>Most likely, AMD has not made the final decision on what to do with FAB 25 and FAB 35. The 2 may be related, but it seems to me that AMD should not convert FAB 25 to flash and be left with only 1 FAB for any period of time before FAB 35 is in production.<
The most rational course would seem to be to have Fab35 ready to begin its ramp as Fab25 begins its conversion to flash. Well, almost.
Possible best case scenario:
Conversion to 0.13-micron is completed at the end of 2002 meaning that, with the exception of percentage yield improvements, AMD's unit production is frozen until some other means of production is brought online. Of course, all increase in percentage yield would probably be absorbed by the increasing die size from the transition from K7 to K8.
Typically, both Q1 and Q2 are down quarters, so AMD could continue to expand market share in in these quarters just by maintaining unit production and sales.
Beginning in Q3 2003, AMD should be ready to expand capacity, either through outsourcing to foundry services or bringing Fab35 online. The disadvantage of bringing Fab35 online is that it is most efficient to ramp Fab35 with cutting edge 157nm lithography, and AMD will not be moving to 157nm lithography until 2004.
One potential option is to strike a deal with IBM, whereby IBM would ramp Fishkill with ASML equipment. AMD could then begin using IBM as a foundry (or leasing fabrication facilities from IBM if AMD's agreed upon cap on foundry services is still a barrier) in H2 2002, possibly utilizing all the capacity brought online at Fishkill by the end of 2002. Then in 2003, as UMC ramps production, AMD could begin temporarily utilizing UMC capacity for the production of processors, perhaps substituting three 300mm wafers of capacity at UMC for every five 200mm wafers of capacity at Fab25 shifted from processor production to flash production.
The next phase would begin with the ramp of production at Fab35. As AMD begins to install new 157nm lithography equipment at Fab35 for the critical mask steps, they team this with used 193nm lithography equipment from Fishkill. And, as AMD pulls this 193nm lithography equipment from Fishkill to help ramp Fab35, IBM replaces it with 157nm lithography equipment that is teamed with IBMs own 193nm lithography equipment at Fishkill.
Once the transition to 0.10-microns is completed, both AMD and IBM would have 50% 157nm lithography equipment for critical mask steps and 50% 193nm lithography equipment for less critical mask steps.
Of course, if AMD builds two hybrid fabs, designed for one third of the steppers being used for the production of flash, they could be at the above stage with just 1/3 of the capacity at the second 300mm wafer fab being ramped. Then, with the ramp of the final processor production capacity at the second 300mm wafer fab, AMD could phase out the foundry services of UMC. And as the next generation of lithography equipment is installed in these two 300mm wafer fabs, the 193nm lithography equipment displaced by this new generation of equipment could be shifted to flash production, completing the ramp at these two fabs. |
