Don't look down
Cheap as chips?
The recession is already hitting the semiconductor business but tight capacity might mean a rapid turnaround in fortunes.
Future Horizons president Malcolm Penn has said at almost all of the seminars he has hosted in the last ten years that the semiconductor business can manoeuvre itself into having a bad year in a good economy. But when it's a bad economy, the chip business has to work hard to not fall into recession.
Most of the problem lies in chip pricing. Penn points out in his latest report that, in the last 22 years, unit shipments of integrated circuits (ICs) have shrunk year-on-year only twice. Once was in 1985 when the general economy was in rude health; the second was 2001 when GDP growth slipped below 3 per cent. The problem is that the value of the industry has fallen more often because pricing slips out of control during recessions.
When the economy sneezes, the semiconductor business does not just catch a cold, it contracts influenza with a side order of pneumonia.
Chipmakers predict bad fourth quarter
After the plunges in confidence and stock markets around the world, warnings from the chipmakers, which are more dependent than ever on the custom of consumers, seemed inevitable. Some of the falls in sales expected for the fourth quarter of this year are pretty large.
The Semiconductor Industry Association (SIA) saw the first indications of a slowdown in the September sales statistics, released at the end of October. Although growth fell markedly across all regions, including Asia-Pacific, the US market shrank 3.4 per cent compared with August 2008.
"The rate of semiconductor sales growth slowed in September as the industry began to feel the effects of the turmoil in world financial markets," says SIA president George Scalise. "We face a near-term period of uncertainty with a steep decline in consumer confidence and caution in the enterprise segment.
"The picture is somewhat brighter in emerging markets. Sales of personal computers and mobile phones - the two largest drivers of semiconductor sales - remain strong in these emerging markets, driven by growing consumer populations and rising income levels coupled with more affordable pricing."
According to the SIA, which collected statistics from a range of analysts, developing countries now account for nearly half of all unit sales of PCs. Intel, which has benefited from the PC's relentless drive into new markets, said in mid November that it expects its sales to be around 15 per cent lower than its initial projection for the fourth quarter.
At the company's analyst day in November, AMD chief operating officer Bob Rivet refused to provide an update on its guidance: "We are not going to change our guidance. It is too early in the quarter. But it is stormy out there - we all want to forget October."
Gartner claims semiconductor companies mostly met expectations for the third quarter of 2008, but guidance across the board for the fourth quarter continues to drop.
"Semiconductor growth was surprisingly strong until recently, given the very weak economic environment, but this will start to change in the fourth quarter of 2008," said Bryan Lewis, research vice president at Gartner. "Mounting evidence suggests that the semiconductor industry will see negative growth starting in the fourth quarter of 2008, and that this will continue throughout most of 2009."
The result is that Gartner estimates worldwide semiconductor revenue in 2008 to total $279.4bn, a 2 per cent increase from 2007. However, the analyst started off with a comparatively low estimate at the start of the year, based on problems in the DRAM and flash memory markets. Future Horizons was more bullish earlier this year, pointing to tightening capacity as the unit surge outpaced capital expenditure. Earlier this year, Penn claimed the market would growth some 12 per cent in 2008. That estimate has now come down to between 4 and 5 per cent.
In September, ahead of the latest turmoil, IC Insights lowered its projection for 2008 growth to between 1 and 5 per cent, but claimed the conditions could - as long as the financial collapse did not get worse - lead to a bounce-back as early as the second half of 2009. This scenario would lead to single-digit growth for 2009, with a much stronger 2010 in prospect.
Gartner expects worldwide semiconductor revenue growth in 2009 is expected to be 1 per cent, down by approximately 7 per cent from previous estimates. In the third quarter of 2008, worldwide semiconductor revenue for 2009 was forecast to be $307.7bn. Gartner analysts now expect worldwide semiconductor revenue in 2009 to total approximately $282bn. The SIA is now forecasting a 5.6 per cent fall to $246.7bn in 2009.
The semiconductor business is doubly cursed this time round. Although the industry seems set to enter a very bad phase, the previous eight years could hardly be described as good. "Profitless prosperity" was the phrase that analysts have used to describe the way in which the semiconductor business grew since the collapse of the Internet bubble at the end of 2000.
Aside from brief seasonal dips, generally after the rush to build kit for Christmas, the trend in IC unit shipments has been relentlessly upward since the start of 2002. From the start of 2005, output did not stop growing until the start of 2007. And, immediately after a quarter-long slowdown, unit shipments accelerated again, growing at up to 10 per cent quarter-on-quarter. Yet prices continued to drop. The average revenue per wafer - measured in terms of 8in equivalents - fell as much as 10 per cent quarter-on-quarter in the year-and-a-half from the start of 2007, only recovering briefly at the end of last year.
Despite the increasing complexity of semiconductor production and the number of functions that can be squeezed onto a chip, the revenue per wafer in the middle of the year sunk to less than what it was at the end of 1990 - the last period when a semiconductor and full global recession aligned. And this recession has barely got underway. What is going to happen to chip pricing once the consumer market sinks?
Sicas sees Fab fall
Fab utilisation has been robust through 2008 but began to drop off in the second quarter, according to figures from Sicas, the group that monitors production levels at the world's major chipmakers and foundries. For leading-edge nodes, from 45nm to 110nm, utilisation hovered around 95 per cent since the start of 2005, benefiting from a rapid migration from the 130nm process for the past few years. Utilisation for 130nm dropped off from 2006 as the cellphone, high-end consumer and PC chipmakers moved to 90nm and below.
