The shift to Asia: Cause for concern in the US?
02/01/2004
The shift of IC manufacturing to Asia — always an issue — has cropped up recently as a growing concern. While significant, it is not as overwhelming as some might think, as the data in the table show. Because equipment is the font of all chip growth, the data makes it clear where market share trends will flow in the future. Most think of this shift to Asia as being comprised of Korea, Taiwan, and Japan, but that is really not the case. The biggest move is to "other Asia," which is mostly China.
The difference that I have seen recently is the degree to which equipment makers are noting the shift. They report a huge shift to Asia in buying activity that is unlike that seen in any other cycle. It's not because Asia is buying so much; it's because so many IDMs are not spending, apparently giving up manufacturing. In the past, Asia added to the buying increase in a cycle. In this cycle, Asia accounted for the bulk of the big buys and the long-term planning. Outside of a handful of the largest IDMs, most are simply not spending, nor do they have plans to spend. US share of spending is down by 9 points and Europe's is down by almost 3. Japan is the only major chipmaking area that is seeing a revival in spending and thus projecting confidence in the future.
 |
At the same time, the large foundries are also not spending much and appear to be ceding their leadership to China, as can be seen in the steep drop in Taiwan's equipment purchases. Taiwan has dropped by almost 7 points since its peak in 2000 and is actually below its spending share in 1998, while other Asia is well above it. It is more confirmation of what we have been saying all along: that this cycle's winners and losers will be very different [1].
Historical perspective on equipment orders
China is such a huge factor in the growth, but has an unknown track record when it comes to spending consistency. This leaves open the question of how long the current upturn will last. Historically, it has taken cumulative growth in spending of over 100% to generate gluts. That would mean that the equipment market would have to grow by over 100% in 2004 to generate a glut in 2005. The probability of that happening is extremely low. Even though the industry has broken growth and decline records since 1998, I do not think it is possible. I doubt that the equipment industry has the ability to double in size in 2004, nor the will to repeat the last cycle where it doubled in two years and then halved in two. Though it does not mean that the chip business won't decline in 2005, it would, however, have to be demand-related.
It is difficult to believe that there is double-ordering when the book-to-bill ratio continues to hover around one, but I've just about given up on orders as a useful metric. The reason is that the industry has become very conservative in booking orders over the last two years. What is a booking today was a billing 10–20 years ago. In some of the most conservative examples I've seen, orders were not booked until after the manufacturing slot was open, the materials had been ordered, and the tool build had already started. The only other way to be more conservative would be to wait until the tool has shipped and then accepted. I have not heard of anyone going to these extremes, but I would not be surprised, considering that companies never want to show cancellations.
In this last cycle, it was not uncommon to hear of chipmakers canceling orders for tools that had been delivered, installed, and were waiting for acceptance. In the worst cases, they refused to pay for the tool and refused to return it — the latter would not have been useful at any rate, since systems are so customized. They were told to deal with it or lose a customer forever; their customers were doing the same to them. It will be interesting to see what happens to these chipmakers' ability to get equipment in the upturn. The main point is that the quality of bookings data is very poor because the basis for its definition has changed in time. In 2000, an order was generally booked when the purchase order was received, and 20 years ago it was often only a verbal commitment and a handshake.
That said, there is a wellspring of unrecorded activity as our weather index indicates [2]. But customers are still very reluctant to issue purchase orders. If I were to grade the book-to-bill ratio today on the old metrics, I would put it at 1.1 based on booked-at-PO and 1.4–1.5 based on booked-at-verbal-commitment. So this is really not a double-order situation. My suspicion is that executive fears about double-booking center on the fact that there is no longer a lot of floor space to put the tools that customers are talking about. Chipmakers are trying to tie up build slots to get in the queue before they actually have approval for the fab itself. So if business softens, it could all fall apart. But I wouldn't worry about double-booking unless you start to see the equipment makers add capacity — we still see some quietly cutting capability. Double-booking, then, just gets pushed out to future slots, making demand the issue — not supply (the first time this has happened in years), which is truly a positive development. As long as the economy grows, we should be in pretty good shape.
Acknowledgment
Chip Insider is a registered trademark of VLSI Research Inc.
References
1. Additional data may be accessed in the appraisal section of the VLSI Research Web site at https://www.vlsiresearch.com.
2. Refer to the "weather report" at https://www.vlsiresearch.com.
Dan Hutcheson is president and CEO at VLSI Research Inc., Santa Clara, CA; ph 408/453-8844, e-mail [email protected].
The material in this article was adapted from VLSI's Chip Insider with permission.