Steven, Every downturn in the history of the semiconductor industry has had a sizable supply component. However, the 2001 downturn was the first ever with a major demand component. The second largest component in the 2001 downturn was excess inventory (semi company and supply channel), which in some cases was substantially longer than a year. I've written extensively about these events, what has changed since 2001 in the supply channel and in the industry as a whole (plenty has changed in inventory management).
Many industry executives (I was in the semi industry for over 20 years) who have read my report on Semiconductor Cycles have stated its the best paper ever written on the topic. I think you would find it interesting.
Please note, I'm not suggesting that a shred of your data is bad, but the analogy between now and 2001 is faulty - this is more like 1996, which is a decline that most analysts have misinterpreted as well.
BTW, another interesting point about TSMC is that the company maintained operating profitability throughout the 2001 downturn hitting a low of 1.7% pro forma operating profit in calendar Q2 of that year while the other major contract fabs reported staggering losses.
Perhaps some context is necessary... While demand was certainly weak in 2001 as represented by a ~20% decline in global chip units, I still believe the 2001 foundry downturn was particularly exacerbated by excess supply.
Some datapoints to ponder:
As a reminder, global semiconductor capital spending increased nearly 80% y/y in 2000 compared with ~5% last year (2007). Specific to foundries, the combined capex for TSMC and UMC increased ~60% in 1999 AND another ~130% in 2000. Clearly there was far too much supply being built, which is why TSMC saw its utilization rates fall to ~40% in mid 2001. In looking at today, TSMC and UMC capex was FLATTISH in 2007 (and now likely down 30%+ in 2008) - which is why we see TMSC utilization rates troughing higher this time (in the mid 60s).
I agree that demand weakened in 2001, but we also shouldn't forget that the 2001 chip unit decline was after seeing a 25% increase in 2000! Far ahead of historical trend line. Clearly, we built far too much. Relative to end device demand, both PC's and handset units declined 4% in 2001. But the compare is also important here too. PC's were coming off of 24% y/y growth in 1999 and 16% in 2000. And PC's quickly snapped back to 15% growth in the two years thereafter. Mobile phones returned to 17% y/y growth in 2002 and 23% in 2003.
My point was simply that growth rates have moderated. And, if there is a silver lining, it is that the foundry supply chain reacts more quickly these days (rather than the past appearance of accelerating capacity growth right as we go off a cliff).
That's my two cents,
"Pelayo compared the current climate to the downturn of 2001. "The 2001 downturn was more excess supply related rather than today's potentially more troublesome problem of weak demand," he wrote."
That statement is factually incorrect. In 2001 total IC unit volumes declined by 20.8% (total semiconductor unit volume was down 21.3%) - that is the only time in the history of the IC we've seen a double digit decline in unit volume. As such, it was the most remarkable collapse in IC demand in history - partly because of excess channel inventory, but mostly because demand simply fell off a cliff.
For example, the downturn in 1996, which was mostly supply driven (excess memory capacity pushed prices down by more than 50%), still saw unit volumes increase by 1.7%.
When it comes to measuring demand, unit volume is the best indicator - not revenue. Revenue is effected by changes in average selling prices that can be driven by many variables.
To read about what has really driven semiconductor cycles through history you can visit nextinning.com and read my report on the topic.
Editor in Chief
Next Inning Technology Research
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...