Re: the $10 chip cost( optimistic estimate by Elmer), already includes FAB depreciation
They invest in their FABs at a rate that loads them with $7,500 million in new CAPEX costs each year and produce parts at a rate of 100 million per year (a nice round number) - that's a $75 per chip capex cost, but I certainly wasn't cutting Intel any slack on that estimate and we're all hoping for more CPU sales next year, anyway.
Instead, lets say that they will sell 150 million chips next year, and that 25% of the FAB investment is for flash and upgrading old facilities that produce chipsets and comm.
You wind up with half the depreciation costs ($37.50). The true number is probably somewhere in between those two (and will be heavily dependant on the number CPUs sold next year - market share, market share, market share). Perhaps the average of the two, $56.25.
But there is no way depreciation is included in Elmer's $10 cost estimate.
In AMD's case, the costs are $800 million in capex and (round number) 25 million chips sold, making the high estimate $32 per chip. AMD talked about making 50 million CPUs at Dresden alone next year, so let's use that number for the combined estimate from Dresden and Austin. I think it's fair to say more of Intel's FAB costs are non-CPU, so let's leave AMD's CPU CAPEX charges at 90% of the total, $720 million.
AMD's low boundry CAPEX per chip would be $14.40. Once again, the true number is somewhere in between. The average of the two is $23.20. But it is almost certain that AMD has a $20 to $40 CAPEX advantage, per chip, over Intel - and that this advantage will continue to hold next year.
Which should put the possible $4 to $6 savings from 300mm in perspective. |
