We’ve been hearing from the CPU companies that their ASPs have been going up a little. This has led to claims that CPU “price wars” are over, and I know that’s left at least me scratching my head a bit asking, “How?”

There’s no doubt a few explanations for this, but here’s one that’s pretty simple to understand.

As the proportion of quadcores slowly increase, absent sizable price cuts elsewhere, ASPs will go up simply because quadcores are priced higher than dualcores.

However, what doesn’t get captured in the overall ASP is the fact that a quadcore costs a good deal more to make than a dualcore and chews up at least twice the resources to make as a dually.

It’s like buying beer by the case rather than the six-pack. Yes, the beer maker will get more money from a case of twenty-four than a pack of six, but it also costs him four times as much to make, and he’ll probably make less profit selling a case for $15 rather than four six-packs at $6 each.

To be truly more profitable in the CPU industry, you need to get more than double the price for the quad than you would for the dual, since you can only make half as many quads with X amount of production capacity as you can with a dual.

Let me give you an extreme example of this:

You have a fab. It can make twenty million dualcores a quarter. Let’s say it costs $40 in actual manufacturing costs to make a CPU, and the overall overhead (financing the fab, paying for R&D and sales and executive bonuses) is a half-billion dollars a quarter.

Let’s assume the ASP for a dual is $70, and you can sell all you can make. The math is simple: 20,000,000 CPUs X $70 revenue per CPU equals = $1.4 billion dollars gross revenues, less $800 million ($40 X 20 million) cost of goods sold, less $500 million overhead, equals $100 million profit.

Let’s assume the fab makes quadcores instead. The fab can make ten million quads a quarter. Each quad will cost $80 to make, and you still have the same half billion overhead.

Just plug the numbers in, and it’s easy to see that you’ll need to get more than $140 per quad to be better off making quads than dually.

Just to explicitly illustrate, if the ASP for the quads were $120, the numbers would be 10,000,000 CPUs X $120 per quad, which would equal $1.2 billion revenues, less $800 million ($80 X 10 million), less $500 million overhead equals a $100 million loss.

As you can see, even though the ASP went up a lot, over 70%, the company gets less money for its quads than it does for its duallies, and loses money as a result.

This is obviously an extreme, oversimplified example, but that’s done to clearly illustrate a point: an increase in ASPs isn’t necessarily going to be a pure plus in the year ahead; it could actually be a negative.

How would you know if that is actually the case? You can tell if that’s the case by what happens to the gross margin percentage.

In the first example, with the 20 million duallies, the gross margin percentage is the gross margin of $600 million ($1.4 billion minus $800 million) divided by gross revenues of $1.4 billion, or 43%.

In the second example, with the 10 million quads, the gross margin percentage is the gross margin of $400 million ($1.2 billion minus $800 million), divided by gross revenues of $1.2 billion, or 33%.

The point to the story? A company like AMD isn’t going to be better off getting somewhat more money for a tri- or quadcore than for a dual, it needs to get a lot more money for them.


Be the first to comment

Leave a Reply