Heard at The ConFab

08/01/2009

For the past five years, PennWell has sponsored an event called The ConFab, which is a conference focused on semiconductor manufacturing issues, combined with a variety of focused “networking” opportunities for sponsors and delegates. It’s held just outside Las Vegas at the serene and beautiful Loews Lake Las Vegas (there are shuttles to “the strip” but most attendees prefer to stay and relax).

Planning for this year’s conference (held June 15-17) started in earnest last year in the late fall, with event director Jay Novack leading weekly conference calls with me and our topnotch advisory board: David Bennet (Global Foundires); Janice Golda (Intel); Bill Rozich (Albany/IBM); Sima Salamata (TI); Kevin Logue (TI); Sitaram Arkalgud (SEMATECH); Hans Stork (AMAT); Paul Edstrom (GE); Jean LeMoin (MCA); and independent consultants Bill Tobey, Ken Rygler, and Takeshi Hattori.

These were the times when it felt that the economy was at the edge of an abyss and our discussions often turned to “will this all be over in June or should it be the overarching theme of the conference?” The semiconductor industry is still climbing out of that deep dark hole, but it now seems safe to say the worst is over. In the end, the conference presented a nice balance of economic data and forecasts with meaty technical topics, such as EUV vs nanoimprint lithography, the cost of test, 3D integration, as well as practical discussions of how to extend fab longevity. For the first time, we also covered the convergence of the semiconductor and solar industries and how photovoltaics technology was evolving.

Keynote speaker Brian Krzanich, Intel’s VP and GM of manufacturing and supply chain, gave everyone a smile to start, sharing research that shows that people are willing to give up about anything before they would give up broadband Internet, including toiletries, clothing and even food. It doesn’t say much about society, but it does mean the Internet and its required infrastructure are here to stay.

Bill McClean of IC Insights further amplified that the demand for semiconductors remains strong, noting that a quarter billion cell phones were shipped in the second quarter. “From the industry’s tone, you would have thought it fell to zero,” he said. He noted that the industry had not added capacity to keep up with underlying demand so he is predicting “boom” years of double digital growth in 2010 and 2011. Historically, that has always been the case. Every recession in the past, without fail, has been followed by strong semiconductor growth.

George Scalise, president of the Semiconductor Industry Association shared similar views on the industry’s long term prospects, noting that semiconductors are the 2nd largest export from the U.S. (behind only oil refinery products). Scalise noted that, according to a recent study, semiconductors are now the driving force behind U.S. energy efficiency gains (www.aceee.org/press/e094pr.htm). The study reports that the need for 184 electric power plants has already eliminated due to more efficient semiconductor technologies, and that $1.3 trillion in reduced electric bills is seen over the next 20 years.

The great debate at the conference was over what lithography approach will supersede the existing approach (ArF immersion with double patterning). The two technologies presented as viable approaches: EUV lithography and nanoimprint. EUV came out as the clear winner with good progress reported by Nikon and Toppan. However, it was also clear that massive amount of investments and technical advances were still required for either to be truly workable in high volume manufacturing. What’s missing is advanced mask inspection tools for EUV and fast e-beam writing tools for NIL: both will require millions in investment, yet little work has been done.

The cost of test panel carried something of a surprise. Gone are the days of million dollar test systems. Today’s systems are more in the $200K range, thanks to design for test (DFT) and built-in self-test (BIST) strategies. The real test challenge may lie in 3D integration: there is simply no way to access all those little vias post-stacking to make sure they’re good. Could be a show stopper. Solar? Dan Hutcheson showed a chart where a tiny area in the middle of Africa could power the entire world. It’s definitely the future of energy.