2004 Economic Forecast – and Beyond

People are actually beginning to feel optimistic about the state of the semiconductor and semiconductor equipment industries, and industry analysts and economists are some of those with increasing hope. As part of Semiconductor International's 25th anniversary, we asked the analysts to give us — along with their usual prognostications of the year ahead — some historical perspective on the past 25 years, and some predictions on what we have in store 25 years into the future. For more from industry executives, check out our CEO Roundup .

Advanced Forecasting's quantitative forecast model, which has accurately predicted each of the major turning points in the industry for nearly two decades, shows growth for 2004 to be healthier than 2003. While the underlying demand for semiconductors is forecasted to be strong throughout the year, the industry leaders' actions will determine whether the industry is capable of maintaining this strong growth, or overshooting and propelling itself into another recession.

In our contribution to the Semiconductor International economic forecast published last year, Advanced Forecasting stated that "drivers" are not necessary for an upswing to occur. We continue to maintain this position. Growth in the semiconductor industry is fueled by the underlying demand for chips, which will then allow the drivers to emerge and become viable. Only in hindsight will one be able to pinpoint the driver of a particular boom.

An inevitable repetition of the 1999/2000 boom looms on the horizon for the silicon cycle. Before dollar signs blind the masses and the feeling of jubilation overshadows reality, we advise companies to exercise caution during this period to prevent the pitfalls of the 1999/2000 boom: Avoid over-hiring, over-building, over-ordering and over-heating.

During the past 25 years, the semiconductor industry has endured extreme volatility resulting in blissful booms and debilitating recessions. Each cycle has differed dramatically from those that came before it and those that followed. The pattern of one cycle cannot predict the amplitude of another. The next 25 years of the silicon cycle will show that history does repeat itself. The industry will continue to forget the pain of a recession when in the midst of a boom, and discount the repeat of a strong recovery in times of downturn.

2003 has been a year of transition for the semiconductor industry. After turning the tide from a steep 2001 decline during last year, strengthening semiconductor demand is expected to have returned chip sales to double-digit growth (at 11.7%) in 2003, setting the platform for a significantly stronger 2004. Gartner Dataquest forecasts semiconductor revenue to expand by 21% in 2004 over the prior year.

Increased chip demand combined with a cautiousness in increasing capacity in 2003, has finally allowed the supply-demand balance to swing the other way, causing a capacity crunch and likely spot shortages in 2004. Given a conservative and cautious investment pattern by chipmakers, the wafer fab equipment sector remained slow in 2003. In fact, it was a year in which it was good just to be able to see a flat year for revenue with a strong upside in orders for the third and fourth quarters of the year. This strong growth coupled with the improving demand for chips and tight fab capacity should enable a strong sales growth in 2004 for wafer fab equipment.

With current capacity close to being fully utilized, and most of the 300 mm fabs only partially filled, there is the potential for a fast ramp depending on the order rate from the semiconductor manufacturers, and the capability of the equipment manufactures and the supply chain to fill orders. Initially, it appears that the growth will be gated by the ability of the equipment manufacturers to ramp the supply chain and gear up the manufacturing engines to deliver equipment. Gartner Dataquest estimates that 2004 should see 34% growth for wafer fab equipment, and there is an up side.

So, the cycle rolls on. Yet, there are larger and longer-term issues to ponder. During the past 25 years, many technology advances have pushed the industry forward, bipolar to CMOS, WSi to TiSi to NiSi to CoSi, aluminum to copper, and from 100 to 300 mm wafers, just to mention a few. During this period, innovation and the drive to build technology has driven the industry forward and continued to keep Moore's Law on track.

As the industry looks to the next 25 years, many technology advances will take place that we have not yet even considered. While sometime during this next 25 years, there will be a physical limit to the scaling due to the natural limits of science, the ability to manufacture new creative applications will continue to drive the industry forward into the future.

From 1970 through 1995, a span of 26 years, a decline in the worldwide IC market was a rare occurrence. During that time, the semiconductor industry endured only two annual declines — 1975 and 1985. However, from 1996 through 2003 (a span of only eight years), the semiconductor industry endured three negative years — 1996, 1998 and 2001.

IC Insights believes that overall lower semiconductor industry growth rates during the past eight years, as well as the increased frequency of negative growth years, is due to a combination of factors. These factors include weak IC average selling prices (ASPs), a significant growth rate reduction in the semiconductor content percentage in electronic systems, and a decline in the average annual growth rate for electronic system sales.

