Would Intel benefit from following AMD's lead and spinning off its fabs as a foundry business?
This is argued in a most interesting piece by Jag Bolaria in the latest edition of the Linley Group's Processor Watch.
Bolaria argues:
That TSMC is making more net margin than Intel - 35% to 26%.
That, in the last decade Intel's market cap has been flat while TSMC's has grown 42%.
That between 2005 and 2010 Intel grew revenue 12% whereas TSMC grew revenue 72%.
That the foundry industry will grow from $30bn in 2009 to $50bn in 2015.
That Intel enjoyed a lead in process over TSMC dring the last decade but did not, on these figures, appear to benefit fully from it.
That potential foundry customers are deterred from using Intel becasue of competition concerns.
Bolaria's conclusion is that Intel is not reaping full value from its process technology, and that this value could be unlocked if Intel spun off its manufacturing operation in the same way AMD did with Globalfoundries.
I suppose that's one way to start killing off Intel
Yes it sounds like an idea concocted by the financial community, Dr Bob, probably the same guy who thought up fab-lite.
I guess it depends what you mean by "not reaping full value from its process technology"...
Intel has the highest performance CMOS process around but it's more costly than CMOS foundries -- which is the right decision for a company whose business is selling expensive high-performance CPUs.
I'm sure they'd get more revenue and profit from the fabs by opening them up to foundry customers who don't mind paying more than for TSMC wafers in return for better performance.
But if their competitors then use this to make competing CPUs they've shot themselves in the foot, since they would probably lose more by selling fewer (higher-value) CPUs than they'd gain by selling more (lower-value) wafers.
Well, Ian, I think the argument goes that when Intel sells a chip it sells not only process technology but IP, whereas when TSMC sells a chip it is selling only process technology. Therefore, it is argued, TSMC's profit margin on a chip should be lower than Intel's but, in fact, it's higher. That's how they get to the conclusion that Intel is "not reaping full value from its process technology".
Are we missing the point here. An Intel chip consists of process technology plus its chip design IP. On these figures, it is not only making less money from its process technology than TSMC but is also getting no recognition for the value of its chip IP. Is Intel charging enough for its chips?
No way Jose, I mean Peter, Intel charges too much for its chips - computers would be a heck of a lot cheaper if there were ten companies selling x86 processors.
The margin isn't the real issue -- to get an edge in the market Intel uses more expensive/advanced technology than the foundries, the payback is better performance for their processors, the cost is that they have to pay all their own process development costs.
So long as they make plenty of margin selling high-priced processors on an expensive process this works fine, if prices drop then they'll be in trouble.
But so long as they're protected by the x86 near-monopoly this won't happen, the market pays the high prices because of the software lock-in.
I'm not convinced how much cheaper PC CPUs would be with a lot more competitors, it costs a huge amount to develop an advanced x86-class CPU and if you split the market 10 ways instead of 2 and drop prices there's less money to be made split more ways.
Should Intel spin out its fabs? Absolutely yes; the day after TSMC and Samsung does!!
Well you could say TSMC's fabs are already spun off because they're a manufacturing operation unconnected with any other operation, Malcolm, and I wouldn't be surprised if Samsung -after their increasingly bitter disputes with Apple - don't see value in avoiding future potential conflicts of interest by putting their foundry business into a separate operation. And Intel is apparently saying OK to foundry involving x86 cores but not-quite-so-OK to foundry business involving non-x86 cores - a conflict of interest which could hobble their foundry business if, indeed, they decide they want to have one. So it's not an absolutely outlandish proposition.