What’s wrong with Semiconductor Equipment Manufacturing? - Part I

We have a guest contribution today from Israel Beinglass, the CTO of MonolithIC 3D Inc. Israel was at Applied Materials for almost two decades, and served as Chief Technology Officer and Chief Marketing Officer for many groups there.

---

In this blog we will try to understand interesting phenomena of the relationship between the semiconductor equipment business (CapEx) and the semiconductor manufacturing business. For many years now I was watching with awe the fact that some of the semiconductor equipment manufacturing companies don’t grow their revenue over the period of the last decade while at the same time the overall semiconductor business has grown.
.

Let’s look at the trend of the revenue of the 4 major semiconductor CapEx companies since the year 2000 through 2010. The companies that we will be looking at are: Applied Materials, ASML, Lam Research and Novellus. Together these 4 companies represent most of the front end CapEX industry. During the period of 200-2010 we have gone through the post DotCom fiasco in the year 2001-2003 and the big down turn of the world economy and the banking debacle of 2008-2009. Also during this period the semiconductor manufacturing industry has gone through several major changes

1. Completion of wafer size change from 200mm to 300mm
2. We lived through 4 technology nodes: 90nm, 65nm, 45nm, and 32nm
3. Channel stress both using selective epitaxial deposition and stressor films deposition became the standard mobility enhancement knob.
4. High k Metal gate was introduced and though is not a the standard everywhere it is on the way to become a standard in the coming years

With all these amazing changes it is quite surprising to follow the equipment business and try to understand yet another amazing phenomenon - there is no real growth in several of these equipment companies.

Figure 1 is the Applied Materials revenue trend (excluding Displays and Solar) since 2000. One can obviously see a big "down" in the “bad years” but no growth at all for the whole decade.

.Figure 2 is the same period's trends for Lam research and Novellus. Similar trend for these two companies, the trend is overwhelming similar to Applied Materials in the case of Novellus and less in the case of Lam where the revenue went up during 2005, 2006 and 2010. Explaining that is probably the market share increase in their etch business, adding the wet clean revenue from the purchase of SEZ and double patterning becoming more prominent.

Figure 3 is the ASML trend for the years 2000-2010. In this case, we see a definite different trend. While in the other cases there was no growth (Lam little different) we see ASML sales are increasing over the period of 10 years. It is actually very clear and very obvious, it used to be that a stepper cost was $5M and later it was $10M. However the latest 193nm immersion is in the $30M range. Needless to say that if EUV becomes a reality in the next few years, a price tag of $80-$100M will be the new standard….Obviously ASML is (almost as a single supplier) enjoying very good market positioning.

At this time the industry is utilizing double patterning as a fall back position due to the lack of a viable solution for EUV, which in turn forced manufacturers to buy more steppers and etchers.

Let’s review again the reason why within a range of 10 years revenue of some leading companies didn’t grow at all, though the total number of units has gone up as well as the total semiconductor revenue. This is in spite of Applied Materials leading in PVD, Low k deposition, RTP, Epi and CMP. At the same time Novellus has a leading position in several CVD depositions and ECP.

The reasons for these are:

1. Less Fabs are built in the world due to the success of the foundry business and consolidation in the memory business.
2. Reuse of same equipment in new Fabs, as the equipment is being transferred instead of retired.
3. Conversion of 200mm to 300mm, increasing by 2.5x the number of chips on each wafer, hence less equipment is needed.
4. The throughput of most of the new equipment has gone up dramatically within the last decade by the equipment companies due to pressure from their customers.

There are more old Fabs running non advanced node devices (see my posting Moore's Law, the bifurcation of the semiconductor industry and 3-D integration) so there is no need for new equipment. Those Fabs can keep running analog, power electronics, MEMs etc. with fully depreciated equipment or using used/refurbished tools.

Figure 4 shows the quarterly unit shipment during 2000-2011 for all semiconductors world wide, it definitely shows upward trend of about 3X during this period

IC Insights updates semi forecast, discusses China's stimulus success

In summary, in new Fabs, the Litho portion of the total expenses is getting larger and larger and the companies who are selling to this market will increase their % of the whole new Fab expenses.

In my next blog I will continue to develop these ideas and I will discuss several options of how the industry will continue on the Moore path: using double patterning, direct write, EUV, and 3D devices. Also we will revisit the effect these new technologies will have on the future revenue of equipment companies.