Home > DRAM, DRAM market, fab spends, fabs, Future Horizons, global chip market, global semiconductor industry, global semiconductor market, Malcolm Penn > Global semicon industry update Mar. ’10: Time for a reality check…pessimism has swung too far, says Future Horizons

Global semicon industry update Mar. ’10: Time for a reality check…pessimism has swung too far, says Future Horizons

March 30, 2010

January’s WSTS results continued to follow the underlying industry recovery trend, with ICs sales up 4.8 percent versus December (on a five-week month adjusted basis). They were also up 73.7 percent versus January 2009, a relatively meaningless number other than to recall just how bad things were this time last year.

The real significance of January is its potential impact on first quarter sales. Were this run rate to continue through February and March, first quarter sales would be up 8 percent versus Q4-09. That would make 2010 grow a staggering 40 percent on 2009. This is by no means a forecast but it does serve to illustrate the strength of the recovery from the abyss this time last year.

Ignoring the structurally (and typically) wild individual monthly fluctuations – which simply means no single month is a good indicator of the underlying trends – the month on month numbers will not settle down until the second quarter of 2010. That being said, given the likely strength of the first quarter versus Q4-09, our current 22 percent forecast for the total year now looks far too low.

Our 22 percent forecast for 2010 was based on the relatively benign quarterly growth pattern of -1.0, +1.0, +6.2 +2.0 percent; in essence a very weak year. No one we speak with is seeing a negative first quarter, with a consensus now building for at least 3 percent positive growth. That alone would bring the year on year growth up to 28 percent.

At the same time, almost everyone is also boasting a strong Q2 backlog with price stabilisation, even increasing; low inventory levels; and tightening supplies, which places severe doubt on the credibility of our plus 1 percent second quarter growth forecast. Were this to be say plus 3 percent, the year on year growth would be 30 percent.

It does however give us further confidence in our analysis and now places our forecast at the low end of the forecast range. Barring an epic 9/11, Act Of God or immoral banker style disaster, growth of anything less than 22 percent in 2010 is now all but impossible.

We fully expect to be increasing our forecast to around the plus 30 percent level at our forthcoming IEF2010 International Electronics Forum in Dresden, May 5-7 bringing the 2010 market within spitting distance of $300 billion in revenues.

The real irony behind this recovery is it is taking place in the first half of the year when things are usually quiet and the strength of the recovery is therefore understated. In addition no one believes (a) what they see it or (b) that it will last, even though there is not a shred of evidence to support a second-half year market collapse, quite the contrary. This is really a very serious problem indeed.

With everyone still running on empty – neither hiring nor spending money – the industry is in a very weak structural position to grow. Capacity is maxed out, wafer shortages are becoming rife, lead times are stretching and some firms are even paying their foundries a price premium to jump the delivery queue.

Some are also finding the low-ball orders they took at rock bottom margins are now loosing money due to foundry wafer price increases. Both of the issues (wafer deliveries and cost) are a fundamental problem of the fabless, and now fablite, business models. Never forget the sole reason for the FSA’s (now renamed GSA) formation in 1994 was to address the wafer shortage issue during the then market boom, following three years of low market growth and capacity under investment.

The chip market sentiment pendulum has clearly swung to far towards pessimism, driven in part by the 2000s decade of ‘lost’ growth. The overall IC market CAGR for 2000-2009 was a paltry 0.8 percent, the worst decade ever for the semiconductor industry, prompting cries of despair that the chip market glory days are over. We strongly disagree. Clearly, from a mathematical perspective the 0.8 percent CAGR number is correct but the conclusions to be drawn from this need to be interpreted with care.

For a start, the data range covered happens to measure a peak (2000) to trough (2009) period; the CAGRs one year either side of this period were 7.8 percent for 2001-2010 and 5.4 percent for 1999-2008. Moreover, looking at the corresponding values for IC units rather than US dollars shows the underlying annual 10 percent IC unit growth rate intact.

