In the last couple of months there have been a number of changes in the views of some of the leading semiconductor industry analysts. Dr Mike Cooke surveys the semiconductor flight path

In the last couple of months there have been a number of changes in the views of some of the leading semiconductor industry analysts. Dr Mike Cooke surveys the semiconductor flight path.

It is always best to be sceptical of analysts and their forecasts. This is not to say that they should be ignored - their work can give valuable insight and guidance. However, they can get it spectacularly wrong - many lay the blame for oversupply in the DRAM market at the door of overoptimistic forecasts in that market during the mid-1990s. And the technology boom/bubble that came to a crescendo in 2000 . . . ? In the final analysis, it is decision-makers (and events over which no one has any control) that determine the state of an industry.

Most commentators see 2004 revenue growth in the range 24.4-37.7%, but next year, as one expects, is far more uncertain. Two companies see the market contracting, the most extreme decrease being 6.8% (Figure 1). The most optimistic estimate for next year is growth of 28.0%. While the commentators gather a number of external factors - GDP, world affairs, industry and consumer demand, and so on - we will focus more on the semiconductor production industry itself.

Fig.1: Semiconductor revenue growth forecasts 2004-2005

Peaky

Gartner sees IC revenue growth in 2004 at 24.8% and in 2005 at 13.8%. It has recently released capital expenditure and equipment revenue forecasts up to 2008. The market researcher expects a peak and the slide to the next downturn to begin next year.
"The surge in new equipment sales in 2004 is a direct result of demand visibility and capacity tightness," says Klaus Rinnen, vice-president for Gartner's semiconductor manufacturing and design research group. "However, now that capacity increases are keeping pace with unit demand, equipment orders are slowing. It seems the industry is attempting to more closely match its supply and demand ramps to maximise much-needed profits."
According to Gartner, capital spending in 2003 of $29.661 billion is set to grow to $44.763 billion (+50.9%) in 2004 and peak in 2005 at $50.767 billion (+13.4%). The period 2006-2008: $43.058 billion, $35.693 billion, $39.872 billion. Capital equipment revenues in 2003 were $22.824 billion. This is expected to increase 63.5% in 2004 to $37.317 billion and again peak in 2005 at $42.912 billion (+15.0%). For 2006-2008: $35.230 billion, $27.806 billion, $32.439 billion. Package and assembly equipment is expected to lead the downturn with growth in 2005 at the 2.4% level. The regional distribution of spending is estimated at 44% for Asia-Pacific, 23% for the Americas, 21% for Japan and 12% for Europe and the Middle East.
Rinnen is hopeful that if the semiconductor industry is mindful of what it brings on line, it could achieve the controlled or "soft" landing that has eluded it in the past.

Sweet spot

iSuppli revised its forecast for 2004 semiconductor revenues to $226.5 billion, a 24.4% increase over 2003's $182 billion. This market researcher believes that the industry has hit the "sweet spot" economically with no signs of the kind of overheating that led to the extended downturn from 2001. Moving to 2005, growth is expected to slow to 11.8%. In 2006, the market is expected to flatten to a nominal 0.1% increase - a soft landing. The new cycle is forecast to begin in 2007 with 9.2% growth to $276.6 billion and 10.4% to $305.4 billion in 2008.
"In the future, we'll look back to 2004 and say 'Those were the good old days'," says iSuppli's Gary Grandbois.
On the demand side, all major electronic equipment application markets will experience growth simultaneously this year - this is the first time this has happened since 1999.
iSuppli had previously expected semiconductor production oversupply to be reached within two years as a result of installations beginning in H2 2004 and continuing into 2005, leading to declining chip prices and a stalling of the industry in 2006.

Cooling system

VLSI Research's revenue growth estimates are 26.2% for 2004 and 11.8% for 2005. Looking mainly at semiconductor equipment sales it warned in April 2004 of "overheating" and a repeat of 2000's "2 for 1" deal, with two years of growth packed into one.
In June, VLSI Research reported its May estimates with the comment that: "The equipment industry cooled off somewhat in May 2004." The market researcher's world semiconductor equipment book-to-bill ratio for May has since been finalised at 1.08 (bookings $4.28 billion, billings $3.97 billion) and front-end capacity utilisation put at 99.7%. The billings figure represents a 74% increase on May 2003 and the bookings 48% growth. A preliminary book-to-bill for June is 1.12 (bookings $4.9 billion, billings $4.4bn) and July is forecast at 1.13 ($4.78 billion, $4.23 billion).
Front-end capacity utilisation rate has been trending downwards slightly - 99.7%, 98.9%, 98.1% - for the period May to July. However, VLSI expects utilisation to remain above 95% for the remainder of 2004.

