The latest industry semiconductor forecast from Malcolm Penn, Future Horizons. Find out what’s in store for the year ahead.
Blue Skies Or More Troubles For The 2007 Chip Industry
Any company in the microelectronics industry requires analysis to help future decision making. Malcolm Penn of European analysis company, Future Horizons provides his view on probable industry moves for 2007
The only real certainty in the semiconductor industry is its ability to constantly deliver uncertainty, and clearly 2006 results were a test of industry vision and faith. Entering 2007, the fundamental issue - “it's different this time” - remains cloudy and unclear. Yet, with the various 2007 forecasts now laid on the table, it is somewhat surprising to find a remarkable degree of consensus on the growth-rate number for 2007. Estimates vary from plus 4.6 to plus 12.0 percent, Figure 1, with Future Horizons, not surprisingly, at the top end of the spectrum. The overall consensus is for 9 + 2 percentage points which, given the margin of forecast uncertainty, in reality shows remarkable degree of forecast agreement.
The big problem with statistics is they can be very misleading. An 8.9 percent year-on-year semiconductor market growth does not seem very dynamic, especially following on from an equally modest 6.8 percent 2005 growth year, yet 2006 was a strong year for the industry.
The chip market fundamentals have never been stronger, with global economic growth strong, and capacity more or less balanced, albeit tilted to excess rather than under-supply, with IC ASPs now starting to trend upwards, this being the one bright spot in the December/Q4 results.
The fundamentals, however, as always are over-shadowed by “issues”, primarily demand/inventory imbalance and the dynamic mis-match between supply and demand. These “issues” are part of the industry fabric, and will always sit on top of the underlying trends. It is this interaction that makes the industry so difficult to navigate.
The strong GDP growth, Figure 2, clearly drives demand across the full semiconductor board, with so-called “killer products” occasionally distorting some product sectors in favour (detriment) to others.
Demand is thus far from linear, driven by seasonal and inventory (lead-time) issues and the killer product surges (and wanes). As such, the industry vacillates between periods of accelerating and de-accelerating growth, both on a monthly and yearly basis, Figure 3.
Semiconductor market growth is subjected to a further industry dynamic, namely ASPs, and this has been a major distorting factor in determining the industry growth potential, Figure 4. Whereas the market growth by value shows two distinct 10-year long-term growth rate trends over the past 20 years, with several smaller sub-trends within each period, the unit growth has been much more consistent.
The “slow down” in market growth has thus been an ASP, not a demand, issue which means it could just as easily bounce back the other way into a period of accelerating value demand. Statistics per se are bad enough, but statistics on statistics on statistics are almost akin to random number generation.
The difference between the value and unit growth rates is ASP, Figure 5.
IC ASPs are primarily a function of complexity, and are driven by new IC designs. These, however, typically take three years to come to market. The industry has also experienced a one-year delay in bringing new designs to market due to the 130nm transition impact, which meant the full impact of post-2001 crash new design activity was not felt until 2006. Supply and demand/elasticity pricing issues sit on top of this dynamic causing either dramatic ASP reductions (in times of severe excess supply) or more gradual price increases (in times of product shortage).
Sometimes these can be seriously distorting, as with the mid-1990s DRAM and 2000 dot-com price bubbles. One thus needs to look at the month-on-month, quarter-on-quarter and year-onyear trends and how these are interacting to throw light on their significance, Figure 6.
The really good news is that ASPs are trending up, both on a month-on on month and quarterly basis. The even better news is that both of these trends have long-term periods, so there impact on the year-on-year growth rate is inevitable. In other words, if the monthly ASPs rise, the quarterly ASPs will follow, which in turn will drive the yearly numbers up, Figure 7.
With Q4-2006 up 0.4 percent on Q4-2005, the 2007 quarteron-quarter trends look set to be all positive, which in turn will drive 2006's 8.0 percent ASP decline positive.
Figure 8 summarises our 2007 market outlook, which calls for an 8 and 4 percent growth in units and ASP respectively.
Overall, we can expect to see a seasonally weak first half of the year, but with nothing unusual anticipated, with an associated oversupply trend, followed by a strong second half absorbing the first-half year capacity excess. Indeed, given the investment trends since mid-2006, if unit demand comes in higher than the 8 percent growth currently forecast, we might even exit 2007 capacity tight.
2007 should see the ASP recovery hold, structural and mix driven, with ASPs potentially coming in much stronger than our modest 4 percent increase.
The most remarkable aspect of post-2001 crash wafer fab investment trends is the ensuing restraint and moderation being shown by the industry. The mere sniff of a slowdown has people stepping on the investment brakes, with a serious degree of caution exercised before reapplying the gas pedal. This is very good news for the industry overall, and should help moderate the year-on-year cap ex growth swings.
It will do little, however, to moderate supply and demand on a quarter-by-quarter basis, due to the fundamental hysteresis in the investment dynamics, Figure 9.
The slow down in Cap Ex that started in July will thus not impact capacity until mid-2007, with capacity growing faster than demand in the meantime. This will be especially aggravated by the seasonally weak factors, meaning utilisation rates will fall in Q4-2006 through Q2-2007.
The underlying trends, as with unit demand, continue to look healthy, with units up 12.3 percent CAGR between 2002-2006, versus a 8.8 percent increase in utilised capacity, the difference being accounted for by a parallel 3.2 increase in IC units per wafer start, Figure 10.
The fundamental driver for new capacity investment is IC unit demand, offset by line width reduction (more chips per wafer) and increased complexity (less chips per wafer). Whilst there are obviously structural ups and downs, the overall trend here is remarkably flat, Figure 11. This means, in the long haul, wafer fab capacity must track unit growth, which is currently trending at 10 percent CAGR, 1986-2006.
The ‘safe' cap ex expansion growth rate is thus 10 percent per year; any growth over this amount must be viewed with extreme concern. In absolute terms, over the long term, this translates into 90c per incremental IC unit, or approximately $720 per 200mm equivalent wafer start, Figure C12.
Current trends since the 2001 crash have averaged only 60c per wafer start, primarily the result of the efficiencies gained by the conversion to 300mm wafers. A similar gain was experienced at the 200mm conversion build out in the early 1990s. A 5 percent Cap Ex increase for 2007 (i.e. US$ 2.1 billion) would thus yield an incremental 3.5 billion MOS IC units, representing a 5.1 percent overall IC unit growth. This is less than the 8 percent growth we are currently forecasting, meaning an increase in capacity utilisation by 2007 year-end.
What could go wrong? As always, the economy, capacity spend and inventory overbuild. Right now the economic growth looks sound, with oil and inflation worries seemingly under control, but economic wobbles and recessions are notoriously unpredictable by nature. The bottom line is, if the global economy collapses, it will take the chip market with it.
Over build in capacity seems an unlikely scenario based on current behaviour, so capacity excesses will likely be restricted to the inevitable impossibility of balancing short-term supply and demand. A reduction (increase) in Cap Ex take around one year to impact supply. Inventory, inevitably, will be the minefield in the equation; it is impossible to get right and to react fast to demand changes. Never underestimate the ability of this industry to throw surprises; there are no safe havens or lucrative highvolume niches. Competing in the chip industry is not for the feint-hearted.