Herein lies the fundamental danger of statistics though. You can derive any CAGR value you want simply by choosing the right start and finish points. In so doing you can then ‘prove’ virtually any scenario you like, providing amply fodder for the optimists and pessimists alike.

The fact that IC units showed growth in line with their average points to the fact that the ‘problem’ with the 2000s was one of declining average selling price (ASP) not growth. The 2000s were thus a decade of depressed ASPs. The real question is thus not the low market value growth – this was the effect – but whether the cause – an above average decline in ASPs – was a bell weather of things to come, as many believe, or a temporary occurrence, the result of a coincidence of events? We believe the latter, as first reported in our November 2009 Report.

Just to recap and bring the situation up to date. ASPs are a very complex issue, driven not just by price increases but also product mix, IC innovation, fab capacity and production techniques. For sure the industry has seen declining ASPs since the 2001 crash but it is fundamentally flawed logic to extrapolate this into the future; a little like saying real-estate prices will forever keep on rising.

They do not; neither will IC ASPs keep on falling. We see the 2000s ASP fall as one side of a cycle; the coincidence of events rather than a sign of more bad news to come. It is vital therefore to understand the events that caused the problem.

First, the industry experienced a major yield bust at the 130nm node, delaying its introduction by a year and destroying the ASP price enhancement it would otherwise have brought. Second was the transition to 300mm wafers, the sole purpose of which was to reduce IC costs. A 2x plus increase in gross die per wafer for only a 40 percent wafer cost increase means a 40 percent die cost decrease.

As is typical in our industry, all of this cost reduction was immediately passed on to the customer meaning all 300mm wafer sourced ICs were reduced in selling price by up to 40 percent. By the end of the decade this was over half of all silicon made.

Third, for DRAMs, where fabs must always be kept fully loaded, the increase in die output due to the 300mm transition was more than the market could use meaning rampant oversupply and the mother of all price wars. It is only now that this massive one-off incremental capacity increase has been absorbed that pricing has return to its ‘normal’ pricing curve.

With DRAM demand hot – 4Gb is the entry point for 64-bit/Window 7 systems; strong demand for servers; new Intel processors in prospect; and a two to four year backlog in enterprise workstation upgrades – and Flash growing too, driven by exploding demand for Smart phones, the memory market has entered a positive cycle for growth and profits.

The last two years of DRAM Cap Ex restraint has now triggered a fab famine, the like of which no amount of die shrinking can offset. ASPs are already now double what they were just 12 months ago, with a minimum two-year period of positive ASP news now in prospect. The DRAMeXchange experts even say three. A fourth factor was the brutal Intel-AMD 32-bit MPU price war that saw ASPs fall around 30 percent from their more normal $100 level to $70. With AMD now bloodied, bruised and losing money, we can expect to see MPU ASPs trending up.

Finally, excess capacity also played its role but is already no longer a factor due to the significant slow down in new capacity investment over the past two plus years. Wafer fab capacity is now essentially sold out, with allocations, extended lead times and price increases the new industry norm. As mentioned earlier, some firms are already paying a price premium in order to jump the foundry wafer delivery queue.

Those that refuse will simply not get their parts. No wafers, no sales; yes it really is that simple. Interestingly overall industry revenue per wafer start increased to $7.70 per sq cm in 2009 from $ 6.96 in 2008, despite 2009 being the worst recession year in the history of the chip industry. Watch for this number to hit its US$8.00 to US$9.00 long-term average in 2010.

With current wafer fab capacity tight, additional capacity will now be driven primarily by Cap Ex, not one-off gains such as wafer size transitions, and this will be governed by industry’s willingness to invest – they currently are not – which translates into no new capacity for at least the next 12 months, due to last year’s incredibly low level of investment.

Even a 50-80 percent increase in 2010 Cap Ex – the current top end of the forecasts – will not significantly increase 2011’s net new capacity; the current starting point is so low. Prices, and therefore ASPs, will rise. 10 percent IC unit growth (the underlying growth trend) coupled with any positive ASP growth means double-digit growth at the IC value level.

