USS Clueless              Voyages of a restless mind

Main: normal long no graphics Contact Log archives Best log entries Other articles Site Search

Stardate 20010801.0831 (On Screen): Motorola, one of the biggest high-tech companies on earth, is in deep trouble financially, and the running wound of its operation is its semiconductor division which has lost a whopping $800 million in just the first half of this year. I think what we're really seeing here is the death of the old model of the semiconductor business. The new model, which is becoming increasingly common, is to divide the business into foundries and fabless designers. Taiwan Semiconductor Manufacturing Company (TSMC) is an example of a big-name foundry; they don't design chips, but they do make chips for other companies -- lots of other companies, including ones you've heard of. The foundries concentrate on what they know: how to take masks and make good chips from them. The designers, on the other hand, have the ability to start small and grow as their business justifies it. This means companies like nVidia, Via or Qualcomm, who know a specific business and know it well, and can create the designs but can't afford the immense capital investment involved in owning their own fab. Fifty or a hundred companies like this effectively spread the gargantuan cost of the fab by contracting with the same foundry.

That's the problem: fabs have reached the point where no companies can actually afford to operate them exclusively for their own products. One of the few which does this successfully is AMD, but they've done it by staying starved. AMD makes its own chips, but it also contracts outside for a substantial part of its production, so it is saturating its own organic capacity and even if it faces an economic downturn in business its own fab will remain saturated. Other major semiconductor companies like IBM and Intel create a lot of chips for themselves, but they're also major foundries and because of this can spread the capital expense around.

Until now, Motorola has been designing its own chips for things like cell phones and using them exclusively in-house. Their new plan is to also sell those chips outside, but the industry is skeptical about this kind of conflict of interest (I know about this first hand) and I'm not sure I believe that this is actually going to bring in $billions in new chip business. Equally, the PPC is reaching the end of its useful life, being supplanted in embedded by the ARM (which Moto recently licensed) and on the desktop by the amazingly resilient x86 architecture (which is also at the end of its useful life, soon to be replaced by Itanium and Sledgehammer and ARM). Volume in the PPC isn't high enough to justify the awesome expense which would be required to keep it competitive, so it has been languishing.

If there's any solution to Moto's semiconductor woes, part of that is going to have to be getting into the foundry business big time (as Intel is) to subsidize the capital cost of equipment. It may also include a radical change in business model, and abandonment of entire product lines or even businesses, and a substantial deemphasis on in-house designing. With the recent announcement by Palm that it would port its OS to the everpresent ARM, Moto will no longer have the inside track on the PDA business. Moto is going to make ARM-based PDA chips, but it will be competing with Intel and probably with Palm itself, who could go to foundries and make itw own chips. I see no future for Motorola in the PDA chip business and I predict they'll be completely out of it (except as a foundry) within two years.

I don't know what the complete answer might be for Moto's chip business (or even if there is an answer) but they are going to have to completely rethink their business model. Their current business model is the one which was successful in the 1980's, back when all chip companies owned their own fabs. That was possible when a fab cost less than $100 million. When a modern bleeding-edge fab can come in above $3 billion, it's no longer viable. (discuss)