Computers 1963

As you probably already know, AMD’s revenues for the first quarter will drop by 15% rather than the seasonal 5-10% drop, and they’re going to lay off 10% of their staff.

It doesn’t take a rocket scientist to figure out why: last quarter they sold almost entirely ancient X2s at every lower prices, and sold very little K10-class CPUs.

What’s important about this is not that AMD lost money, or even that they’ll lose an extra hundred million more than previously expected. It’s important because it illustrates AMD’s core problem for at least the next year, and one that a layoff which will only save AMD around $60 million a quarter once fully implemented will not fix.

For AMD to do well from a new generation of processors, they have to hit three progressive milestones:

1) They must be able to make the processors.

2) They must be able to make a lot of processors.

3) They must be able to make a lot of fast processors.

Over the last ten years, AMD has usually had problems doing this. Normally, they don’t have a problem with step one, steps two and/or three have been the problem, which usually have been resolved by redoing what they probably should have done in the first place.

This time around with K10s, AMD had big problems with step one. It took them a long time to fix that, and now they have, but it’s hardly a triumph because they have two more to go.

Right now, they aren’t capable of step 2, and probably no longer plan to except as a backup plan if 45nm does poorly. Why revamp your fabrication to make 2.3GHz-2.4GHz K10 duallies for a short period of time if you can make 2.6GHz-2.7Ghz X2s with no extra investment that perform just about as well (especially when you’re not going to get paid diddly for either)?

AMD has flat-out given up on step 3 for 65nm; they’ve said B3 will be the last 65nm K10 stepping.

Given the circumstances, these aren’t bad decisions by AMD. When you have no money, it’s perfectly reasonable to stop tossing good money after bad.

So they’ll make a relative handful of 65nm K10s (almost certainly less than 10% of production) just to make it look like they’re achieving something, try to make a few extra dollars, and try to get it right with 45nm.

So there certainly will be no CPU-led recovery for AMD from 65nm processors.

Some think and have even written me to suggest that AMD could have a graphics/mobo led recovery. Sorry, but no, and here’s why:

In the combined AMD/ATI, the ATI segments only account for about 25% of revenues. Since the merger, these sections have done a little worse than breakeven operationally.

However, if you include the interest from all the debt AMD incurred to buy ATI in their results, ATI starts looking like an anchor to AMD’s boat. Even if ATI starts doing considerably better, all that will happen is that ATI will stop being an anchor. It will not and cannot become AMD’s main sail, that would be like the tail wagging the dog.

Nor could it be reasonably expected for the former ATI to become a sail. The main point to buying ATI was to exhance CPU sales by providing the total package to system builders. That’s not a bad idea in principle, but AMD paid way too much for what boils down to a nice-to-have (AMD could have built two new fabs for what they paid for ATI), and having first-rate mobos doesn’t help much when you have second-rate CPUs.

The best thing AMD can do for the ex-ATI is to get the CPU business flying right.

But how likely is that to happen?

Assuming nothing goes seriously wrong, AMD should be able to get the first 45nm chips out, with a top speed of 2.8-3GHz, by the last quarter of the year.

Fine, that’s step one. How about step two? Again, being optimistic and assuming historically normal rampup rates, 45nm chips won’t start making a significant impact on AMD’s sales until the end of Q1 2009, and wouldn’t approach crossover until about the middle of 2009.

So “recovery” in the form of plentiful 45nm chips is a year away, even if you completely ignore the very inconvenient fact that Nehalems give every indication of maintaining the current performance gap between Intel and AMD and keep even new 45nm chips in the bargain basement.

This is why I’m pessimistic about AMD’s future prospects. Getting to step one is just showing up for the game; steps two and three are actually playing it. You don’t expect the fans to start carrying around a player in triumph just for showing up on the field.

It’s looking at the prospects of steps two and three that make me pessimistic about AMD. It’s not because I’m assuming the worst when I look into the crystal ball, it’s because what I see looks ugly even when I assume the best (consistent with the available evidence).

Do you know what’s really frightening about all this? Looking back at what I’ve said about AMD and the K10s, I’ve been consistently wrong about the company: I’ve been consistently overoptimistic. I’ve said, “This is going to be bad,” and they’ve proven me wrong by doing worse.

Again, looking down the road from what will probably prove to be an overly optimistic viewpoint, I don’t see how this company can continue to exist in its current shape and form for another twelve months. Push is going to come to shove, they don’t have the money to last another year, and the current condition of the financial world precludes them getting any more by conventional means.

I’m not saying they’re going to go out of business in the next twelve months; I am saying something radical has to happen in that time.

We’ll talk about that tomorrow, for now, I’ll just say that the why is more important than the what.

Ed


Be the first to comment

Leave a Reply