Although the future average annual semiconductor industry growth rate is forecast to be lower than historical figures, the volatility of the industry is alive and well. For example, the 63-point semiconductor industry growth rate swing registered in 1984-1985 (46% up in 1984, and 17% down in 1985) was exceeded 16 years later in 2000-2001, which displayed a 69-point swing (from 37% growth to a 32% decline). Thus, while semiconductor market average annual growth rates may be "maturing," the volatility of the industry has yet to show such behavior.

IC Insights is a firm believer in the continuation of economic and business cycles. Whether it is worldwide GDP growth, electronic system sales, or the semiconductor market, we are confident that the cyclical nature of these segments is still intact.

For 2004, we believe that a healthy worldwide economy (+4.2%) and strong electronic system sales (+11%) will spur a 27% or better semiconductor market increase (surpassing the all-time semiconductor sales high set in 2000).

Over the 2003-2008 time period, the IC market CAGR is forecast to be 13%. During this period, IC unit volumes are expected to continue growing at an average rate of 9% per year while IC ASPs are forecast to increase an average of 4% per year. Thus, in the next industry cycle, IC market growth rates are expected to be driven more by unit volume shipment growth than by ASP increases.

Those that are waiting for the semiconductor industry to return to normal after the 2001 crash will be disappointed. The industry is recovering (or is already recovered, depending on whether you measure recovery by revenue or by unit shipments), but the post-recovery period will see a new kind of "normal."

From the market perspective, the biggest change is the absence of a single end-product driver. Unlike earlier periods, when products like mini-computers, PCs and mobile phones were in large part responsible for semiconductor market growth, there is no single new product on the horizon with both the semiconductor content and potential market size to account for several tens of billions of dollars in semiconductor sales. This is a good thing for the industry, because it reduces the chance of a massive meltdown when a single market cools off.

On the supply side, increasing commoditization of the CMOS process, along with a reluctance to invest in fabs on the part of many IDMs, will accelerate the trend toward foundries. Long term, this is expected to depress ASPs (and semiconductor industry revenue growth) as it increases competition at the wafer level. It will also exacerbate the swings in ASPs throughout the cycles, as foundries compete on price during soft markets and attempt to recover their losses during tighter markets. Increased design, verification and mask costs will provide a filter that reduces the portion of the total semiconductor product set that chases the latest process geometries, increasing the opportunities for foundries to compete at the sweet spot.

Within the context of this new reality, we expect that 2004 will be a good year for the semiconductor industry as demand, driven by recovery in the overall economy, outstrips supply. The combination of rising unit demand and rising ASPs, caused by short-term capacity shortages, is expected to result in 25% revenue growth. This robust growth is expected to continue into the early half of 2005, as a combination of cautious semiconductor producers and a decimated semiconductor equipment industry result in delayed equipment orders followed by extended lead times. When capacity finally catches up with demand, we can expect another downturn, but that is not expected to occur before mid-2005.

It's finally taking hold! The unit volume growth and the proliferation of applications and end markets are, after a rather long hiatus, gripping the chip sector. Now we need new capacity.

If the last few months of 2003 were any indication, the first few quarters of 2004 will be a barnburner for the semiconductor capital equipment field. In talking with players in the supply chain, it appears as though new orders for capital equipment could be up more than 50% in the first and second quarters. These numbers are very attainable because the industry is coming off an extremely low base.

I am concerned with the possibility that this capital spending cycle will be a violent — one year or less — phenomenon. Certainly, this concern can be offset but stronger end demand. Given that we are heading into an election year, one should lean toward continued economic strength. That should light a spark under the corporate spending arena — which still leaves a lot to be desired. The consumer markets have been the home of the action this year. New display technology will be the big driver during the next few years. I can add to this, rather confidently, that the personal computers, particularly portable computing platforms, will see high demand.

My main concern for 2004, like last year, rests with industry profitability. Profitability in the capital equipment field continues to be weak. Profitability with chipmakers is better, but not great. This weakness is hamstringing new technology development. Yes, the industry is transitioning to smaller geometries and larger wafers, but for the pace of new development to stay in line with Moore's Law we will need to see higher R&D spending levels. Stronger bottom lines would help on this front.

In the semiconductor industry, it seems the more things change, the more they stay the same.

In 1979, analog ICs represented 19.8% of the total IC revenues. In 2004, after 25 years of digitization, analog ICs will account for 19.9% of total IC revenues. The market has changed in many unexpected ways, but the demand for all device types continues remarkably strong — far longer than many prognosticators anticipated. Picture, video and voice information drives the market, and the PCs, handsets and consumer devices that embody this information will drive the semiconductor market to 17% revenue growth in 2004. In 2004, the electronic equipment market will double its 2003 annual growth rate to give an 8% equipment growth push to the semiconductor market recovery. Stabilizing device prices and increasing semiconductor content will further fuel the 17% semi growth.

