The 450mm transition: Many unanswered questions
The move to a larger wafer size -- from 300mm to 450mm ??? has been the focus of an interesting debate for the last five years. The largest semiconductor manufacturing companies in the world ??? Intel, TSMC and Samsung ??? have been strongly advocating the move, urging equipment suppliers to start development (which has started) and SEMI to create the necessary standards (which is done). Equipment suppliers have been understandably slow to embrace the change, given the long time for them to see the return on their investment during the last wafer size transition (200mm to 300mm).
The move to 450mm took on some new urgency at this year's Semicon West. Applied Materials announced that it would spend $100 million on developing 450mm tools (which some say is just the cost of doing feasibility studies), companies such as EVG and KLA-Tencor introduced new 450mm products, and several panel sessions were devoted to the topic.
Many questions remain unanswered. Although most people seem to agree that it's no longer a question of "if" but only "when," there's still a long way to go.
Speaking on one of the panels was Gartner's Bob Johnson, research vice president, semiconductor manufacturing, who noted that the move is purely a cost play, with a target goal of a 30% reduction in production costs. The problem, said Johnson, is that there is a fundamental conflict of interest between semiconductor and equipment manufacturers. "The way it looks right now is we've got the opportunity to spend billions of dollars on R&D that could go for other projects, and end up cutting the market (for equipment) by about 30% going forward. This is not a heck of a lot of incentive to put a lot of the R&D money out there."
Johnson also questioned the real cost savings that have been gained by previous wafer size transitions. "If you look at the 30 year period from 1980 to 2010, we had 14 major technology node changes, each one generating, over roughly a two-year period, about a 50% reduction in cost. That's 30% per year," he said. During that time, there were three major wafer size transitions. "Each wafer transition was about the equivalent of one year's worth of scaling. Wafer size changes really didn't have a lot with reduction of costs. What happened instead was shrinks, scaling, cleverness in making smaller cell sizes and a lot of the other things that process engineers do to keep driving things forward," he said.
Presently, although some tools have already been introduced, 450mm is still at the feasibility stage. "We don't really know if we can do it or not," Johnson said. "Once we start putting pilot tools in, the costs start going through the roof. This is fairly standard. Feasibility is relatively cheap. Hopefully, we can decide one way or the other whether this will be worth it before we start spending cubic gigabucks on R&D."
Another unanswered question is how many technology nodes will be left by the time 450mm wafers are in production. Intel is starting 22nm in production at the end of this year or early 2012. That means that they'll be at 7nm right after the first 450mm production fab is slated to be in operation. "We hear people talking, ???We can do 32nm, we can do 22nm with some of these tools.' So what? When 450mm comes into production, we're going to be at 14nm or 10nm or below technologies. That means we need EUV, immersion for mix-and-match, and probably dry litho all to be economically viable," Johnson said. "The next question we have is how many nodes do we have? How long is conventional silicon going to keep going after we pass 10nm level, and I don't think anybody knows the answer to that."
Johnson provided an interesting look at ROI considerations, comparing the market and thinking during the transition to 300mm in 1997, to the way it is today. "In 1997, we were convinced, looking in our rearview mirror, that the industry growth that was going on at 17% per year was going to continue forever. We didn't realize that we already passed the inflection point two years earlier, and we weren't going to realize that for another for our five years. Now we know that the long term growth trend in the industry is mid single digits, somewhere around 6%/year. If you look at the ability of the industry to generate funding for fabs and everything else, it's much less than it was back in 300mm days." We could reasonably look at about ten nodes over the product life of 300 mm tools. Now if we look forward, maybe 3? Maybe one? I submit if 450mm is really going to end up being a single node product, it ain't worth it." Truer words have never been spoken.
Solid State Technology | Volume 54 | Issue 8 | August/September 2011