Online information source for semiconductor professionals

There is a murky grey mist covering most of the UK at the moment and visibility is rather poor. The revised Gartner CapEx spending forecast may be accurate from the point of view of what has happened this year, but there remains a considerable lack of clear visibility over the revised projections for 2008.

Initially, our own research was at odds with Gartner as we entered 2007, with Gartner projecting a slight CapEx decline in 2007 compared to 2006.

However, we at Fabtech had plotted that approximately 17 new 300mm fabs could come on stream this year while the significant number that started production in 2006 would be further ramped to somewhere near full capacity in 2007.

This all meant that it was very unlikely that CapEx spending would be down compared to 2006. By June, Gartner had revised projections and noted that it expected a 5 percent increase in spending, pretty much close to but still below what we had expected for 2007.

I think we may have been the first to highlight that the foundry ramp at 65nm was not happening at the expected pace, which was limiting the need for the major foundries to add new capacity at the leading-edge.

Though fab utilization rates were rising, the major foundries had yet to increase capacity, but remained adamant that they would meet spending levels previously set for the year.

Gartner’s latest report takes note of the foundries’ dilemma at 65nm and the impact on equipment sales. This is due to demand being required at 90nm and above rather than at the leading-edge node.

Overall, I think Gartner’s view of how 2007 will pan out in this latest revision is inline with our own analysis. A conversation with Gartner’s analyst Dean Freeman yesterday showed that we both project 13 300mm fabs coming on stream in 2007.

The point here is that these projections rarely get into alignment, especially since both Gartner and Fabtech have both recently revised the number of first phase tool installs.

In the most recent edition of Semiconductor Fabtech (#35), I highlight in the ‘300mm activity report’ that fewer than the expected number of 300mm fabs are coming on-stream this year.

There is a difference between Gartner’s own view of the knock-on effect in 2008 and what may actually happen in 2008. Gartner sees less CapEx being spent, but we also see the possibility of more fabs starting initial production!

There is also the issue of a number of paper fabs that have yet to be confirmed this year, which could still break ground in 2007 and start tool install in 2008.

There is also the question of whether Samsung’s and Hynix’s 200mm memory fabs will actually be converted to 300mm production in 2008. Though both Korean memory manufacturers have openly stated that conversions were being evaluated, little has been said with regards to which fabs - and when, if so - these facilities will be converted?

Freeman believes that conversions are inevitable but that this may be delayed to enable one more shrink for flash devices in 2008. Aggressive process tweaking via the Applied Materials or Lam Research solutions announced this year may come into play, but the jury is out on that right now!

The bottom line, as I see it, is that 2008 can go three ways (see below), with a negative CapEx year on one side of the logic fence as projected by Gartner certainly feasible!

However, it seems that with the current lack of new fab tool installs as highlighted in the latest 300mm activity report, the picture could actually be far worse than the situation depicted via Gartner’s small decline.   

What we don’t have in alignment is the number of 300mm fabs we expect to see entering first phase tool install for 2008. My number is half the number Gartner is projecting!

My point here is that they see a slight negative level of spending in 2008 due to various factors, taking into account the number of new fabs the company expects to come on stream in 2008.

Our own data suggests that there is a significant upside to the number of new 300mm fabs that could come online in 2008. The problem is that the real view is cloudy and much uncertainty exists as to how 2008 will pan out.

The third and final scenario is that the number of new fabs coming on stream will actually be above our own limited number but also above Gartner’s number, pushing CapEx levels above the 20 percent level!

This scenario rests on fabs’ reaching very high utilization levels in the first half of 2008, forcing the majority of chip manufactures to spend more on adding capacity than they had planned.

Ultimately, it is the market that dictates all three scenarios, but chip manufacturers also help or hinder such dynamics. The wild card has yet to be played for 2008. However, the visibility level is so poor, and we at Fabtech will be tracking these developments as closely as possible, as it is clear that things could swing one way or the other!