Thanks for your comments. You bring up very important points. So important, in fact, that I wrote a new post to discuss them in greater detail.

For example, 2001, 2002, 2003, and 2004 were all smaller years than 2000. Demand fluctuated. Hell, from 2000 to 2001 it cratered.

Granted, over a 10 year perspective, we’ve seen consistent growth decade to decade. But we have seen occasional decreases in demand and we’ve always seen changes in the demand mix.

For example, INTC was the #10 semi company in 1987, but #1 by 1992. AMD was #15 in 2005, but #7 in 2006. AMD could have been higher in 2006; however, they were supply constrained because they didn’t have enough capacity.

The semiconductor market isn’t a seamless whole. For example, if INTC moves to 300 mm platters and 45 nm processes, they’ve just increased their capacity/reduced their cost by a factor of 4. AMD won’t be able to effectively compete with 200 mm platters and 65 nm processes if their cost is 4 times higher.

Not to mention that INTC’s parts will likely go faster at 45 nm *and* use less power which means that they’ll be able to charge premium prices. Or, as they’re doing now (before they even finish the 45 nm migration), charge normal prices and deny AMD the ability to build and equip a new fab without seeking outside financial assistance.

Standing pat doesn’t win success in the semiconductor market. Companies that hop off the treadmill pretty much are destined to long-term obsolescence.

By the by, at least as recently as 2004, many companies considered 85% to 87% utilization as full capacity, 90%+ as supply constrained, and 95%+ as critical shortages. So with utilization only now falling into the historic full capacity range and with most people calling for a bottom this quarter’sh, it only makes sense to bring capacity online now to ensure that it’ll be ready for next year.

I agree that the whole thing is anti-intuitive and in some ways capex spending is a house of cards.

But if you don’t spend and your neighbor does and the demand is there, you’re done. And unless you can spend enough to catch the next cycle, you’re moving several steps down the Top 25 semiconductor list. And in this market, volume is power. So better to overspend and drown you and your neighbors in an oversupply problem than to underspend and let your neighbors thrive in an shortage scenario while you do without.

reinharden

If the semi makers consistently invested in 10% growth industry-wide there would be no more cycles and I would no longer have to raise the alarm when they start acting stupid. They would always have on order the equipment they would need just when they start to need it.

But I don’t think I have to worry. They’ll continue to be stupid and I’ll continue to be able to make money trading around their supply issues.

If you don’t overbuild during the period of oversupply, then you don’t benefit during the period of overdemand.

Is it totally rational? Perhaps not. But unless every semi company got rational at the same time, you’d set your own up for bankruptcy by failing to play nice with the other kids.

Anyway, you’ve also got to keep in mind that you’ve got to order semiconductor equipment at least 6 months to 18 months ahead of when you’re going to get it delivered and add another 6 months to 12 months to get it installed and the problems with your processes resolved. Which means that you’re ordering equipment now in the hopes that you’ll need it two or three years from now.

Anyway, semiconductor companies can’t add new capacity instantly, so they spend their way through most downturns in preparations for the upswings. ’cause it’s always worked before.

Maybe it won’t one of these times. But I think this beating the drum of yours is pretty alarmist. And for someone who’s apparently been an analyst since 1996 seemingly without the benefit of hindsight.

reinharden