When Microsoft announced at CES that Windows will be available on
ARM, it was the result of Microsoft lacking a modern tablet OS and
Intel lacking a low power x86 chip and major design wins in mobile.
It's the usual divorce, in that Microsoft and Intel need to work
together for the sake of all the Wintel offspring, but both need other
partners to live happily ever after. It also gives Intel the
opportunity to design a new CPU on a clean slate that doesn't need to
support the x86 legacy.
Windows Is in Decline
x86 has grown up around the needs of Windows and Intel, but the PC -
at least in the developed economies - is a market that has peaked. From
now on, as the sales of iPads and other tablets ramp up, PC sales will
fall. iPads and Macs have already sucked all the growth out of the
Wintel PC market, and the more people like using touch on phones and
tablets, the less they will want to use any of the current versions of
Windows on PCs.
Also, Intel's current approach of shrinking Atom to a 24nm design to
reduce power doesn't seem to be picking up many new customers, so it
should look at an approach that offers more for mobile.
iOS alone will require about as many ARM
processors as the number of x86 processors Intel makes for desktops and
laptops.
Intel already has Apple's x86 business, but in calendar 2011 that
will be at most 18-20 million CPUs. This year Apple will probably sell
more than 80 million iPhones, 35 million iPads, and 25 million iPod
touches - the 2011 iOS opportunity should be over 140 million CPUs. If
the September refresh of the iPod nano makes is an iOS device too, add
5 million to that, and in 2012 Apple should sell more than 200 million
iOS devices. iOS alone will require about as many ARM processors as the
number of x86 processors Intel makes for desktops and laptops.
A New Apple/Intel CPU Makes Sense
That is why it makes sense for Intel to design a new CPU with Apple.
Indeed, Intel should forget about Windows and design the most efficient
chip it can for iOS and OS X. Intel can treat the x86 market as a
cash cow and keep pulling in the money while its new 3D transistor
design and new 22nm process makes AMD uncompetitive for at least a
couple of years.
All the design wins with the new CPU will be with other operating
systems. It needs to be more power efficient than the current Atom and
offer better throughput than ARM, so the future of this is a commodity
chip that can run iOS and Android better than ARM, and run OS X
and Linux (particularly on server versions) better than x86, and
ideally allow HP to migrate its Unix business and leave behind all the
costs associated with Itanium.
If it has the same throughput as x86 but uses less power, it will
make inroads in computing centers and high performance computing, where
power consumption and the resulting need for cooling is a major
constraint; if Intel doesn't go this route, it risks losing the power
efficient server market to ARM based designs.
Windows on ARM
There is also currently no large financial penalty for Microsoft
making Windows run on ARM (WARM). The expenditure of time, effort, and
money will be worth it if WARM delays and reduces corporate defections.
However, if this new Intel iOS/OS X chip gets traction, Microsoft will
be forced to react or risk more erosion of its IT user base. As it is,
the more IT goes mobile, the more IT people are getting used to
non-Microsoft solutions. This is why Apple is starting to make moves
into the enterprise and building briefing rooms into Apple Stores.
With this new chip, Intel and Apple could be seen as a newer, safer
bet for sizable parts of the future of corporate IT.
This new chip could also attract Nokia, RIM, and HP's WebOS for the
phone/tablet market, as it gives them a way of showing better
performance and can split - and therefore further fragment - the
Android manufacturers.
The Potential Market for a New CPU
Nokia is still the largest cellphone manufacturer in the world,
typically selling over 100 million a quarter, even though it has little
US presence and has lost a large part of its smartphone market. But for
the fallout over the move to Windows Phone 7, away from the partnership
with Intel for MeeGo tablets, Nokia could have been in Apple's position
to drive this chip design forward. The volume is there now, but
management needs to concentrate on the OS transition, if Nokia is to
survive as a major manufacturer, and as the largest Windows Phone
customer for the foreseeable future, it can persuade Microsoft to port
Windows Phone 7 to a new CPU.
