Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Manufacturing/Packaging

Commentary: A few fabless fables

Posted: 28 Dec 2007 ?? ?Print Version ?Bookmark and Share

Keywords:fabless fables? commentary? fabless supply-chain?

By Shankar Pennathur
Oxford Semiconductor

The semiconductor industry is commoditizing with opportunities to differentiate being few and far between. As the fabless companies jostle and compete for market positioning and market share, they are subject to the larger trends that transact in the fabless supply-chain world.

As a particularly competitive year ends and a new one is on the horizon, this author would like to reflect on a few trends in the past decade, and ponder about the lessons one can learn from such trends. As you read these fabless fables, any imagined similarities you may draw between real companies and the references herein, are well, a figment of your imagination.

Foundry competitive landscape is changing: The foundry industry was getting very boring and uncompetitive, essentially being dominated by a couple of players from Taiwan early this century. However, things got very interesting when an emerging China got involved in the mix. A popular up-start and some smaller start-ups entered the foundry industry to make things very exciting for the fabless customers.

Taiwan would not let cutting-edge technology transfer to China and the U.S. Congress would not approve sale of some process equipment to China. Despite all this, economics found a way around to render a semblance of early viability to these new ventures. The main strategy chosen by these new players from China turned out to be competing on price.

After lawsuits, settlements and more lawsuits, the foundry landscape has been changed over time in a way that has caused foundry pricing to come down significantly (good for consumers), but where winners have remained winners and newcomers have not made a dent in either the profitability or the market share of the incumbent market leaders.

Moral: Never compete on price with competitors with deeper pockets.

Where do I invest my manufacturing dollars? (Wafer fab or assembly-test plant?): The 1990s were a story of profitable wafer foundries and perennially struggling assembly, test and packaging (ATP) vendors. Fabless customers were loyal to foundries and yet were changing ATP vendors constantly. A couple of things changed at the turn of the millennium, however, that completely reversed roles.

First, the communication bubble burst. Semiconductor demand was subsequently driven predominantly by consumer applications. Cost of silicon came under attack and came down significantly as a result. The entry of new foundries and aggressive process migration to finer geometries helped to fuel the trend. Cost of ATP services as a percentage of total IC cost increased to a level where silicon was not necessarily the dominant cost component anymore for many applications.

Second, borrowing a page from the foundries, ATP vendors began managing capital spending very closely. They also began to bring supply down aggressively relative to demand to shift the economics more in their favor. A substrate factory going up in flames here, or a couple of substrate factories going out of business there, certainly increased the supply-demand imbalance.

Today, very few wafer foundries are making money (top 2-to-3 maybe), and yet a lot more ATP vendors are profitable (at least the top 4-to-5). The ATP vendors seem to be sufficiently attractive financially, even the private equity community is taking notice and taking action in some cases.

Moral: Old dogs can learn new tricks.

IP business: The viability of the semiconductor IP business is a popular topic for debate. Except for a few dominant players (mostly processor IP providers), it is rare to find profitable and scalable IP businesses. The key to viability seems to be a royalty-bearing IP revenue base that can grow significantly from a business model perspective, with some degree of IP customizability from a technical perspective.

Except for processor IP, most other types of IP do not survive these business and technical 'filters.' For example, the free or paid standard cell IP business is already a commodity business that is increasingly under threat from foundry-provided libraries that are likely closer to silicon-real than any vendor-provided IP. The PHY IP businesses generally start out profitably, but over time and given their standards-based nature, morph into commodities (quickly for certain IP, somewhat more gradually for others).

The lack of industry-accepted IP quality standards in general is an issue, although incentives are frequently misaligned between customers (who want standardization so they can drive costs lower) and IP providers (who often want to compete on quality and differentiation).

Further, the lack of willingness on the part of customers to pay large (if any) IP royalties (especially for consumer applications) and the constant lowering of the price/performance barrier due to the entry of lower-priced IP providers also play some role in the general unattractiveness of this business. It is not a surprise that this landscape has seen a lot of consolidation already with surely more to come.

Moral: Business model innovation is at least as important as technical innovation.

Innovation gap: The silicon creation or electronic design automation (EDA) houses and fabrication (foundries) portions of the supply chain constantly innovate. A lot of process technology innovations such as tricks to push the lithography limits, DFM/DFY, and novel transistor engineering including innovation in newer materials are being pursued in the relentless push to keep extending Moore's law.

This focus, aided by literally billions of dollars spent in R&D, characterizes the foundry side of the innovation equation. The EDA vendors typically play catch-up to process innovations, and try very hard to cope with the constant curve-balls thrown at them from the silicon manufacturing side.

Yet, the EDA vendors manage to develop and market the best tools enabling the creation of designs that can model and predict silicon performance accurately. The flip side of this is the story, however, is with the ATP vendors (as a group), where on a relative basis, they have not innovated a whole lot.

QFPs and BGAs have been around for quite some time. Once in a while, comes along a surprise like the flip-chip or a QFN, however for the most part, things have been rather innovation-challenged or devoid of any serious creativity of any sort on the back-end side of the business.

Is it really a surprise that one has to wait 4-to-6 weeks to get a 4-layer package substrate, when it only takes a little over a month to get a silicon wafer with 30-odd mask layers processed? Can this imbalance in innovation continue? Or will eventually the "Innovate or Perish" slogan catch-up with the back-end IC manufacturing providers?

Moral: Sometimes the tortoise really does beat the hare.

A fabled conclusion

The increasingly consumer-driven semiconductor industry is characterized by its dynamic and ever-changing supply chain landscape. It is very important for the industry participants to step back and notice the gradual shifts and learn. There are business model challenges that need to be addressed, innovation gaps to be filled, competitive landscapes to be improved, all leading perhaps, to continued industry consolidation. A competitive, yet stable, supply chain is critical to the industry's graceful aging as its maturation process continues.

One last moral: Study history to avoid mistakes of the past.

About the author
Shankar Pennathur
is VP of Operations and Manufacturing at Oxford Semiconductor Inc.

Article Comments - Commentary: A few fabless fables
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top