Is the pure-play foundry model broken?
EBN recently explored this issue in light of the business conditions that the top three silicon contract manufacturers are battling, as well as claims by analysts and a few top independent device manufacturers (IDMs) that the cost of doing business is undermining the ability of foundries to roll out new technology for their customers.
At first glance, the obstacles are daunting. Slack capacity utilization is challenging the ability of foundries to maintain their capital expenditure schedules, while at the same time they are under growing pressure to field
complex new process technologies that both generate a return on investment and serve the needs of a diverse fabless client base.
Recent disclosures that high-profile customers like Nvidia and Xilinx have transferred production of their latest designs from Taiwan Semiconductor Manufacturing to IBM would seem to support the concerns raised in the EBN article.
For its part, TSMC defends its position, stating that a well-publicized delay in ramping up its 90nm production process was an economic decision and not based on any technological shortcomings.
Let's face it, pure-play foundries will face a hard road when it comes to advancing their process geomteries if they can't keep their fab lines well stocked. The shift to 0.13-micron and smaller linewidths is presenting a host of new technology hurdles and contributing to the inexorable rise in the cost of mask sets. The transition to 300mm wafers carries a separate set of problems, given that the sheer size of the platters necessitates fabless clients commit to fairly large chip volumes.
However, none of these issues is really unique to the foundry model. Scaring up demand in a 300mm-wafer world, fab construction costs, and the increasing complexity of new production techniques are challenges shared by the entire industry--including any number of top-tier IDMs. The foundry model has served a large and growing constituency for at least a decade and to this point represents the only viable outlet for the majority of companies developing semiconductor components, particularly start-ups. Moreover, IDMs like Motorola and Agere have come to embrace the outsourcing concept as a means to defray capital expenses.
Only time will tell if the problems that foundries like TSMC have encountered are purely business related or if there are deeper technology issues that will allow IBM and other IDMs to skim the cream off the fabless milk pail. But given the winning track record of the foundry model to date and the fact that a universally bad economic climate is roiling the entire semiconductor market, any effort to read dire consequences into the foundry sector's travails is premature.
E-mail comments to Andrew MacLellan at firstname.lastname@example.org.