Before I tackle the future of Intel and what needs to happen post Otellini, let's get some market context. Let's say it's 2002, and you're pondering your investments. You've got a few hundred dollars to throw around, and so you invest in the high-tech blue chips IBM, Oracle and Intel. For your last hundred, you go the risky route and buy ARM Holdings.
Flash-forward to current times. How did you do? Your $100 in IBM is now $245--nicely done. Your $100 in Oracle is now $315--no complaints there. Your Intel investment didn't do so well--it's worth just $107 now. But ARM more than made up for your bad choice with Intel, because that $100 is now worth $1,200. That's right, ARM up 1100%, Intel up 7%.
You're probably saying that's not fair. Intel is a well-established company, ARM is more of a startup--of course ARM will outperform. Let's test the notion that ARM is a startup. First, it's been around too long to be considered a startup, but, even more important, look at market capitalization: Intel is worth about $100 billion, and ARM is worth $50 billion. Yes, according to market sensibilities, ARM is worth three times what Dell is, about 80% more than VMware and about the same as EMC--and the company doesn't actually make a thing. That's some darn good intellectual property there.
There are few lessons we can take from this. The first is the gross, insane nonsense of market capitalizations for high-flying tech stocks. Sure, ARM gets a ton of credit for being in the right place at the right time, and it's doing a masterful job of providing the right technology to its customers, but a $50 billion market capitalization seems a bit extreme.
Back to Otellini. Intel finds itself in the ironic position of being on the wrong side of Moore's Law. Gate sizes are now so small and performance is so good that for most applications, the market will no longer reward Intel for squeezing more performance onto a single chip--whether it be through more cores, or added cache, or even bringing some of Intel's other parts like memory managers and I/O controllers on board.
In fact, the market hasn't been rewarding it for years. While ARM was in the right place at the right time with low-power designs, its licensing model is the real secret. Licensing allows others, including Nvidia, Qualcomm and Texas Instruments, to use some of the chip real estate for what they do best--like video, radio and sound processing with specialized circuitry. Intel not only was terribly late to the very-low-power CPU market, it also failed to see that, because of Moore's Law, the rules had to change with regard to its customers. Intel must now find a way to let others stake out some on-chip real estate.
Whether Otellini sees this or not is anyone's guess, but the enigma of the market is surely one driving force in his early retirement (for all I know, his golf handicap may play a much bigger role). It's clear that Intel's vast investments in developing new processes and the intense capital investment in retooling for each new fabrication process every two years or so is seen by the market as a bad thing.
One radical idea is to break Intel into two parts: a fabrication house and a design house. Of course, there's no guarantee that such a move would produce better results--there are plenty of fabless semiconductor houses that struggle to make a profit, too.
While Moore's Law takes away (killing old business models and design goals), it also brings new opportunity. The challenge for Otellini's successor will be to change its business models just as quickly.