RIM, although it is the well established leader in a very profitable
niche messaging/email market, has problems too. It won't reach its
targets for the quarter and is losing market share in the US to Apple
and Android. Like Nokia, it is moving to a new OS, but as QNX is a
Unix, it should port effectively to this new Intel CPU. At current
volumes of Blackberries, RIM would take over 50 million CPUs without
any PlayBook success.
HP, since it took over Palm, has done little to commercialise WebOS.
As it profits so little from PCs despite being the No. 1 vendor, it
makes sense if HP sees more of its future away from Windows. While
WebOS can be part of that future, it has a lot of catching up to do to
become a large enough market to attract many developers.
Android was always Google's best way of blocking Microsoft from the
mobile market and continuing to control mobile search. Google would
therefore want to make a good port of Android available as soon as any
Apple exclusivity for the new CPU ends, and those Android manufacturers
without a strong contractual commitment to ARM chip suppliers, like
HTC, LG, and MMI (Motorola Mobility), will move as quickly as they
can.
What About Samsung?
However, this would cause difficulties for Samsung, now Nokia and
HTC's biggest rival. What will Samsung do with its foundry business,
which produces ARM chips? Is Samsung, as a vertically integrated
manufacturer and components manufacturer, willing to hand over a key
component of its phone business to a design partly controlled by Apple,
a key rival, or will it concentrate on using its own OS, Bada, on ARM
and look to keep up market share and profits with that.
Whatever decision Samsung comes to, it will lose ground and revenue
and Apple's ARM business. With the new fab in Austin, TX supplying up
to 50% of production to Apple, a move to Intel could leave Samsung
stretched. Apple currently buys over $6 billion worth of components a
year from Samsung, but that isn't enough to stop lawsuits between them.
While Samsung wants to be the biggest cellphone manufacturer in the
world, it will pay Apple to source elsewhere and stop subsidising a
major competitor.
Disrupting the Market
Why would Apple spend time, effort, and money helping Intel produce
a new chip, which would then be sold to competitors too?
Because this chip can disrupt the market.
Currently Apple licenses reference designs from ARM and customizes
them for iOS to get better battery life and performance where needed,
but all ARM architecture licensees can go through a similar tuning
process for their designs. As this includes iPhone's major competitors
and the ARM foundries, iOS advantages on ARM are likely to be
temporary. A new chip designed for iOS and OS X that outperforms ARM
and gives good battery life should give Apple a much stronger advantage
for at least a few chip generations.
With a similar agreement to that made for Thunderbolt, Intel could
own the intellectual property and Apple have an initial period of
exclusivity. This would provide time to show off the advantages of the
new chip, let Intel pick up design wins, and give competitors like RIM
time to port and test their operating system on the new CPU. As the
design evolves through future generations, building the testing around
iOS and OS X would keep an Apple advantage. Apple, though, would
probably keep its ARM license, so there would be an iOS port running on
the latest designs. This would keep Intel from raising the chip price
too much and keep it developing the new design.
Apple would also be guaranteed the lowest price of any buyer, as
well as substantial quantity discounts, and with Intel's eight
worldwide fabs available, Apple's production lines would no longer be
at risk from lack of CPUs.
Could It Work?
Making a project like this work is to a large extent cultural. Both
Intel and Apple are engineering companies, and both are dedicated to
making the best designs. They already have a successful collaboration
to build on (Thunderbolt).
The only reason Wintel worked so well for so long was because of the
revenue it brought Microsoft and Intel, and money papers over many
cultural differences in business. But Windows is now a fading platform,
Intel can make even more money with Apple without the friction of
working with a marketing culture like Microsoft.
Of course, as the platform gains traction, Microsoft will want to
come on board to protect Windows and Windows Server revenue. Their
biggest difficulty will be accepting the role of junior partner, with
companies where they used to drive the agenda.