In the second quarter, utilisation levels fell almost across the board and foundries found their book-to-bill ratios slipping to below 0.8, according to IC Insights. The only process nodes to see an increase in utilisation were around the 0.25µm node. But this increase is illusory - Fabs have been shutting down or converting these lines. Actual production of 0.25µm-class silicon fell in the second quarter compared with the first quarter of 2008.
For trailing nodes, all the way from 0.7µm and larger to 0.35µm, utilisation has dropped to 80 per cent or below from around 85 per cent.
Capacity has, historically, controlled the pricing of semiconductors. In times of plenty, vendors fight for sockets and will frequently build market share by doing deals. When silicon is scarce, they have the ability to hold pricing, particularly on higher-margin new products. When new chips enter the market in an environment of stable pricing, the average selling price moves up according to Penn's view of the semiconductor business.
In this scenario, there is some room for optimism despite the chaos about to happen. Spending on production equipment had already fallen to historically low levels even before the current recession started. As a proportion of annual IC sales, spending on chipmaking tools has, according to Penn, fallen to historically low levels.
The long-term view
Gartner claimed in October that capital expenditure would fall by a quarter in 2008 and another 13 per cent in 2009 before finally recovering in 2010.
Over the long term, capital expenditure has tended to run at around 20 per cent of annual semiconductor sales. However, this number was distorted by the massive spending encouraged by the Gold Rush atmosphere that surrounded the Internet and the 'Asian Tiger' economies in the mid- to late-1990s. With a longer-term perspective to consider now, having had eight years since the excesses of the Internet bubble, analysts such as Penn have reduced this figure to around 17 per cent. Even this may be too high.
If you compare the amount spent during the boom periods with the actual increase in wafer production capacity that resulted, it is clear that capital spending was much more efficient in the period from 1990 to 1995 and in the wake of 2001. The picture is distorted by fab closures during these periods but, on average, capital expenditure is at least twice as effective now than it was in 1999 to 2000.
Based on the spending patterns outside the periods of excess to calculate the capital actually employed per wafer-out, long-term spending has trended far closer to 15 per cent for the past 20 years. This lower level has been perfectly adequate to support double-digit growth in IC units and in an environment where pricing pressure has suppressed chip-market revenues. In fact, looking at the unused capacity currently in the market, this apparently lower level of spending is slightly outpacing unit growth in all but the most advanced processes, which are normally almost sold-out except in the deepest recessions.
Although Sicas does not have figures for the third quarter yet, SEMI has reported a small fall in wafer shipments from the second quarter, which indicates that the gap between capacity and output is opening up. Guidance from many companies suggests that the gap could widen further during the fourth quarter.
The question for the second half of 2008 as well as for next year is how far fab-related capital expenditure will drop below the 'real' long-term average. Analysts such as McClean expect unit shipments to catch up with infrastructure investment within the next year unless unit growth goes into reverse as a consequence of the wider recession.
If capital expenditure stays low for a year, the industry could move straight from recession into a shortage - it will not take long for the current utilisation gap to close - which will push prices up rapidly. If a process of consolidation takes place during next year, which seems likely given that many semiconductor companies will seem cheap relative to their cash value, the stage is set for a reversal of 'profitless prosperity'.
Given the greater efficiency of capital spending now, the silicon shortage may not be as pronounced as in earlier boom cycles. But, unless expenditure increases in 2009, it could be 2011 before customers of foundries and end-user equipment builders can breathe easy as capacity takes many months to bring online once the necessary kit has been bought.
Industry casts a wary eye on consumers
On top of a possible silicon shortage, a reason for expecting semiconductor prices to rise overall is the change in attitude among many of the semiconductor companies. During the most recent cycle, the focus has been on the consumer market and trying to establish market share in its many sub-sectors. If consumers do not have the spending power they once had, it does not make sense to chase them at any cost.
Similarly, the high cost of design for the most advanced product classes, such as cellphone baseband and application processors has convinced companies such as Freescale Semiconductors and NXP Semiconductors to quit those markets. Freescale has put its phone chipset business up for sale and NXP has agreed to sell the entirety of its operation to STMicroelectronics. Even Texas Instruments, which has a substantial market share in this business is focusing less on expensive digital-only parts in favour of mixed-signal products, which often carry higher margins.
Intel, NXP, TI and others see greater potential now in relatively low-cost, simple devices used for health monitoring and similar 'ambient intelligence applications than in complex gadgets. The target remains the consumer but in markets that may carry substantial government support as they try to rein in the cost of primary healthcare. The reasoning is that, if devices encourage people to take care of themselves, that puts less stress on the medical infrastructure. These devices need a higher proportion of analogue circuitry but are less complex – they simply use cellphones and PCs as information conduits.
|To start a discussion topic about this article, please log in or register.|
"The Internet of Things used to be a buzz phrase in tech circles, but it's already so last century. Brace yourself for the Internet of Everything"
- Can we make further productive use of Wylfa Magnox Nuclear Station site at Angelsey? [05:25 pm 12/12/13]
- The Nuclear Innovation Research Office (NIRO) and its new role [03:37 pm 12/12/13]
- Will we ever be able to generate energy from invisible dark matter? [12:49 pm 12/12/13]
- PCB Design Courses [07:46 am 12/12/13]
- 11kv overhead lines question [08:24 pm 11/12/13]
Tune into our latest podcast