Business Week has an article about AMD which has one rather amazing revelation, and some perspective on what AMD can expect from Opteron.
SOI: A $46 Million Non-Problem
Did you know that AMD had to pay IBM 46 million dollars last November to help them get SOI working to at least some degree for them?
You didn’t? I didn’t either. Nor did anybody else, including the investors.
Yes, AMD did make mention of it. Do you know how they did it? They buried it in their restructuring charges. I quote:
Perfectly clear, isn’t it? 🙂
This happened well before AMD and IBM made their research agreement, that was early January.
After observing the company the last few years, I’ve noticed that AMD’s consistent policy has been to never admit having a problem until after it has been solved.
For instance, when the initial Thoroughbred broke its leg, AMD denied it was having any problems until they admitted, “Yeah, we had to redesign it,” after they had redesigned it.
With Hammer, AMD refused to admit any problem with that until just a few weeks ago. There were no problems at all until the end of January
What that means is when AMD doesn’t do something they said they were going to do, you can safely laugh at any “there’s no problem” statement. When you hear an AMD spokesperson admit to a problem, that tells you the problem has been solved.
Did that $46 million solve AMD’s SOI problems for good, or only up to a point? Who knows? And in all honesty, with their track record would you believe them if they say it did without a real CPU around to prove that they did?
What Was That About Windows?
The article also divulges that Microsoft will have Windows x86-64 ready and shipping by the time Opteron gets introduced in late April.
That’s certainly good news for any early Opteron buyers, but what about all this nonsense AMD was spewing just a few weeks ago about Windows not being ready until the fall. Server versions of Windows are normally a superset of the desktop version.
I think $46 million didn’t solve all of AMD’s SOI problems, and the delay on Athlon 64 is just a means to buy time to get SOI to run more quickly and likely also do a rethink on what Clawhammer is going to be like, because it’s starting to look like a 64-bit Duron.
What can I say? This company just can’t be believed a lot of the time. If they say something, believe it when you see it.
P.S.: Opterons and The Server Market
The Business Week article also talks about what AMD expects out of Opteron, and what analysts expect Opteron to mean to the bottom line.
AMD hopes to go from 5% to 10% of the server market this year, and 20% by 2005. These don’t seem to be unrealistic goals.
More importantly, the article cited analysts as estimating the revenue AMD could expect from its 10% share: $100 million. That may look great until you realize that AMD needs over $3 billion a year in revenue just to break even.
There’s no knock on AMD meant here; Opteron sales will be a nice profitable niche for them which they certainly should pursue. I mention this because some seem to have the notion that Opteron sales can sustain AMD all by themselves, and they just can’t.
About 130 million PCs were made in 2002. Only 4.5 million servers were built the same year.
Granted servers are more likely than PCs to have more than one CPU in them, but as AMD itself says, 80% of servers use only one or two CPUs.
The CPU market for servers just isn’t that big. AMD can’t possible meet its bills with server sales alone (neither could Intel). Hammer must also eventually do well on the desktop for AMD to survive and thrive (which is no news to them).