Resurgence in memory IC growth will occur in 2004 as DRAM, SRAM and flash memory will each show revenue growth exceeding 25%. DRAM will go from 17% revenue growth in 2003 to 27% in 2004. The revenue growth in 2004 will be aided by slowing ASP erosion as megabit growth has hovered around 50% for both years.

Beyond 2004, the semiconductor market will slow slightly as equipment growth settles; semiconductor revenue growth will be 12.9%, 1.2% and 9.7%, respectively, for 2005, 2006 and 2007. The slowdown in 2006 will have two causes — an expected equipment growth slowdown and an excess of memory fab capacity, which will further push growth down as memory ASPs slip. And for historical confirmation, 2006 will be the 25th anniversary of the 1981 industry downturn.

What can we expect 25 years into the future? Well, I'll still be talking up analog ICs. Power management devices, such as voltage regulators, will have carried on the tradition of a 17% CAGR, even as the general semiconductor market deviates from this historic growth pattern and settles into a 10-12% CAGR.

Semiconductor shipments will increase about 25% in 2004. The increase consists of 12% more parts, 8% higher prices and a 5% depreciation of the U.S. dollar. Every market will share in the expansion. Early in the year, parts for consumer imaging and wireless products will enjoy the fastest growth. Sales of these inexpensive end products respond quickly to a shift in the economic cycle from decline to expansion, and are getting a further boost from the rapid, non-cyclical market penetration of developing countries. But by the end of the year, capital goods end markets will be providing the major stimulus to semiconductor sales.

This is the first full year of an economic expansion that had reached almost every key country by the end of 2003. Worldwide, consumption, investment and government spending are rising together for the first time in three years. So semiconductor manufacturers will enjoy rising volume and modestly rising prices intermixed with price spurts for parts temporarily in short supply. This will lead inevitably to tighter supply conditions in 2005 when price inflation is expected to pick up to more than 10%, volume increases slip to less than 10%, and dollar deflation ends.

Production capacity tightened late last year, but began 2004 generally adequate. There is enough capacity expansion available well into the year from die shrinks, new fabs soon to start up and outfitting partially built facilities. Then the economic recovery will reach semiconductor production equipment suppliers and persist for several years.

Given the choice of looking back or looking forward 25 years on the semiconductor industry, looking back is perhaps the easier of the two tasks. The industry has experienced a tremendous amount of growth over the past 25 years, and we all can celebrate in how semiconductor technology and its breadth of end applications have changed the way we work, learn and play. Looking forward 25 years is the challenge, as looking ahead even one year has proven to be a trying exercise. Yes, technology will continue to advance, but forecasting the pace of change and the growth outlook is the dilemma.

The outlook for capital equipment market growth is finally brighter after contracting for two years. The recently issued SEMI Capital Equipment Consensus Forecast shows the equipment market will grow 8% in 2003, led by investments in assembly and test equipment. Wafer fabrication equipment has experienced strong growth in Japan and Korea, with equipment orders now strengthening in other regions as well. The consensus forecast calls for 39% growth in 2004 globally, and 18% growth in 2005.

Strong growth is expected for materials as well and, as demand increases, concerns are surfacing regarding possible shortages for large wafers and package laminate substrates. The SEMI Silicon Manufacturers Group is forecasting 15% growth in silicon wafer area shipment in 2004, lead by an 84% increase in 300 mm wafer shipments. The package substrate market will experience strong demand, though material suppliers and assemblers are struggling with how to maintain profitability with some of these packages.

The growth in 2004 will contribute to the technical advancement and the prosperity of the industry in the next 25 years.

It is very difficult to predict the future whenever record fluctuations occur with regularity. All forecasts must be based on the past, as no one can see the future. Fluctuations outside of the range mean that there can be no statistical basis behind any prediction. One should focus on indicators and drivers instead.

The indicators are clear: Chipmakers have visibility into the first quarter for the first time in years, and they are out of capacity. Most have believed others were building capacity, but those others were not. Foundry capacity is short, as is capacity in general. At the same time, chip inventories are low and lead times are stretching out.

The drivers are equally clear: The economy is in an upturn and the industry has never turned down into a rising economy. There is a Y2K backlash upgrade cycle underway. Businesses can't use old systems forever and the typical upgrade cycle is 3-5 years. The new law that cellular service providers can be switched without losing your phone number is driving a huge wave of demand. Many were waiting to upgrade until this law took effect, plus easier upgrades means more demand. There's lots of new technology to buy. Be it wireless access in laptops that have usable battery life or cell phones with digital cameras, there are lots of reasons to buy something new next year. Overall, it should be a very good year.