"Will 300mm contribute to over capacity in 2002?"
This is article is a few months old, but worth the read IMO. Thanks to Didi on the SI Notes thread for first posting the article.
icknowledge.com
Published 11/11/201, updated 11/12/2001 and 11/13/2001.
The last few weeks have seen a steady stream of market research firms announcing new lowered expectations for 2002, with most now forecasting less than 10% sequential growth - see, IC Insights, Dataquest, SIA, and some even forecasting a decline - see, Future Horizons.
According to the SIA - SICAS report, the current worldwide capacity for 200mm wafers is 1,324K wafers starts per week (wafs/week), of which 850K wafs/week are actually being produced (64% utilization).
IC Knowledge estimates that linewidth shrinks from 180nm to 130nm could add the equivalent of 85K wafs/week to the existing capacity by mid 2002, and another 85K wafs/week by mid 2003.
IC Knowledge estimates that 300mm wafer production capacity will end 2001 at 107K wafs/mnth, end 2002 at 199K wafs/mnth and end 2003 at 406K wafs/mnth. This equates to an additional average capacity on a 200mm equivalent wafer basis of 82K wafs/week for 2002 and 163K wafs/week for 2003. It should be noted that since a significant portion of 300mm capacity will be at 130nm linewidth, the actual 300mm effect may be even greater.
The totals are 1,324K wafs/week (current capacity) + 85K wafs/week (linewidth shrink) + 82K wafs/week (300mm) = 1,491K wafs/week for 2002, and 1,324K wafs/week (current capacity) +
170K wafs/week (linewidth shrink) + 163K wafs/week (300mm) = 1,657K wafs/week for 2003. This is in contrast to current demand of 850K wafs/week.
This analysis suggests that to maintain ASPs the semiconductor industry needs to grow approximately 67% for 2002 (to reach 90% utilization). Assuming the revenue forecasts are in the right ball park, either Fab utilization or ASPs must drop. Since semiconductor companies are loath to allow low utilization on new Fabs, either older Fab utilization will drop, or some old Fabs may be taken off-line and still there will be significant ASP pressure.
Some ASP decline is of-course business as usual for the semiconductor industry. DRAM and Microprocessor ASPs have on average dropped 35% per year for the last 3 decades (see Cost per function trends), and in fact we expect to see continued drops in processor prices next year as AMD attempts to wrest market share from Intel. The DRAM market place is a different story. DRAM prices are currently below what IC Knowledge calculates the manufacturing costs to be. All of the DRAM manufacturers are losing money with Hynix in danger of bankruptcy. We expect to see consolidation in the DRAM business in the year ahead.
If we assume that 2002 will show a 5% revenue gain and a 20% ASP reduction, then wafer volumes required should ramp up to approximately 1,116Kwafs/week. In 2003 if we assume revenue growth of 25% with 20% ASP decline, then the required wafer capacity could reach 1,744K wafs/week resulting in under capacity. This analysis suggests that unit volumes at the leading edge must almost double next year to maintain ASPs, which also suggests a near doubling of revenue. Given the unlikely hood of revenue doubling and the desire of semiconductor companies to fill leading edge 200mm Fabs and new 300mm Fabs, there is a risk of huge drops in ASPs for 2002 Alternately we may see further push-puts and delays in 300mm Fabs. The largest number of the delays announced to-date have effected Fabs due to come on-line in 2004 (see 300mm push-outs), to avert over capacity we believe that additional 2002 and 2003 300mm Fab delays are warranted .
This simple analysis points out that 2002 is currently a weak spot with over capacity threatening to derail any fledging recovery. 2003 could actually see a capacity shortage based on current investment levels.
Related articles
300mm push-outs - our latest 300mm Fab forecast based on a complete review of our data. The graph contrasts the current forecast to prior forecasts. (10/29/2001)
300mm transition - cumulative number of new 300mm Fabs by year from our October forecast. (10/1/2001)
300mm transition - cumulative number of new 300mm Fabs by year from our August forecast. (8/8/2001)
300mm cost comparison - comparison of 130nm technology - 300mm DRAM and microprocessor wafer costs. (8/14/2001)
Back to Top
Copyright © 2001 IC Knowledge - all rights reserved.
AdvocateDevil |