Negative sentiments

IC Insights has led the downwards sentiment with a forecast for 2004 revenue growth at 31.0% and a shrinkage of 6.0% in 2005. Market research vice-president Brian Matas points to a number of factors leading to this conclusion. Despite the shaky revenues of the past few years, unit shipments have shown double-digit percentage growth for three years, including 2004.
Matas comments: "Since 1978, there has NEVER been a 4-year period of double digit IC unit volume growth."
In 2002 and 2003 the growth was 15% and 2004 looks set to reach 24%. Other three-year double-digit growth periods were 1982, 17%; 1983, 23%; 1984, 50%, and 1986, 18%; 1987, 16%; and 1988, 16% (Figure 2).
Matas points out: "In each of the two previous cases where there has been three-year double-digit growth, unit growth the following year experienced negative growth (1985, -16%) and soft growth (1989, 6%), respectively. Since 1978, the world-wide semiconductor market has NEVER grown more than 8% in the year after world-wide semiconductor industry capital spending increased greater than 50% - the 2004 semiconductor cap ex forecast is 53% growth."
Matas feels that the electronics industry needs time to digest the number of chips that have been produced over the past couple of years. Other factors feeding into IC Insights' prognosis are the traditional economy shrinkage following a US presidential election, the expiry at the end of the year of a US tax incentive to buy electronic office equipment and the Chinese government's decision to reign in its GDP growth to 8-9%.
"Our 2005 forecast may not pan out exactly as we show, but based on some of these inputs, historical trends seem to show that much slower growth is on the near-term horizon," Matas adds.

Fig.2: Despite the shaky revenues of the past few years, unit shipments (left) have shown double-digit percentage growth (right) for two years - three if one includes 2004. On previous occasions that have exhibited such behaviour, deep downturns have followed

Adding its voice to the negative 2005 view, Semico Research forecasts that IC revenues will grow 33.8% to $222.6bn this year. Additional strength in pricing could even push this into the high 30s. Based on this, next year the market researcher sees a decline of 6.8%. The company has therefore increased its pessimism from a previous forecast decline of 5.0%.
"The 2004 semiconductor market conditions are very similar to what occurred in 2000," says Semico president Jim Feldham. "Both 2000 and 2004 are characterised by strong GDP, an election year, and upgrades and growth in the cellular phone market. New cellular subscribers, smart phones and camera phones are a key factor in driving semiconductor demand."
In 1999 and 2000, revenues increased 18.9% and 36.8%, respectively. Last year's growth of 18.4% already matches the first part of the pattern, to which Semico adds a 34-39% prediction range for 2004.

Revisionism

The US Semiconductor Industry Association (SIA) has also revised its predictions with a compound annual growth rate (CAGR) of 10.4% through to 2007. This year is expected to be above average at 28.6% resulting in world IC sales of $214 billion. This compares with the SIA's previous forecast for 2004 of 19% made in autumn 2003. This beats the $204 billion sales of the previous record year, 2000.
"While this [compound] growth rate is lower than the historical growth rate of the past several decades, it represents very healthy growth for a $200 billion-plus industry," comments SIA president George Scalise.
Next year, growth is expected to be 4.2% and in 2006 sales are expected to fall by 0.8%. The year 2007 is expected to show a rebound to 11.7% growth with sales reaching $250 billion.
Regional market share for IC consumption is expected to reflect further moves to Asia Pacific, going from 40% now to 43% by 2007. The Americas is expected to lose share, going from 19% to 17%. Europe is expected to have a steady 20% share over the period. Japan is also forecast to maintain its IC consumption at 22%.

Buoyant chip sales forecasts in 2004 and warnings of slower growth next year will probably negatively impact chip equipment sales this year, according to the Information Network.
"Negative sentiment from analysts and forecasts of a relatively flat 2005 from the SIA are pointing to eventual pushouts and cancellations of equipment," says Dr Robert Castellano, president of the market research organisation.
In a forecast issued in November 2003, the Information Network put 2004's front-end equipment sales growth at 21.0%. The firm is looking at its figures as a result of recent IC sales forecasts. The Information Network says that it anticipates that orders will decelerate quickly towards the end of the year.

