|
UMC released second quarterly financial results yesterday (July 27,
2005), which didn't do much for equipment companies expectations that
the second half of the year would see a return to capital equipment
spending on the back of improving foundry utilisation. Granted, UMC
seemed relatively happy that fab utilisation had reached 65 percent
compared to recent guidance of 60 percent but in terms of historical
correctness utilisation figures need to pass the 85 percent to trigger
fab expansions.
Though UMC painted a picture of improving conditions for the
second half of the year the stats did the real talking. Capital expenditure has
now been reduced from $ 1.5 billion US Dollars to no more than $ 1.0 billion US
dollars for the year. But considering that UMC has only spent $ 332 million US
dollars in the first half of the year, real demand must improve significantly
for them to spend the rest of the CAPEX budget in 2005.
The foundry did highlight that the majority of CAPEX would
go on its two 300mm fabs (Fab 12, Taiwan & Fab 12i, Singapore). This is for
slow but continued expansion of 90nm devices. However, in the wafer starts
figures UMC produced, both 300mm fabs are not expected to ramp any further in
2005! Estimated figures listed by the foundry show that for the next two
quarters Fab 12's wafer starts per month will stay at 22,667 while Fab 12i will
be static at 13,926.
This is a drastic change in the forecasted ramp projections
for both 300mm fabs. Previous to this particular quarter, UMC had projected
that capacity at Fab12 would reach over 25,000wspm by year-end and Fab12i would
reach over 20,000wspm.
Demand going forward for leading edge processes would seem
to be limited. One possible explanation is the continued inventory glut that
UMC's biggest customer is still having to deal with. Xilinx reported recently
that it was holding 4 months (132 days) of inventory, a problem it has been
carrying for nearly a year. Its volume products are mostly fabricated using
130nm processes on 300mm wafers but UMC's figures showed clearly that 130nm
production had dropped from 20 percent of revenues to only 14 percent of
revenues.
Now for the punch line! UMC is in trouble of repeating the
same mistakes as it made in 2003 when it held back major tool orders to the
extent that it missed a large chunk of the last upturn (2H03 to 2H04). This was
due to not ordering lithography tools until late December 2003, which everyone
knows (except perhaps UMC) take 10 months to build and ship. 2006 now looks
like being a very good year for the industry, but UMC is in danger once again
of missing the wave and of course getting caught with its trousers down.
|
Blogs 
