-high server designers/manufacturers to
reduce time to market.
We
expect HP to update its product line in a timely fashion, and has already
started with the LC2000, the LH3000. However, producing "more of the same"
type of products will be insufficient to move NetServers beyond its current
spot. When the LPr was released, 1-2 years ago, it was something new and
different, with the potential to have a big impact on the marketplace.
HP needs to continue in a that vein, and the LH/LT 6000 series is a strong
start.
We
do not see HP being acquired by any of its hardware competitors - there
would be insufficient extra capability to make the deal worthwhile. We
do not see HP acquiring any of its hardware competitors. (70% probability
of HP staying essentially "as is").
Vendor
Recommendations
Improve price/performance
The
NetServer product line cannot afford to be significantly higher in price/performance
figures than Dell and Compaq. Both of those competitors have better price/performance
figures, Compaq has three times the market share, and Dell has the momentum.
Status quo is not viable if HP wants to regain third place, or even increase
its share percentage. HP must compete head-to-head if it wants to stay
in the game. The excellent results for the NetServer LH 6000 indicate
HP is trying to remedy this situation; we expect it to continue, but will
monitor the situation.
Investigate direct sales (and benefits) more thoroughly
Investigate
direct sales (and benefits) more thoroughly As Compaq has discovered (we
think), emulating Dell's direct sales model will reap benefits. Reduced
pricing will come from at least two areas: reduction of outflow to VARs,
and improved return on capital through shorter inventory and production
cycles. If HP decides to go through with this, it will take them at least
a year to implement it, but the payback will be worth it.
Revamp
product line
HP's
server products need some updating. The recently-released LH 6000 has
posted impressive numbers in benchmark testing, for both performance and
price/performance, and LT 6000 has set a new CPU density standard for
4U-high (1U=1.75") rackable systems. Its new mid-range servers, the LC
2000 and LH 3000, are solid but "unexciting" in the sense of groundbreaking
new functionality. Unexciting is not necessarily a bad thing, the ability
to be rock-solid is underrated. However, in today's highly competitive
server market, customers can get stability in a number of places, and
are looking for more in the way of features and benefits. The rest of
HP's product line is getting a little "long in the tooth" - HP should
work on those products next. One way this might be accomplished is through
focusing on highly dense packaging. HP has been able to do it before with
the LPr server (2 CPUs in a 2U-high rack box), the LT 6000, and their
highly dense Rack Storage/12, a unit which set a disk drive density standard
which had it ahead of the competition for 1-2 years.
As
mentioned earlier, we do not believe NetServers will disappear from the
Intel server landscape (80% probability), but they will have a difficult
time regaining lost ground. However, we believe they can do it (65% probability)
if they decide to make the necessary changes (75% probability).
User
Recommendations
Despite
the challenges listed above, users should include HP in their Intel-based
server selection process. The users with the greatest affinity for HP
will be:
Looking
for solid systems, not necessarily concerned about TCO
HP produces well-designed systems that contain a fairly standard set of
features. Their products are generally solid, with neither ground-breaking
functionality nor performance. The LH 3000, one of HP's two new servers,
does provide more functionality than much of the competition, For example,
its storage and I/O capabilities equal or exceed that of the major competitors.
But with the rackable version 1U larger than the Dell PowerEdge 4300,
the user ends up losing some "density" of functionality, which increases
cost and TCO.
The
LH 6000's price/performance is a strong attempt to reverse this trend,
but it is only one product - HP needs to ripple this through its entire
NetServer line.
Looking
for one-stop shopping for a heterogeneous environment
As
mentioned often herein, HP has arguably the most diverse product offering
outside of IBM. This allows customers to buy from a manufacturer, not
just a VAR or consolidator, when they need all manner of computing hardware.
Customers should weigh the one-stop benefits against anticipated pricing
benefits from a vendor like Dell.
Although
Dell (for example) can provide Unix systems as a semi-custom order, it
is clearly not their main focus, so the lower upfront price may exact
a higher price down the road. Given that HP's Unix servers are considered
excellent, and HP-UX is one of the more robust operating systems, customers
have less worry when combining Intel and Unix through HP.
In summary: HP's NetServers are solid, reliable products, but have had
price and performance shortcomings. Whether the solid results from the
LH 6000 will be a one-time occurrence, or become a trend, customers must
monitor the situation and factor it into their decision-making process.