Minority report

The current forecast from European researcher Future Horizons for 2004 is +32%, unchanged from its numbers first published in January 2004.
Malcolm Penn comments: "Our 32% 2004 number, was at that time, the highest of all but since then, most other forecasters, including the SIA (originally in the low teens) have all raised their numbers, either equal to or much closer to our 32% number. We did not need to change our 2004 number because our analysis and reasoning was, we believe, much more accurate than the others, and the current spate of upwards revisions vindicates our position. "
Penn also says that his company was the only one to get the 2003 number and analysis right. "Our January 2003 forecast called for an 18% increase in the market - again the highest number then on the block - and we refused to change our number mid-year, at the time when all of the others, were constantly downgrading theirs." The final number turned out to be 18.2%!
As for 2005, Future Horizons sees two clear scenarios - the "doom and gloom" short-lived recovery prognosis of the majority of forecasters and a more optimistic minority scenario championed by itself. According to Future Horizons, the pessimist view is that a collapsing global economy and over investment/excess capacity (300mm driven) will tip the 2005 market into recession. Penn strongly disagrees and finds himself flying against popular 'wisdom'.
He adds: "The minority scenario, championed by us, believes that the global economy will not collapse and that it is impossible for the industry to over invest in 2004. Granted there are economic uncertainties, but we believe these will merely dampen - not kill - the rate of economic growth in 2005 and, as such, our 2005 number reflects a sizeable slowdown in IC unit growth compared with either 2003 or 2004. On the capacity front, it is virtually impossible to over invest this year, despite the fact that capital spending plans are up significantly on 2003. The recession-decimated equipment industry and its outsourcing infrastructure simply cannot deliver the goods fast enough to kill 2004, or even the first half of 2005, and with half the current year already over, there is not much time left to influence 2004. Remember, it takes between one and two quarters from receipt of equipment delivery to the start of producing saleable, qualified first wafers out, so gear delivered today, equates to first IC sales earliest in January 2005."
At a push, and with a following wind, Penn even sees capex spend possibly ending up at around $50 billion in 2004, up about 60% on 2003. Even that spending level only puts the industry back to its long-term average level, according to the Future Horizons' analysis. This does not begin to start to compensate for the fact 2002 and 2003 were both way below the average, with 2003 being the lowest level on history since records began.
"We thus foresee under supply lasting well into 2005, causing IC ASPs to harden quite strongly, hence our 28% 2005 forecast (13% in units and 15% in ASPs) with the potential for over-investment not biting home until the second half of 2005 at the earliest. The precise danger point will be clearer once the 2005 cap ex run rate is determined at the end of this current year."
Penn sees the down side to the 2004-2005 growth spurt as being that IC lead-times will inevitably extend (they already are) triggering an increase in OEM inventory levels. This additional IC demand is a one-off adjustment, sitting on top of the real unit demand and artificially inflating unit orders for suppliers. This sort of behaviour reached a staggering 60% level in 2000 just before the market collapsed. Fab investment will inevitably target the higher unit demand number - overshoot is inevitable. Once capacity surpasses the order book (consumption plus inventory adjustment) unit demand, lead-times start to drop back, inventory gets liquidated, orders start to fall and capacity instantly goes into excess, starting a downward cycle and triggering the recession. The timing and severity of the inevitable market collapse will depend on the slope of the capacity ramp and the disparity between real and perceived unit demand.
"Future Horizons' minus 8% forecast for 2006 is based on the assumption that over investment will not be so bad as it was in 2000, and that the economies remain steady in the 2005/06 time frame," comments Penn.

Conclusion or confusion?

In the past couple of months, there has been some gathering in of the majority views on next year around a "soft landing" prospect. However, normal semiconductor industry behaviour tends to the extreme. In that case, the lead question is to which extreme are we heading short-term - boom or bust?
Malcolm Penn at a recent review in London suggested that one of the considerations that leads his group to expect a delayed, but significant downturn in the 2006 frame is that caution currently reigns - "not enough champagne has been drunk yet".

Fig.3: 40 years of IC revenues. The approximate straight line on the logarithmic scale version (right) indicates the power of the long-term growth factor

Fig.4: 40 years of IC sales growth