14 Sep 2007
How is Europe placed in the global image sensor market? Tom Hausken of market analyst Strategies Unlimited looks at regional revenue shares and the consequences of a fabless versus integrated manufacturing approach.
The image sensor market is continuing to show strong growth with sales expected to increase by 14% to nearly $7 bn (€73.5 bn) by the end of 2007. When you consider that the image sensor market crossed the $1 bn mark in 1999, some 26 years after commercialization, and that growth alone has exceeded $1 bn per year in recent years, these figures are nothing short of phenomenal. But now that the market has passed the inflection point in a very long S-curve, how are the suppliers reacting?
European companies have regained some worldwide market share. We have estimated that in 2006, European-based companies were responsible for about 10% of the worldwide production of image sensors. Although this is about the same level of production as in 1996, it is an increase from less than 5% in previous years thanks to strong sales by STMicroelectronics (ST) in CMOS arrays for mobile camera phones and other consumer products.
The regional revenue shares for recent years are shown in figure 1. The graph shows manufacturing in non-Japanese Asia (Taiwan and Korea) for 2006, whereas in earlier years it was counted with Japan. Similarly to Europe, the share held by Taiwan and Korea is expanding. It is striking to see how Japan is losing market share. Japan's image sensor revenues grew through the period, but not as fast as newer players such as ST and Micron. The net effect is that Japan's relative share is declining.
The regional share is calculated by using the country of the company's headquarters. By this definition, Atmel, Cypress Semiconductor and Texas Instruments are classed as US companies. If we counted only where the chips were fabricated, the share held by Asia would be much larger.
Fabless vs integrated manufacture
An early trend we are following is how fabless manufacturers are losing share to competitors that fabricate their chips internally. Figure 2 shows this, with the share of CMOS fabless manufacturing falling between 2005 and 2006 compared with integrated CMOS manufacturing. Fabless production grew but not as fast as integrated production, so its relative share declined.
CCD manufacturing is another matter. Revenues from CCDs have grown, but more modestly than CMOS arrays. In turn, the relative share of CCD production has declined, taking the share of the Japanese CCD manufacturers with it. Historically there has been little CCD foundry manufacturing, with most of the volume being in the hands of a few Japanese suppliers.
One interpretation is that this trend reflects a slight but real advantage in favour of the integrated manufacturers. For example, both the fabless designer and the foundry expect to take some profit margin. An integrated manufacturer can take the entire profit margin for itself, or alternatively, it can sacrifice more margin to compete on price and gain market share.
However, there is no convincing reason why an integrated strategy is fundamentally better than a fabless strategy – it all depends on communication and co-operation between the design team and the fab engineers. Even for integrated suppliers, politics and short-term priorities can overrule good engineering. Other important factors include the wafer sizes used in the fab, the skill of the design team and the strength of the patent portfolio. Either strategy can work well or poorly.
The fabless model has worked for US-based Omnivision, which has been fabless since its inception and operates through TSMC in Taiwan. Kodak has also gone fabless for its recent CMOS image sensor effort, using TSMC and IBM, but it remains integrated for CCD production.
Europe's e2v technologies acquired Atmel's image sensor and camera group, expanding its scope across CCD and CMOS. e2v's CMOS manufacturing is fabless and done through Tower Semiconductor, a foundry in Israel, however, its CCD production remains integrated, like Kodak's.
Fairchild Imaging is going a step further. It began with fabless production for its CMOS arrays in 2003, but is now making its CCD production fabless as well, working through the Supertex foundry.
ST made a particularly surprising move this year when it looked to foundries for its image sensor production. The company has been manufacturing its image sensor chips in 8-inch fabs in France, while performing module assembly in Europe and Asia. These fabs have been competitive because the equipment is depreciated, but the growing worldwide demand pushed ST to expand capacity. As a consequence, this summer, ST looked to TSMC and ProMOS in Taiwan (both have 12-inch fabs) to expand its image sensor production.
While Sony's sales continue to climb, in part due to consolidation of its share of CCD sales, it too is scrambling to bring its own CMOS image sensors into cameras.
Customers and volume are also important. Another interpretation is that the falling share of fabless production is simply an artifact of the changing fortunes of the customers: the good or bad fortunes of a major customer may tip the market share of the image sensor supplier along with it.
Now that mobile camera phones are such a big part of the image sensor market, the market share depends heavily on the fortunes of the handset manufacturers. Nokia is a key customer for ST and Motorola is a key customer for Micron. So, for example, when Motorola's share declines so does Micron's. Other leading suppliers are Omnivision in the US, Samsung and Magnachip in Korea, and Sony, Sharp and Matsushita in Japan.
In view of the importance of the handset manufacturers, small changes in the market share of image sensor suppliers are insignificant in the short-term.
Volume is another big factor, with the image sensors for consumer products tipping increasingly in favour of the largest suppliers. As those suppliers successfully ramp in volume, they can spread their costs over a larger number of chips, allowing them to cut their prices to the customers. Smaller volume suppliers will find it increasingly difficult to compete in those markets and the number of competitors will gradually decline.
The global nature of imaging
This evaluation of the image sensor market raises a perennial question: how is Europe placed in the market? It is holding its own with respect to some of the design and ST's strength in CMOS arrays. On the other hand, Europe's share of the market remains low. But is that a reason to be concerned?
In fact, the entire business is global and this has probably helped Europe to hold its position. It is safe to say that most imaging customers today are buying and selling worldwide, either directly or through distributors. This is true for applications as diverse as mobile camera phones, automobiles and digital radiography.
It is worth remembering that internal operators are global. For example, an image sensor company may have its design centre in Europe, its fab and module assembly in Asia, and the OEM customer in North America for distribution to an end-user possibly in South America or Australia.
A recent economic study shed some light on the paradox of globalization. Who Captures Value in a Global Innovation System? The Case of Apple's iPod was authored by Greg Linden, Kenneth Kraemer and Jason Dedrick from the Personal Computing Industry Center, University of California at Irvine, US (http://pcic.merage.uci.edu).
The authors used a teardown study from Portelligent to evaluate the contributions of value added to the Apple iPod. Apple is not unusual today in that it does not actually manufacture anything in the iPod itself – everything is purchased or outsourced to contractors. In other words, Apple only supplies creativity in the design and marketing of the product.
What is striking is that about 27% of the $300 value of an iPod is returned to Apple and its shareholders. China appears to add only about 1% of the value in labour. While the Apple iPod may be an exception in the electronics industry in that it reaps a high margin, it nonetheless serves as an important lesson: manufacturing can move to Asia and much of the value can still come home. Essentially, Apple is using resources to produce something of much greater value. Does it matter if 1% of the resources come from Chinese labour?
Europe is not a dominant player in image sensors and it appears that even more of the actual fabrication will be done in Asia than ever before. But if the iPod example is any guide, the value returned to Europe in direct costs and profit will still be important.
Even more important may be the close coupling of the image sensor design in Europe to European camera makers, especially in speciality industries such as machine vision, automobiles and radiography. That market is even larger, since it includes not only the image sensor but the added value of the optics and image processing, all adapted to the application. It is not easy to discern the value of having a close local coupling of the image sensor design to the camera manufacturers, or to their customers, but it is certainly something that should not be overlooked.
• This article originally appeared in the September 2007 issue of Optics & Laser Europe magazine.