Semiconductor International is highlighting the release of a couple of forecasts on the foundry industry.   Those forecasts are  quite optimistic for that segment of the semiconductor market.    The two reports, one from iSuppli and another from IC Insights, are quite consistent in the actual revenues through 2008 and are fairly consistent in their view of the years ahead.    Both show foundry revenue of about $20B in 2008, about a 15% decline in 2009 to around $17B, followed by three solid years of 20%/annum growth.    Doing the math, at that rate, there will be a pure-play foundry industry of about $30-35B in 2012 or nearly double the business of the segment in 2009 and 75% above 2008 levels.    Wow!  Corks must be a-poppin’ in Taiwan and around Asia!!

But wait a minute, something’s up here … !?   

Let’s look at a few structural facts that underlay this forecast.    First, let’s look at the pure-play foundry industry itself.      As noted in the SI article, this industry is characterized by increasing consolidation.    The market shares in the industry as reported for Q2 of 2009 are shown below.   Note that TSMC is more than half the market and the top 5 foundries which include UMC, Chartered, SMIC and CRM (China Resource) comprise just over 90% of the pure-play foundry segment (source: GSA).     The SI article also highlights the increasing consolidation pressure going forward as the investment cost in leading edge technology winnows out the weaker players.   Prima facia, then, these five leading foundries should outgrow their industry segment and be enjoying about 20-25% growth for the next three years!!   Does that mean a $20B TSMC in 2012!?   Wow!  More champagne (?) … or maybe something stronger.

Let’s look further.     We can compare these pure-play foundry actuals and forecasts to actuals reported by the Global Semiconductor Alliance (GSA) and forecasts by the SIA.     The graph below shows three independently developed actual and forecasted series on the same graph.     First is the series that is the IC Insights view of the pure-play foundry market, second is the actuals for the fabless industry as reported by the GSA and the third is the percent that the fab-less segment represents of SIA actuals and forecast of the overall semiconductor industry.     Observe some relationships:     First, the fab-less segment has grown to 20% of the overall semiconductor industry and been fairly stable at that percentage for the past 4 years.        Note also that the pure play foundry business has, at the same time, been in “lock step” with the fab-less segment in the years 2006 through 2009 at 40%.  This makes perfect sense as the foundries basically serve the fab-less segment’s cost of goods.   But if the IC Insight and iSuppli forecasts are right then something structural is about to change and change dramatically.     Assuming the IC Insight and iSuppli forecasts are right and assuming that “40% of fab-less” generally holds, the resulting forecast for the fab-less segment is shown (yellow “hatched” columns).    That forecast as a percent of the SIA forecast is also shown.    If right, that says the fab-less segment will expand dramatically to become about one third of the overall semiconductor market!!    Again, wow!!

Fab-less forecast implied by IC Insights pure-play foundry forecast

Frankly, I doubt it.    But IC Insights and iSuppli are not naive and two independent forecasters have developed surprisingly consistent views of the market.    Something else is happening.    Is this because IDMs are entering hybrid supply models?    Certainly the ATOM initiative at INTEL is a strong indicator in that direction.     Regardless, these numbers are dramatic and more questions need to be asked.    Simple solution is to buy the reports and find out what’s driving it.     Unfortunately, not in my budget this year but I will ask some more questions and find out what’s going on.