This is all really good news for the industry as a whole but not for all companies. For a start, the OEMs will need to get used to a capacity (supply) limited market with increasing, rather than decreasing, IC buying prices. Secondly, the fabless and fablite firms will need to adjust to a world of tight foundry wafer supply and increasing prices. It will be a sanguine moment when they suddenly realise that they are no longer in control of the delivery times and prices they quote to their customers; their business is now at the mercy of their foundry partners. Better start to learn the new industry lexicon: “Please Sir… may I have some more?”

Industry capacity

Overall MOS wafer fab capacity increased marginally by 0.4 percent in Q4 versus Q3-09, from 1,877k 200mm equivalent wafer starts per week to 1,884k. Only 300mm leading edge capacity showed any increase in the quarter, at around 3.7 percent growth. This increase was not enough to offset the previous quarter’s 0.7 percent decline but is a reversal of the 1.6 percent quarterly decline reported this time last year. Overall MOS capacity is down 12.3 percent from Q4-2008 and on a par with where it was in the first half of 2007. Capacity has been essentially flat for the last three consecutive quarters.

At 640.4k wafer starts per week, Q4-09 200mm capacity continued its absolute value decline, from 666.8k in Q3-09, a fall of 4.0 percent. 200mm capacity is now down 22.5 percent versus the same period last year.

300mm wafers now account for 56.3 percent of the total MOS capacity, up from 54.5 percent in Q3-09 and 48.2 percent from the same period last year. 300mm wafers now account for over half the total capacity, with 200mm in second place at 34.3 percent, down from 35.5 percent in Q3-09 and 39.1 percent in Q4-08. Advanced capacity (i.e. 0.06 micron and below) grew 6.9 percent or 40.8k 200mm equivalent wafer starts per week, as leading-edge designs migrate to the 5x and below nodes.

As correctly predicted 15 months ago in our June 2009 Report Capacity review, the combination of capacity cutbacks and recovering IC demand caused total MOS IC Q4-09 utilisation rates to reach near sold-out levels, reaching 89.2 percent, up from 87.0 percent in Q3-09 and 68.4 percent for Q4-08. Advanced IC capacity, i.e., 0.06 micron and below, reached 96.2 percent (from 93.8 percent in Q3-09), whilst 300mm and 200mm wafers checked in at 96.7 percent (Q3 = 96.1 percent) and 82.4 percent (Q3 = 80.2 percent), respectively.

It doesn’t get more ‘sold out’ than this… and this at the START of the IC recovery cycle. Given the further 46 percent cut back in 2009 Cap Ex spending, 2010 capacity will be scarcer than hen’s teeth. Foundry price rises, extended lead times, allocations and premiums for priority delivery will dominate the landscape… watch out for an awful lot of fabless and fablite firms to be caught with their trousers down committed to woefully low IC ASPs based on anticipated continuingly low foundry wafer prices.

2010’s capacity cannot increase much beyond today’s level, so any increase in die output is dependent on shrinks and yield improvements. 2011’s capacity increase will depend on 2010 Cap Ex, off to a flat start on Q4-09. This means capacity will be tight through at least mid-2011 yet industry is STILL in collective denial.

We have said it before and we’ll say it again. There is already not enough capacity in place to meet 2010’s demand, 2011 will be even worse… the fablite model will be the worst hit by this shortage; depending on who you are, the fabless firms won’t escape unscathed either. Never forget the FSA (now renamed GSA) was formed in 1994 as a direct result of the wafer starvation caused by the early 1990’s Cap Ex underinvestment and the 1993-1994 market boom. Déjà vu?

Part two of this update will look at market trends and semiconductor developments.

Advertisements
  1. Ram Dwivedi
    March 31, 2010 at 3:32 am

    Hi Pradeep,
    Just loved your analysis.

    Thanks,
    Ram

  1. No trackbacks yet.
Comments are closed.
%d bloggers like this: