appliances (also called "PC appliances" and "Internet appliances") are
the latest name for the type of computing similar to that pushed by Oracle
CEO Larry Ellison a few years ago under the name of "Network Computer".
Whatever you call it, this market is new and growing. In recent months,
most of the large hardware vendors (Compaq, Dell, Gateway, IBM) have started
to announce products or plans for products. In addition, COMDEX '99 highlighted
a slew of new PC appliances, from Microsoft, Intel, AMD, IBM, and others.
market has grown due to the desire to reduce the price of computing, primarily
from a hardware viewpoint, but also with respect to desktop support. Hardware
cost reductions are expected to come from lower equipment and maintenance
prices, support cost reductions are expected to come from the centralization
and standardization of application software available to the user/client.
A subtext to this is an attempt by some vendors to move away from a Windows-centric
computing model and loosen Microsoft's control of the desktop. Presently,
the primary target buyer for these products is the business/corporate
user looking to control cost at the desktop. The secondary target is the
consumer market segment where budgets may be limited.
market is immature and pricing is still largely undefined, but estimates
run from $299 for a low-end model from Netpliance, up to more than $1600
for a fully loaded Dell WebPC with monitor. The sales cycle is still largely
unknown, but we expect it will be equal to or shorter than that of a traditional
desktop machine. The percentage of revenue derived from service will be
smaller than for a traditional PC, since these products will be significantly
less expensive than traditional desktop devices, as well as less complex
and more "commoditized".
do you get for your money? Presently, there is no single answer. Some
have suggested that the information appliance market can be segmented
into three groups: stripped-down PCs, Internet appliances, and thin clients.
Although we understand the rationale, we hesitate including thin clients
in this area, except as a comparison. In general, appliances include a
CPU, operating system or browser (or both), memory, network card, keyboard/mouse,
and sometimes a media device (CD, floppy, or hard drive) and (less often)
I/O slots. However, as more storage and I/O capability is added, it becomes
an anorexic PC, not an appliance. Appliances are generally small - a typical
footprint is 6" wide and 10" deep.
you generally do not get is a graphics card, more than one
or two I/O slots (if at all) or more than two drives of any type (HDD/CD/DVD).
present market size approximately $600 million [Source: IDC], but we expect
it to grow beyond $5 billion by the end of 2002, a compound annual growth
rate (CAGR) of more than 100%.
is an emerging market, the "failed" Network Computer effort notwithstanding,
with no clear leader yet. However, Compaq has already announced its Windows-based
iPAQ Internet device, and Dell and Gateway are expected to ship Windows-free
devices in early 2000. These companies will have a head start on the rest
of the field. In addition, if Gateway leverages its relationship with
AOL and its tremendous network, it will gain instant market share - as
long as it ships a product within six to nine months.
As mentioned above, Compaq had a head start on the most of the rest of
the field, as the business-focused iPAQ started shipping in mid-January,
2000. Since Compaq has recently lost PC market share, this move should
help it regain market share, as well as restore lost sales revenue. The
unit's small footprint (6.4" by 10.5") means a customer can keep it on
the desktop, away from dust and within easy reach, and its "MultiBay"
hot pluggability and compatibility with its counterpart in the Armada
notebook line should allow easy transport of data between the notebook
and appliance. Compaq is also focusing on "deployment speed" (how quickly
a customer can be up and productive) and extended product life (up to
a year vs. the typical 90 days for a Celeron CPU), although we do not
necessarily see the one-year timeframe as being a "killer" benefit.
the negative side: Compaq announced this product as an Internet Device,
but this appears to be just a stripped-down PC. For example, the base
model ($499) has a 4.3GB hard drive, the system can take up to 256 MB
of SDRAM, and it runs on an Intel Celeron with an Intel 810e chipset.
Compaq's use of the term "legacy-free" is limited to: no PCI/ISA slots,
and USB for the keyboard/mouse. Although it is small and low cost, those
two factors alone are insufficient to make this an appliance.
Dell has recently started shipping its "WebPC", marking its consumer-oriented
entry into this market. Dell's traditional strength, delivering low-cost
product of above-average quality, will give it a built-in advantage. The
original base price of $999 was high enough to make this product fall
into the realm of "stripped-down PCs", rather than Network-Computer-like
appliances. However, Dell has now reduced that to $699 (including monitor),
making it now competitive with Compaq's iPAQ vis--vis price and functionality.
We still viewed it as a re-packaged PC, though.
Gateway presently has the #4 PC market share position, but focuses more
on the consumer market than the corporate market. However, its recent
alliance with America Online (AOL) should provide it with access to the
largest ISP in America. It will be up to Gateway to provide the low-cost
hardware for the appliance. One area of weakness: Gateway's servers (gained
from their acquisition of ALR) are not considered serious market share
contenders in corporate America.
IBM's thin client Network Stations have been around for some time, but
IBM is also developing an appliance under the EoN codename. The Network
Station series has good features and emphasizes network/communications
and versatility, with a minimal focus on storage media. However, the price
structure is high, with the Network Station costing approximately $999.
As mentioned earlier, IBM is expected to release a new appliance early
in 2000, but neither features nor pricing are available at this time.
Currently, Netpliance's consumer-oriented "I-Opener" is one of the few
true Internet appliances. It does nothing but connect to the Internet
- no Windows applications, or anything else. If all you want is Internet
access, this will do it for $99/$199[See Note] plus a dial-up fee ($21.95-$26.95
per month). It has additional limitations, such as dial-up being the only
access method. For the consumer (i.e., home user) only wanting to surf
the 'Net, this may be enough, and the price is extremely attractive. However,
we feel this is not a viable long-term strategy, and expect that Netpliance
will enjoy about 12-18 months of independence before losing market share
to one of the big players, or getting bought.
Netpliance recently changed the price from $199 to $299 to $99 (special
deal through 6/20/00) then back to $199. Netpliance has found that demand
for the I-Opener is "not inelastic", which in English means that which
means that the market was willing to pay $199, but not $299, for the I-Opener's
Although Oracle/Liberate was an early, vocal proponent of this general
computing style, we believe they will be at a market disadvantage. Liberate
was among the first players (under the name Network Computing, Inc.),
but its $1 Million in revenues in 1998 are still minimal. In addition,
although Larry Ellison has been very vocal about the NC, there is no hardware
product from Liberate other than the TV-top unit, which is more like a
cable TV unit than a PC/NC.
Oracle has "spun off" the New Internet Company, with its ostensible purpose
to develop and ship an Internet appliance. Finally, we believe Ellison's
focus (sometimes characterized as an obsession in print media) on beating
Microsoft and its CEO Bill Gates, will distract Oracle from focusing on
the job at hand. [Note that although Liberate is in fact a separate company
from Oracle, its genesis is from Oracle, and Ellison is on Liberate's
board of directors.]
Hewlett-Packard had announced their Entria line of thin clients. However,
the Entria has now been cancelled due to HP rethinking their strategy
- they'll be focusing more on delivering product through their NetServer
Division. Although we commend HP management for killing a potential loser
before committing significant resources to it, the delay in entering this
market will hurt HP's chances for grabbing a large share of the market.
But, given HP's ability to deliver quality products, we believe it will
still have a credible entry in this space - if it can deliver within six
to nine months. [HP's recently-announced "e-Vectra", its corporate entry
into this space, came too late to be included in this article. This product
is due to ship in April, 2000.]
Like but Really Isn't" Market
Terminals (as in terminal/server computing, precursor to client/server
computing) are not really Internet appliances, and probably should not
be called information appliances. They can be small and feature-poor (similar
to some I-appliances), but they run off of a mainframe or large server,
as opposed to connecting to an ISP or ASP. Note that there can be some
terminology overlap, to wit: isn't the ISP/ASP acting as the server? We
make the distinction that an I-appliance does not require a centralized,
in-house server to work, and has an Internet component as a key feature,
rather than as an afterthought.
Microsystems: The "Sun Ray" does not fit the term "Internet appliance".
It is small and relatively inexpensive, but is essentially a dumb terminal,
having only enough "intelligence" to repaint/refresh screens (when commanded
by the server) and interpret keystrokes. This is evidenced by the need
to connect to the server to run any programs/applications - there is no
OS resident in the box, only a Java Virtual Machine. In addition, except
when connected to the 'Net through the main server, there is really no
Internet component to this unit. Although the Sun Ray does have some interesting
features, such as its "Hot Desk" feature, we do not feel it fits here.
market will grow significantly over the next 3 years. We believe that
Internet devices/appliances will ship more than 5 million units by 2002,
for a CAGR of more than 100%. This growth will come at the expense of
traditional PCs. We also believe that this market will help grow the rental
PC-software market (also called "subscription computing"). Microsoft's
announcement that it will start renting software meshes very well with
the Internet appliance and Application Service Provider (ASP) model.
believe a large part of this growth will be driven by a herd mentality,
catalyzed by: (1) financial managers looking for a PC infrastructure cost
"magic bullet"; and (2) Microsoft pushing the rental model for application
software such as Microsoft Office. We do not believe this is due to this
computing model being fundamentally better than the existing model(s).
Failure (10% probability): This would result from a combination of excessively
high pricing and excessively slow service delivery through the network
or excessively bad performance. Complete Acceptance (<2% probability):
This would result from Microsoft abandoning its present licensing structure
(i.e. applications sold/licensed through retail, one to a machine), combined
with the failure of alternatives such as Linux (an suitable applications)
to penetrate the desktop market. Partial Success (60% probability): In
this scenario, the greatest acceptance will be in corporate managed computing
environments and in consumer low-end markets. Mainstream computing (as
it exists today) will exhibit a slower acceptance rate than these niches.
Big Success (25+% probability): Would require a major attitude shift in
the computing marketplace, combined with major acceptance by those (potential)
users who currently have no access to computing and the Internet.
of Likely Success:
mentioned above, we believe the areas most susceptible to acceptance of
this computing model will be the low-end consumer market and the managed-PC
corporate market. We believe resistance to computing-model change will
delay acceptance (though not necessarily implementation, i.e. through
management dictum) in other areas.
Vendors should provide both Windows-based and non-Windows appliances,
and make each as low-cost as possible without sacrificing quality. [Of
course, the regular Windows license fee, combined with the hardware overhead
required, should drive the price of a Windows-based machine $50-$100 higher
than a Linux-based machine, even more if running Windows 2000. This would
drive Windows-based appliances toward the Windows CE operating system,
rather than Windows 98/NT/2K.] However, the most important factors at
present are time-to-market and grabbing market share.
believe that Linux will find another home in these appliances, and we
also expect BeOS will gain market share at Windows's expense.
suggest that vendors try not to have "one box fit all". This leads to
market confusion ("It's a floor wax! No, it's a dessert topping! Wait,
it's both!") which leads to lost sales opportunities.
should pay attention to this market because it will grow significantly
in the next three years. Since this computing model is not for everyone,
customers should analyze their computing needs thoroughly - it is a mistake
to assume that all end users will accept this model with open arms, no
matter how beneficial management believes it to be. The effects of end
user resistance (and training) are rarely given sufficient attention by
management, so users should exercise caution. Spending extra time before
committing to this course of action will prevent false starts, or, in
extreme cases, dumping the network device model and returning to the "traditional"
server/thick client model.
who find that many of their users are so-called "power users" - using
many different applications, or applications which are considered "compute
intense" - will find the appliance model unsuited to their needs. Customers
whose users stick to the basics - e-mail, web surfing/research, some mid-range
word processing or spreadsheet usage - will find greater use for this
the user has decided that the network appliance model is the best solution,
the type of device (or devices) should be chosen. Customers need to determine
if fully functional Windows is a necessity, or if Linux, Unix, or Windows
CE will be sufficient. (The distinction is made, not to imply that Linux
or Unix is less functional, but because of the applications and market
share present on the corporate desktop.)
the device type has been chosen, users should focus on whether the applications
will be served from within their company, or from an external ASP or ISP.
If the choice is an external A/ISP, then the user should determine which
package/suite of applications and functionality bests suits his needs.
Presently, with the exception of Micron (who has no thin client offering
per se, only a PC with "subscription computing"), there is no provider
market clearly defined. If the applications will be provided internally,
then the user needs a vendor who can provide a full client/server solution
bundled with a high bandwidth network infrastructure. The only large vendors
who provide that currently are Compaq, IBM, and Dell.