Having listened to the conference call, here are some points in no particular order:
1. I'll get the subjective aspect out of the way first. It was my feeling that most analysts will not change their opinions, but that some were favorably impressed, while none were negatively impressed, therefore I predict that there will be a few revisions upwards and none down.
2. Inventory is still low = they are selling everything they make and have no plans to change in the near future except that inventories will rise for the next several weeks due to holidays.
3. PC100 yields now at 80%
4. New contract between Crucial and CPQ to supply upgrade memories (about 1/3 of CPQ's needs).
5. Cost benefits of the shrink to .21µ are just now showing up
6. TI fabs are now moving from .27µ to about .25µ due to an upgrade program that TI had started. There will be cost reductions at the TI fabs this quarter, but they will not be to .21µ for 6-9 months. In the future they will probably trail Boise by about 6 months due to the learning curve.
7. Analysts were surprised (as was I) by the fact that even with TI there was only a 10% sequential bit increase. This was due to shutting down the 6" line in Italy and replacing it with 8". Starting in the next quarter there will be about a 20% sequential bit increase.
8. Analysts seemed to believe that MU would have made money w/o TI fabs, but MU specifically refused to answer that question.
9. PC builders have low inventories compared to the same time last year, and are expecting stronger than usual demand for the next two quarters due to Y2K, and therefore MU is not expecting prices to fall much. Also PC makers are planning to put 128MB in most units.
10. Samsung is back to shutting down a week a month for the next 3 months.
11. Analysts were surprised that Op expenses weren't up more, and MU said that they will rise some more, but not a lot.
12. Savings on royalty payments to TI will not begin until Jan. 1, 1999.
13. They are still not shipping much on the spot market but as bits increase they expect to do so again.
14. They are still making some 16MB chips, but those have also been shrunk to .21µ. This enables them to make 166MHz SDRAMs for video.
15. SRAMs are also at .21µ. Now shipping an 8MB chip.
16. They don't expect much RDRAM action until late 1999. Estimates that they are hearing are pretty wide, and range from as low as 5% of memory to as high as 20%. They expect it to be used primarily in servers and to cost about 50% more than other memory. There is a 20-25% increase in die size to make SDRAM, and significantly higher testing cost resulting in a cost that is about 50% higher.
17. Costs would have been a bit higher this quarter had they owned the TI fabs for all three quarters.
18. They will not give any information separately for TI as opposed to MU.
I came away very impressed. I was surprised by how small the number of bits produced by TI was. My feeling is that costs next quarter will be lower than this quarter. 3 factors tend to reduce costs: the yield improvements at .21µ, the TI fab switch to .25µ, and the elimination of royalties to TI. 1 factor increases it - the fact that they will own TI for all three months of this quarter.
I believe that selling prices will fall only slightly this quarter or remain constant. Strong demand combined with supply control by Koreans will keep prices in the $9 range in my opinion. Any decrease in prices will be offset by the fact that with their bit increases they will have some chips available for sale on the spot market, so I believe ASP will remain constant. My estimate for next quarter is a loss of $.10 including TI. Note that there was some speculation that they would treat the losses from TI as non-recurring charges, but they are not doing that. These costs are being blended in.
Anyway, these are my thoughts,
Carl |
