Asleep At The Wheel?
A number of places announced a couple days ago that AMD was going to be delayed a few months on 90nm.
Well, we told you that a couple weeks ago.
That’s not due to any great investigative journalism on our part; as you can see from the link, AMD said so during their analyst meeting a few weeks ago. They didn’t say so too loudly, but they made it clear enough to anyone listening. Guess some didn’t.
More importantly, how significant is the delay? Hard to say, given two other tidbits at Comdex.
First, AMD had a dual 90nm Opteron setup at Comdex, but refused to disclose the speed at which they were running. You might recall the last time they did that, they were trying to hide the fact that they only had 800MHz Opterons available.
Sorry, guys, but a few of us have memories. When the news is good, you blare it out; when it’s not so good, you duck and cover.
Second, the biggest news (at least for this audience) wasn’t what was at Comdex, but what wasn’t: socket 939 boards. There wasn’t even mention of them, as best as I can determine.
Given that AMD was supposed to make a few hundred thousand of them in the first quarter of 2004; that’s a pretty notable absence, and a strong indicator we’re not going to see them anytime soon.
Is the apparent delay in socket 939 related to the delay in 90nm? Maybe, maybe not, but a delay in socket 939 all by itself will hurt AMD sales early next year.
In For A Penny, In For A Euro
AMD announced this week that it will build a 65nm fab plant in Dresden.
The fab plant is made possible mostly due to the national German and regional Saxony governments. They will give AMD about 20% of the (suspiciously low) projected cost of the fab.
The regional Saxony government and some private investors will invest in AMD to the tune of about 13% of the cost of the fab. The exact terms of the equity financing aren’t known yet.
About 30% of the cost will be covered by loans, most of which will be guaranteed by the German/Saxony governments.
This leaves AMD to come up with a bit less than 40% of the cost to come out of their (or at least somebody’s) pocket.
Why should you care about this? You should care about this because the most likely pocket AMD is going to get that $900 million from is yours.
Not that there’s anything wrong with this or that this should come as a shock to you. Eventually, the customers always pay for all the expenses of a company (where else are they going to get them from, the Fab Fairy?)
What is important to note is that if AMD has to come up with around $900 million out of 2004-2006 revenues. They can’t wait until the fab is finished, then pay off the $900 million with the new fab’s production. They’ll have to pay that $900 million from current production, and they can’t do that if current production loses money.
AMD is going to be under a lot of pressure to keep CPU prices over the next few years a good deal higher than if they got loans for that $900 million.
What is most intriguing about all this, again, is not what’s there, but what isn’t in this whole transaction. Where’s IBM in all this? You’d think they’d want AMD to use their fabs.
At the least, AMD apparently doesn’t want to get that close to IBM, or vice versa. Keep that in mind when people consider IBM to be AMD’s Fairy Godmother.