AMD Defabbing . . .

There’s a sentiment among some financial types that what AMD is going to have to do is largely or entirely outsource its fabbing to third-parties.

Outsourcing is a big buzzword these days, and it may work very well in other situations, but does it make any sense here?

First, we must ask, “What is the point of outsourcing?” The reason why businesses outsource is because they believe that they can hire someone else to make or do something profitably for less money than if you did it yourself.

There’s two situations where this can be so:

  • The outside hire can pay people a lot less to perform a labor-intense task competently than you can and/or
  • The outside hire can use economics of scale to do/make something for a lot of people who need something done, but not a lot of it.

    A typical example of the first is making clothes. A large part of the cost of making clothes is labor, so if you can get the job done with people making $1 a hour rather than $10, that’s a big cost saving.

    A good example of the second is tax preparation. You need to handle your taxes, but not very often. Unless your situation is very simple and easy, it’s better to hire a professional who does this every day than taking the time and effort to learn all the applicable tax law yourself.

    Do either of these situations fit AMD?

    1) Saving on labor: Fabbing CPUs is not a labor-intensive activity, the process is mostly automated. More importantly, the cost of labor is only a small percentage of the value of the product made. When AMD had just Fab 30, it made CPUs worth $4-5 billion with about 2,000 fab workers. If we use a ballpark figure of $100,000 per employee, that’s $200 million labor for $4-5 billion of product, or about 5%.

    So even if an Asian fab paid its workers much less than that paid in Germany, the cost saving would only be a few percentage points, which would be immediately eaten up by any profit made by the Asian fab.

    (We’ll note that assembling the CPUs is rather more labor-intensive for the value added, and both AMD and Intel have long done that in Asia for the cost savings.)

    The real cost of fabbing isn’t the labor, but the cost of the fab itself. For each worker to make several million dollars’ worth of CPUs each, you have to spend $2-3 billion first to give them a place to work, then continue to spend a lot more to keep it updated, only to basically scrap/dump the outdated equipment within ten years.

    If you pay somebody else to do your fabbing, they don’t get the fabs for free. They’re the ones who’ll have to lay out the billions, and the only way they’ll be able to pay that kind of money is to include the cost of all that spending is to charge its customers more, and then some.

    If AMD outsourced its fabbing to other people, they wouldn’t have to invest those billions in fabs, but they would end up paying more for each and every CPU than they do now, probably a good deal more.

    Keep in mind that the dollar differences we’re talking about probably are in the range of $10 per CPU, maybe a bit more, so it’s not a killer, but it certainly doesn’t save AMD a lot of money in the long run.

    2) Using someone else’s economics of scale: You can’t make small fabs with today’s or tomorrow’s technology. If you want to fab by yourself, you’re going to lay out a ton of money no matter how few chips you need.

    It’s like buying milk in ten gallon containers when you only drink a gallon a week. Sure, you’ll get a better price, but you’ll waste 90%

    ATI and nVidia pay a lot more per GPU to the third-party fabbers than it costs AMD/Intel to make a CPU, but even after accounting for the differences, it costs them a lot less to buy just what they need than make much more than they want.

    AMD is one of the few chip-making companies where this is not the case. It can keep at least one full-fledged fab working full-time all by itself, and then some. If you have a fab running full blast with the name “AMD” on the door, why would another fab doing the same thing the same way with a different name on the door have much lower costs than AMD? If anything, product from the other place ought to cost AMD more simply because the other guy wants to make a profit, too.

    In the long run, outsourcing part of product makes sense only when the fab you have isn’t enough, but building another one would give you too much. That’s why AMD went to Chartered in the first place, to give them that extra 10-20% for less than building another 100%.

    And AMD has never, ever, ever claimed that chips from Chartered cost them less than chips from Dresden. It’s always been a case of paying more for some extra, flexible capacity if and when needed.

    Downsizing By Another Name?

    There are lots of other issues in play, for a brief but informative overview from someone who actually covers fab plants on a daily basis, go here.

    This isn’t coming from AMD, though; this is coming from the financial community.

    As we’ve shown, fabbing down or even fabbing out isn’t any financial panacea. In the long run, it will cost AMD more, not less money to let others make their CPUs.

    So why would finance business advocate actions that will increase, not decrease costs?

    There’s one very good reason that’s very bad for AMD as we’ve known it. Getting rid of the fabs makes little sense if you think AMD can run them and sell what they make very profitably.

    But if you don’t . . . .

    This is nothing less than a vote of no confidence in the company and the way it does business.

    What these proposals in essence say is, “You’ve never been able to make serious money consistently doing it the old-fashioned way, certainly not enough to justify the billions in investment.

    “Well, the gravy train is over. Your way doesn’t work. Now you’re going to try it our way. Sell your assets (and not so incidentally pay off your debts to us) and let’s see if you can make money without our billions. If you can’t, well, we got ours back, Jack, sounds like a personal problem to me.”

    It’s not clear yet if we’re at the point where the outside beancounters have effectively taken over the company. It’s possible that Morgan Stanley and Friends are really running the company behind the scenes now, but probably not quite yet. At the very least, though, AMD is already under great pressure from these folks and their allies, and will need to do something to mollify them a bit, even if things go as well as can be expected the next year.

    And if things don’t go well and AMD management doesn’t find somebody nicer to take them over, this company certainly will become a puppet on the strings of the big money.

    The least AMD will end up doing is to cut back considerably on what they wanted for fab capacity, and if that means CPUs are going to cost it more, well, they’re going to have to charge more. That means forget about 30% soon, make peace with Intel (which will let them charge more, too). If you still want to seriously compete against Intel, AMD, then you’d better design much better processors that justify higher prices and costs and keep doing it, not sit on your laurels like you did with Hammer.

    This isn’t necessarily a bad thing for AMD in the long run, if not the AMD execs. Barring a miracle, I think Hector and a few others are history by the end of 2008. It will be almost as bad for anybody who likes cheap CPUs.

    AMD will be forced to become a much better (though probably smaller, maybe much smaller) company than they’ve been up to now.

    The problem, of course, is if the company isn’t up to it.


  • Be the first to comment

    Leave a Reply