That rumor about an IBM takeover of AMD has proven to be a rumor run amok.
Nevertheless, it’s probably a good time to think about the underlying issue a little harder, a little deeper.
What would happen if some technology company with very deep pockets, an IBM, a Samsung, a Microsoft (or really, anybody with such pockets) figured out how to buy AMD.
For now, forget any impediments and roadblocks. Just assume Intel Goliath wakes up one morning and instead of seeing David, he sees another Goliath giving him pit bull eyes.
What would happen? It would take a while for the new Goliath to get cranking, but in the medium-term (say three-five years), the following would happen:
No doubt you’d love the first, even less doubt you’d hate the second and are in a state of denial at the moment. Many of you are saying, “No, competition always gets you better products for lower prices.”
That’s not so. Competition usually gets you better products for lower prices provided the activity stays profitable enough.
The irony of the CPU industry is that innovation for it is the opposite of those its products enable for others. Innovation is not nimble, quick and cheap; it’s lumbering, prolonged and expensive.
Even the tiniest change to a CPU takes months to implement. A somewhat serious redesign takes years. Process shrinks take years. You spend billions in R&D, and even more billions and years building the factories to implement the R&D.
What do these little things cost a place like Intel? Well, R&D and capital expenditures will cost Intel about eleven billion dollars in 2008, and most of that is spent paying for those little innovations.
And those costs are rising.
No need to pass the hat around for Intel, but if their prices went permanently down 20-25%, they would be a breakeven company, and wouldn’t be able to afford spending eleven billion and rising every year.
A true duopoly would do that, and initially, probably by a lot more than that, for a while. If, say, Samsung built enough capacity to supply 50% of the world’s CPUs, and Intel kept up with its 80% capacity, we would see what a real CPU war would look like. It would likely put the periodic memory price wars to shame.
The problem is, given the current nature of making the little beasties, such a price war would be nuclear; there would be no winners.
Some of you might say, “Cut costs.” The problem is you can’t where it counts. Yes, there’s probably some fat in Intel’s budget, but the core problem is the nature of the CPU manufacturing process itself. You can’t lay off a fab, or cut its pay.
Neither AMD nor Intel like spending billions on R&D and $2.5-$3-$3.5 billion on a fab, they spend it because they can’t figure out a way to do it for less. A truly duopoly would leave the costs the same, but drop the revenues.
What would happen in a true duopoly, a Goliath vs. Goliath contest, is that after some spectacular fighting, and even more spectacular losses, the two sides would toss all their R&D efforts into coming up with a new, cheaper way to make CPUs and put everything else on the back burner.
That will probably mean coming up with an entirely different way to make something that acts like a CPU, and that’s not going to be easy. We’re not talking redesign, or new architecture, we’re talking about starting from scratch.
We’re probably not all too far from that happening, anyway. At some point, the increasing costs of building CPUs the current way will force that; competition would just make it happen sooner.
I’ll put it another way. CPUs are essentially made the same way today as they were ten-fifteen years ago. I don’t see that being the case ten-fifteen years from now. Not unless everyone gives up on making better CPUs with the current process, and can’t come up with a new way to build CPUs.
I’m not trying to say that this is something terrible that should be prevented. Decades from now, it would probably be considered good.
But a Goliath vs. Goliath fight wouldn’t be more of the same AMD-Intel fight, but better. Ten, fifteen years ago, that would have been the case, but not now.
What would happen today would be far more disruptive, unstable and destructive than that, not because of competition per se, but because the underlying technical and financial infrastructures of current CPU manufacturing are nearing an end, and there’s no easy successor in sight.