300 mm is Not a Blanket Solution
Laura Peters, Senior Editor -- Semiconductor International, 5/1/2003
As high-yielding and productive 300 mm fabs begin to materialize, there is still some controversy surrounding the industry's most difficult wafer size change in history. Companies that pioneered 300 mm manufacturing, including Intel (see "2003 Top Fabs"), Infineon, Samsung, TSMC and Texas Instruments, have demonstrated that a 30% cost reduction or better is possible using these large wafers. In a recent report by iSuppli (see "The Push-Pull of 300 mm Continues"), however, it looks like the full-scale production at these facilities will not be needed until mid-2004. As we well know, this date can come earlier or later than forecasted, a trend that the industry will inevitably continue to struggle with.
Another trend the report addresses is the fact that not all companies need 300 mm at this time, and perhaps will never require it. The issue has to do with needing some minimal number of wafers to justify the increasingly expensive mask sets and 300 mm wafer runs. So, as of right now, 300 mm probably will not be a total solution that can be applied across the entire industry. Just as there are still profitable 150 mm fabs, 200 mm fabs have long lives ahead of them.
One of the issues iSuppli's report does not address, and is certainly a concern with 300 mm, is the differences in yield and productivity that exist from one device manufacturer to the next. Some companies will do a much better job at making 300 mm a very profitable venture, while some of the trailing companies will have a more difficult time making 300 mm pay. Companies will suffer much more from inefficiencies in a 300 mm fab relative to a 200 mm fab.
It is quite likely, too, that the 300 mm ramp-up will affect the semiconductor economy, by swamping the industry and shortening the term of the upswing. China's aggressiveness in building fabs will also contribute to a fast upswing, even if much of this capacity is not at the leading edge.
Jim Morgan, CEO of Applied Materials (Santa Clara, Calif.), recently commented that the semiconductor business is changing and Applied is changing with it, as the company announced its restructuring plan last month. Morgan spoke of a "new reality" and the difficulties associated with turning a profit in the current economic climate. Though semiconductor equipment and materials suppliers are hit hardest, profitability problems are evident across the entire electronics sector. The pressure on companies to reduce costs while maintaining substantial R&D investment through this prolonged downturn has caused enormous layoffs across the industry; many of these workers will be hard to re-attract to the industry once the upturn does arrive.
Given all this, leaders in semiconductor manufacturing will not add 300 mm capacity until it is absolutely needed. Outsourcing to foundries will continue to be attractive. But will there be a mass exodus by integrated device manufacturers (IDMs) to abandon manufacturing and rely exclusively on foundries? For some IDMs with less profitable fab operations, this change certainly seems likely. Maybe the answer to this question lies in the precise assessment of how damaging idle capacity is to companies and how long cash reserves can be maintained, especially during long downturns, the next of which companies have no doubt already started planning for.
One thing is for sure: A 300 mm fab is an enormous endeavor and only the best-performing companies can assume the risk. The builders of the 200 mm fabs probably could not have envisioned all of the problems that were encountered at 300 mm. Nor would they have envisioned the extent of this downturn. But builders of fabs in the 1980s who paid a couple hundred thousand dollars for a stepper also did not think that million-dollar steppers would even be cost-effective. Today's researchers are developing $20M EUV tools. 300 mm is a reality, but only for the manufacturing elite.
