Last week’s shock news from TSMC that it is planning to significantly reduce capital spending in 2008 and indications that this would be potentially become the norm for several years ahead shouldn’t be underestimated. 

I have to point out from the outset that I was surprised by the comments from Rick Tsai during the financial analyst conference call. But when I took the time (provided by a 6-hour flight that took 9 hours with AA last Friday) I started to realise that what I had been saying to people for the last two years had finally been verified by Rick’s comments!

More on that below!

Many people within the industry, including market research analysts and equipment suppliers, had expected CapEx spending to return with some level of rigor in 2008 as it had become blatantly obvious that as utilization rates passed the 90 percent range at TSMC, it would be forced to spend more money in 2008 than it did in each of the last two years.

Equipment suppliers were expecting the major foundries like TSMC to start spending in the third quarter of this year onwards, but as a range of suppliers have now reported, this simply hasn’t happened.

Novellus’ Rick Hill was convinced only a week ago that he expected the major foundries to come back and spend dollars by the second quarter of 2008. Although that is later than most have hoped for, at least the pattern looked good!

As the biggest foundry in all terms this has the potential knock-on effect for the other foundries that include UMC, SMIC and Chartered. TSMC has out-spent the others significantly for over a decade and importantly has the right level of business to ensure that it suffers less than its rivals in any downturn, and can also recover faster. These points shouldn’t be sidestepped as the others have historically mirrored TSMC, albeit at lower levels of utilization, CapEx and profitability for longer than I care to mention!

When the inventory issue raised its ugly head in 2006 I expected TSMC to cut CapEx to better fit the economic climate. It didn’t.

As the chart below shows from official TSMC figures, the foundry spent $2.4 billion on CapEx in 2006. But it started the year with a low level of spending, climbing to a peak in the third quarter. The taps were turned off for two quarters (4Q06-1Q07) as the inventory overhang was highly visible and utilization rates declined.

As can be clearly seen, both 300mm fabs hardly added any capacity, highlighting the lack of demand from customers.

Things started to recover by the end of 2Q07, but don’t be misled by what the chart seems to be showing!

With low utilization rates through 2006 and into 2007, TSMC had added significant capacity at its two 300mm fabs. TSMC had gone through phase 2 & 3 expansion at Fab 12 and had been adding leading-edge capacity at Fab 14.

As the chart shows, TSMC had increased 300mm capacity by approximately 33 percent in 2006, yet utilization rates were declining. In 2007 we saw a similar pattern of investment with the bulk of spending occurring in Q2 & Q3. TSMC is now on track to have expanded 300mm capacity by approximately 43 percent in 2007. However, the utilization rates have only recovered since Q2 onwards.

The realization of all this is that TSMC had been adding significant capacity when demand didn’t justify it. I had been questioning this aspect for some time, as TSMC was out of sync with its rivals.

Compared to the other major foundries, TSMC had ramped its 300mm fabs significantly faster than any of its rivals in the past two years, while many would have expected a closer correlation across the industry. UMC, for example, has been expanding 300mm capacity at a snail’s pace in comparison. Only Chartered and SMIC have seen reasonable ramp rates, both coming from a low starting point. Chartered, for example, has achieved a 1,100 (approx) wafer ramp per quarter in 2007 - hardly a ramp at all!

So is Rick Tsai correct in claiming that productivity improvements as well as the installed base allows the foundry to drastically cut back on CapEx next year?

The chart shows that capacity growth of an extra 10 percent over 2006 has actually been achieved, but yet the total CapEx budgets vary by only $200 million. Considering the larger budget in 2007 has not all gone on 300mm tools it is fair to say that productivity improvements could account for the extra wafer starts.

There is also the strong possibility that TSMC has enough extra capacity already in place that has yet to be ramped at both 300mm fabs. This comes from considering the CapEx spending it has achieved over the last two years.

Then there is the issue of Fab 15, which is expected to be ready for first-phase tool install in 2Q08!

With a CapEx cut will that new facility be mothballed until demand justifies its opening?

Until TSMC provides a real CapEx figure for 2008, this will be difficult to ascertain. TSMC has historically fitted a pilot line but only ramps when necessary. Considering the current cloudy horizon I wouldn’t bet against TSMC following this course of action. The kudos of being able to announce the opening of a new fab regardless of whether it will be ramped straight away may prove too tempting!
It will also be interesting to see what the CapEx plans are from UMC, SMIC & Chartered for 2008. I would be surprised to see significant increases from any of them, which would mean that TSMC would still be spending close to or more than double that of its rivals.

The problem has been that many equipment companies were expecting more of the same from TSMC in 2008, and that’s just simply not going to happen!