Major Structural Changes in Store for the Industry
by Debra Vogler, Instant Insight Inc., Sunnyvale, CA
The semiconductor industry has to move EUVL into volume production, develop new transistor architectures and new channel materials (as a minimum), and transition to 450mm – all within the next several years. While these activities are undertaken, industry consolidation will continue, and some companies may need to retool their business models. Bill McClean, president, IC Insights, noted that Intel, Samsung, and TSMC will probably be able to manage funding at the levels the companies believe to be necessary. Mid-range companies such as GlobalFoundries, Hynix, UMC, and Micron, however, will be challenged, and other companies may choose to go fab-lite (Fig. 1).
Figure 1. Transition of logic production to IC foundries. SOURCE: IC Insights
Intel is able to spend over $20B total each year on R&D and capex combined, observed McClean. “That’s a huge amount of resources and research that can be called upon to make those big leaps,” said McClean. “And there isn’t anyone at this time that can stick with that type of investment.” Though AMD will feel the pinch if Intel’s process gets too far ahead, McClean noted that Intel’s impact on the rest of the industry will not be too great because they are not huge competitors in other areas such as NAND Flash, or DRAM, nor is the company the leading foundry. Therefore, until Intel’s major product is not a microprocessor, McClean believes that it’s alright for the rest of the industry to make progress and to invest at a different rate. “Intel is unique, and sometimes it’s a mistake to judge the rest of the industry by what Intel is doing,” said McClean.
Turning attention to another industry heavyweight, McClean observed that Samsung is playing out the vertically integrated business model to perfection. A key factor to ensuring success with the vertically integrated model is for the people at the top to be behind it. The top executives have to be willing to invest during industry downturns so that when the business cycle goes up again, the company will have the market share, the experience, and new fabs ready to go. Without such support from the top, the model falls apart completely, explained McClean.
Wafer size transitions will impact both digital and analog
Wafer size transitions are not the sole province of the top 10 or so IC manufacturers and foundries that are expected to be able to move to 450mm manufacturing. Analog may be a different world, but McClean noted that even those device manufacturers are getting pressured to move from 200mm wafers to 300mm. “Almost all analog chips are made on 200mm wafers,” said McClean. But TI forced the issue of going to 300mm when it purchased Qimonda’s 300mm tools for very little money when that DRAM fab went out of business. As more companies feel the pressure to make the move to the next wafer size, business models are challenged, forcing adjustments (the familiar chorus of going fab-lite, or maybe even going out of business).
When asked to comment on the alphabet soup of 450mm consortia and whether it might be better to pool resources, McClean said that participants and different regions have their own agendas. For example, unless Intel builds a 450mm fab in Ireland, European equipment and materials suppliers will have no other indigenous European IC manufacturer at 450mm. What these suppliers want to avoid is being shut out of leading-edge production. “If you look 10 years down the road, almost all leading-edge digital will be on 450mm wafers,” said McClean. “If you’re not a supplier to that, you’re out.”
And speaking of being out of business: McClean observed that there are about 70 companies doing 200mm production, 30 companies doing 300mm, and fewer than 10 expected to do 450mm. “If that isn’t a consolidation, I don’t know what is,” he said. “Major structural changes are coming, and we’re right in the midst of it.”