|
Fab capacity is a myth! |
|
|
|
Jan 31, 2007 at 11:41 AM |
One of the first lessons I had in the semiconductor industry concerned ‘fab capacity.' This simple term is meant to describe the number of wafers a fab can produce when fully equipped and operating at full throttle. The cleanroom is bursting with tools, operators and in-process wafers. No matter what you do, the output becomes a constant, hence ‘capacity.'
When chip manufacturers announce the building of new fabs, sometimes we are lucky enough to get details of that new fab's ‘capacity.' That figure is often carefully calculated as it provides a small level of competitive insight. A growing trend is to only provide first phase capacity levels and never actually comment again on what the ‘real' capacity of a given fab is!
Others simply refuse to give any clue as to the capacity, cleanroom space, etc., which makes life that little bit more interesting.
To be honest, fab capacity figures, when given by chip manufacturers, are all too often fictitious. Plucked out of the air, they are as reliable as the London underground system. Within our 300mm activity report we have pages of tables, which list the estimated capacity figures of a given fab. These are revised constantly, while expected tool install dates (a much more fluid affair) rarely change that much!
A classic case of ‘conservative' capacity levels was given by Toshiba & SanDisk concerning Fab3. This dedicated NAND flash 300mm fab is one of the largest facilities of its kind anywhere in the world. The ‘quoted' capacity was ‘set' at 110,000wspm. Looking at the scale of this fab from photographs I had always believed that to be a conservative figure. However, this is hardly the correct way to estimate things, and so we listed the capacity in our reports at 110,000wspm.
SanDisk, for its part in this story, has been good enough to provide ramp rate levels at Fab 3 in recent quarterly conference calls, which help us to project the rate of ramp and the current capacity levels. Three months ago, SanDisk guided on capacity levels, which strongly indicated to us that Fab 3 would reach just over 60,000wspm by the end of 2006. Full capacity would therefore be reached by the end of the third quarter of 2007. Simple stuff really, when you get the hang of it!
Apparently not!
Last night SanDisk noted that Fab 3 capacity had reached 90,000wspm at the end of 2006! That is nearly 30,000wspm more than we guided from one quarter to another. This is actually rather staggering and sets a new benchmark for ramping a 300mm fab, far above previous benchmark levels.
No sooner had I picked myself up from the floor, when it was stated that Fab 3 would reach a capacity level of 135,000wspm by the end of 2007!
Is that the ‘true' capacity of this fab or simply the ramp level expected by the end of this year?
Who knows?
Going back to aerial photos of Fab3 and the under-construction Fab 4 sited next door, it begs the question what capacity will Fab 4 have?
We have it under good authority that Fab 4 would have a capacity of 150,000wspm, making it the biggest fab ever built. But looking at the size of Fab 4 compared to Fab 3 and now knowing Fab 3 can do 135,000wspm, Fab 4 could potentially pump out 200,000wspm!
Unfortunately we just don't know and will just have to wait and see!
In that first lesson I had, which was actually given to me by Bill McClean, now over at IC Insights, he noted that fab capacity was a ‘Pandora's box.' Never a truer word has been spoken!
|
|
|
Blogs 
