Electronic system production
Top 10 growth list
Reducing cost per function
Most people are happy to have 2009 in their rearview mirror and are looking forward to a much better 2010. In "The McClean Report 2010 Seminar" held in Scottsdale, Arizona on January 19th and in Sunnyvale, Boston and London later in the month, IC Insights market analysts put the worst recession in the semiconductor industry in perspective and reinforced the optimism for 2010. The bottom line is a conservative forecast for 2010 of 15% growth with a high probably of even better than that. There are some expected and unexpected changes from previous recoveries.
The 15% growth is in line with projections from other forecasters such as Databeans. Gartner’s projection is slightly more pessimistic showing a 13% growth, and IDC’s Semiconductor forecast is a little vague, indicating double digit growth in 2010.
For the IC industry, the recovery was well underway by the second half of 2009 says Bill McClean, president of IC Insights. He’s also the namesake for the McClean Report, a competitive analysis and forecast of the integrated circuit industry. Electronic system production in 2009 (Fig. 1) of $1.1 billion represents about 2% of the world GDP of $55.6 trillion and was down 11% from the 2008 level.
“We kept telling people ‘think quarterly’—the second half will be a lot better than the first half,” said McClean. After the recession hit bottom in March, the second half brought seasonal strength for electronic systems, the completion of IC inventory adjustments and the start of shortages. A decline of 13% in dollar growth of semiconductors in the first quarter of 2009 followed the 26% falloff in the last quarter of 2008 and preceded a 16% growth in the second quarter and subsequent growth of 16% and 9% for the remaining quarters.
Semiconductors rely on electronic systems for growth. Over the past 25 years, electronic system production has grown around 6 to 7% on average per year. “We are forecasting a 7% increase in electronic system production which is about the same as the long term average,” said McClean. In fact, semiconductors performed fairly well compared to the GDP. “It was the worst global recession in 63 years and the semiconductor market, in dollars, only declined 10%,” explains McClean.
Looking forward the outlook is optimistic. “Our forecast is for 15% growth in 2010,” and I would label this a conservative forecast,” he says. “We think there is quite a bit of upside potential to this number.”
Capital spending was down 41% for semiconductors in 2009 with a forecast of 45% growth in 2010. This growth will be necessary to fuel the growth in semiconductors and avoid pricing pressures that can occur with supply shortages. The materials market, wafers, chemicals and gases as well as packaging materials, slipped 10% in 2009 but should experience a surge of 17% in 2010 “pretty much follows the unit growth” according to McClean.
One of the major differences for predicting future pricing is the lack of new companies coming into the marketplace. China and India will not be investing any further in new semiconductor manufacturing capability as Korean, Taiwanese and initially the Chinese manufacturers have done in the past. When new manufacturers in these regions entered the market, reduced pricing was the incentive to capture business. Besides the $4 to $5 billion investment it takes to get a new fab up and running today, there is no real entry point according to McClean. DRAM, high-end microprocessors (MPUs) or flash memory are not good starting points. “That’s a big change from the past 30 years,” he proclaims.
Major assumptions for achieving a 10 to 20% forecast range include GDP growth of 3.2 to 3.8%, IC capacity at 90 to 92% utilization, IC unit volume shipment of 12 to 18%, IC ASP change of -2 to+2%, a neutral effect of foreign currency on the US dollar exchange rate and no major terrorist attacks or wars.
IC PRODUCTS GROWING IN 2010
Looking at IC product categories provides greater insight into specific growth areas. “Things are definitely on the up-turn,” says Brian Matas, vice-president, Market Research, IC Insights. “We are going to see some exciting numbers in 2010.”
Of the twenty-eight product categories for the IC market, eleven are forecast to provide the Top 10 growth list shown in Figure 2. Automotive is a driver for the second fastest growing item, 32-bit microcontrollers (MCUs). Not surprisingly, with the automotive slump in 2009, MCUs suffered a decline of 22%. Four categories showed growth in 2009 but the highest level was a mere 5% for telecom application-specific analog ICs.
A combination of a 12% increase in units and 17% increase in average selling price (ASP) for applications in growing PC shipments, Windows 7, netbook computers and other devices that require DRAM play a role in DRAM growth of 31% forecast for 2010. But the growth could be even higher. “Even higher than this 31% that we’re showing,” notes Matas. “Capacity situations can really drive the ASP up.” The rebound in applications related to automotive, computer and consumer impact many of the other high-growth categories.
While flash memory experienced additional growth in early netbook computers, the more recently introduced netbooks have gone back to spin drives. An interesting application for flash memory replacing hard disk drives is in servers and data centers according to Matas. In these high-end situations, users can experience a cost savings by using flash memory. Part of the savings is from the electricity standpoint both to power the memory and almost equally to cool the equipment. “By using flash memory, it helps eliminate a lot of the excess heat that is coming off some of these spin-drive type computers,” says Matas. “More so than that, the flash memory is faster, so it is able to access some of this data at the centers a lot more quickly than the spin drives.” The growth in data center usage is reflected in NAND in solid state drives growing from 17% to 26% of the total NAND growth in 2014 of $33 billion.
IC insights’ projects that high-volume technology driver applications of cell phones and personal computers will increase from 1125 million units in 2009 to 1630 million units in 2013 and from 286 million (combined portable and desktop PCs) in 2009 to 439 million in 2013, respectively.
For over 50 years, the value of ICs has been directly attributable to the industry’s ability to reduce the cost per function (Fig. 3) from 25 to 35% per year. “Whether that function is the DRAM bit, or megahertz, MIPS, regardless how you define that, we have been able to reduce the cost of that function about 30% per year,” says Trevor Yancey, vice-president of Technology at IC Insights.
With each generation, increasing the number of bits per DRAM or the speed of microprocessors (MPUs) has become more difficult. Alternate design approaches, such as to multi-core designs for MPUs and multi-level cell bits for flash memory, have helped to minimize the difficulty in maintaining the momentum. At the same time, the design/processing cost per device curve has a natural tendency to increase and is becoming more difficult to restrain.
Research and development costs as a percent of revenue today is about 17 to 18% and only a decade ago was 10 to 12% says Yancey. Wafer fab costs have decreased from 22% of sales down to 20, then 18 and now to 14%, but 14% appears to be the bottom. “The cost to design each new generation is becoming very, very expensive,” he said. In the future, integrating vertically in the third dimension could provide more functionality per given area.
Historically, increased functionality and decreased cost occurred by reducing the minimum IC geometry. This made each transistor smaller, while operating faster and consuming less power. Smaller transistors result in smaller chips and make chips more cost effective to manufacture. Historically, the minimum geometry has been decreasing about 13%/year but this is changing.
The transistor count on the highest density ICs, flash memory, is growing at 64%/year and is expected to reach 128 Gb by 2012 up from last year’s maximum of 64 Gb. However, future benefits in CMOS performance are expected to come primarily from materials innovation instead of just scaling. In addition to improved materials, packaging—specially stacked packaging for more advanced packages—should provide more effective space utilization in increasingly smaller end products says Yancey.
Using larger wafers is another technique historically used to reduce silicon cost. The industry’s use of 300 mm wafers exceeded 200 mm in terms of produced surface area for the first time in 2008 according to Yancey. With the recent recession, the industry’s anticipated move to 450 mm wafer fabrication in production has been pushed out to the 2014 to 2016 timeframe and the number of companies pursuing this wafer size has dwindled to only 14 from the current 29 owners of 300 mm fabs.