The six billion dollar gap
02/01/2006
Economists have been delivering dire warnings lately that the semiconductor industry will soon no longer be able to continue making IC devices smaller, faster, and cheaper at the same pace it has maintained over the past 40 years. The problem, they explain, is not technological in nature, but rather the result of an inability to pay for the research needed to achieve the needed innovations. Now, a new survey of the leading executives in the industry supports this claim by quantifying the impending research gap and explores coping strategies that a growing number of companies are finding they have no choice but to adopt.
The report, “Semiconductor Equipment and Materials: Funding the Future,” - commissioned by Semi and based on in-depth interviews with executives from 45 companies in the US, Japan, and Europe - estimates that by 2010 equipment and materials suppliers, to whom the burden of research has been shifted from chipmakers, will be able to afford an annual R&D budget of about $10.4 billion. But it also forecasts that the research required for the continued scaling of IC devices, even without another wafer size increase, will cost some $16.2 billion annually by then.
One reason for this huge gap can be seen in the challenge of adopting new materials. Because current IC structures have been made virtually as small as possible using traditional methods, new materials with new electrical and physical properties are being explored to allow further scaling. After several decades with minimal changes to the material sets used in device manufacturing, scores of new materials will be needed for high-k and low-k dielectrics, wafer substrates, strain engineering, transistor gates, and so on.
To put the cost of such new materials in context, at the 90nm technology node, just four new materials were introduced. But at the 65nm node, a dozen new materials will be required. And at 45nm, more than 30 will likely be needed. This trend is significant considering that before the year 2000, just 16 different materials were used in semiconductor manufacturing. The bottom line is that the cost of testing these new materials and integrating them with existing processes - which will entail managing unexpected interactions and accounting for process variations among different customers - will quickly spin out of control.
The report concludes that among the most promising strategies for managing rising research costs is for companies to engage in pre-competitive partnerships, both with customers up and down the supply chain as well as with direct rivals. To deal with the number of new materials, for instance, companies must work together, the study contends, to narrow the number of new materials to a few choices, thus eliminating redundant and obsolete efforts, so they can then concentrate on developing competitive products.
How practical is this approach? Can rivals collaborate on creating the very technologies that can offer a competitive advantage? To a large degree, this is already happening, albeit not necessarily by choice. At least that was the message to emerge from representatives of the leading semiconductor manufacturers in a rare display of solidarity at a marathon after-hours panel discussion titled “R&D: Who will do it and who will pay?” at the recent IEDM conference in Washington, DC.
On the inevitable trend toward collaboration, Bernie Myerson, VP of strategic alliance and CTO at IBM, asserted that there has to be a change and that basic research will have to be performed in partnership. Otherwise, companies will collapse under their own weight trying to deal with the R&D funding problem. “The industry has had to grow up,” he noted, “like children who realized that if you don’t first figure out how to make the ball, it’s kind of hard to play ball.”
Supporting that sentiment, Paolo Gargini, Intel’s director of technology strategy, explained, “In any operation, most of the money is spent on trying to find out what to do next.” If you look at the top eight or so IC companies, he noted, you can see that by pooling resources in industry consortia, they were able to gain nearly an order of magnitude return on investment in basic research.
Of course, the challenge remains for individual companies to take the intellectual property shared among many and turn that into competitive products. The unanimous consensus, however, was that this is where the vast majority of R&D resources must be spent and where the competitive spirit can still thrive. It’s almost like a Tour de France bicycle race, one audience participant suggested, in that people cooperate and take turns leading until someone sees an opportunity to break away.
Jack Sun, senior director at TSMC, perhaps summed up best. “Those who want to be leaders will invest and share more, and those who choose not to, or cannot afford to, will have to drop out,” he said. “There’s controversy in terms of whether final development and productization should be done in an alliance or in house. I would just use the Chinese saying: ‘When someone drinks water, whether it’s warm or hot, only he or she can tell.’”
 |
Phil LoPiccolo
Editor-in-Chief