AI
Thursday, February 03, 2011 3:59:44 PM
ARM's business model is a very interesting one to me. I don't see it requiring near as much overhead as Intel's.
Low overhead but also a tiny share of revenue. ARM gets ~1%
royalty for chips sold with its processor cores in it. That $20
processor in an iPad is $19.80 for Samsung (the guy with the
fab!) and 20 cents for ARM. OTOH the $34 Atom in a netbook
is a full $34 to Intel for a chip with a marginal cost likely under
$5. How about them apples?
ARM thrives in nascent markets where no one in sure there how
big the market really is and what features are really needed and
which are useless costs. OEMs can try different things with ARM
based SoCs and see what sticks. But if a new market is real
enough, big enough, well defined enough, and where compute
performance is a important differentiator then look out for Intel.
Intel's two core competencies are designing high performance
processors with a lot of arcane custom circuit and physical
design content, and having the fastest, 100% MPU oriented
logic processes and fabs to build them in. Intel's model is too
expensive and has too long a lead time to make fully dedicated
chips for markets that suddenly appear or are unproven but
once the market is shown to be real and persistent then it can
bury competitors with a pipeline of ever better products.
Low overhead but also a tiny share of revenue. ARM gets ~1%
royalty for chips sold with its processor cores in it. That $20
processor in an iPad is $19.80 for Samsung (the guy with the
fab!) and 20 cents for ARM. OTOH the $34 Atom in a netbook
is a full $34 to Intel for a chip with a marginal cost likely under
$5. How about them apples?
ARM thrives in nascent markets where no one in sure there how
big the market really is and what features are really needed and
which are useless costs. OEMs can try different things with ARM
based SoCs and see what sticks. But if a new market is real
enough, big enough, well defined enough, and where compute
performance is a important differentiator then look out for Intel.
Intel's two core competencies are designing high performance
processors with a lot of arcane custom circuit and physical
design content, and having the fastest, 100% MPU oriented
logic processes and fabs to build them in. Intel's model is too
expensive and has too long a lead time to make fully dedicated
chips for markets that suddenly appear or are unproven but
once the market is shown to be real and persistent then it can
bury competitors with a pipeline of ever better products.
