Re: Now that's what I call a profitable business
Amen. That's why I posted that getting too focused on raw silicon costs might not be the best idea.
I also think your package and test costs are too high (it's probably more like $10).
One additional point on the comparison. When Intel is at 300mm wafers, AMD will be shipping 80mm2 Thoroughbred/Appaloosa parts (same size as Tualatin).
Intel claimed they'd initially see only a 0% to 5% cost reduction from the 200 to 300mm transition, but I think they were sandbagging and will do somewhat better than that right off the bat.
If you compare your estimated cost for Tualatin (Same size as Athlon) with your estimated cost for Northwood, you get the 40% cost advantage that Sanders was saying AMD expected for next year (even with Intel using 300mm and AMD using 200mm wafers).
But, the main thing is, a cost difference of $3+ dollars is pretty much irrelevant when most of the parts sell for $50 to $250.
The other interesting thing is trying to allocate "overhead." Just for a starting point, let's say Intel ships 120 million processors next year and AMD ships 40 million. Intel has been running annual Capex of $7 Billion while AMD has been running annual Capex of $1.2 Billion. Intel has R&D costs of $3.6 Billion, while AMD has R&D costs of $0.7 Billion. Intel has administrative costs of $4 Billion while AMD has administrative costs of $0.6 Billion. Without trying to guess what Intel is including as R&D, AMD seems to be using its R&D money more effectively.
Bottom line (in millions except final number), per chip direct overhead would be Intel ($7,000 + $3,600 + 4,000)/ 120 = $122 and for AMD ($1,200 + 700 + 600)/40 = $62.50.
This is why I keep harping on how critical it is to fully understand and account for Capex, R&D, and admin costs (and how unimportant silicon costs are). It also explains how AMD can lose money while selling for $80 CPUs with a cost of $20, and Intel basically broke even selling $25 parts for $150.
Both companies make things other than CPUs, but (unless and until the flash crash ends) neither company is making enough money on non-CPU products to cover overhead costs - so I'm just ignoring those as either costs or revenue.
This also shows why AMD is so anxious to keep its units up, and why Intel is so determined to take back the market share it lost. If Intel can drive AMD down to 12.5% market share, AMD's per chip overhead costs will double to match Intel's. Right now, AMD has a pretty significant cost advantage - which has absolutely nothing to do with the costs of making, testing, and packaging the chips themselves. The cost advantage is what allows AMD to survive while it's ASPs are so much lower than Intel's. |
