SAN JOSE, Calif. – Despite a sudden surge in fab tool activity for the next-generation wafer size, 450-mm fabs will emerge rather later than sooner, and the production target appears to have slipped.
This comes to no surprise to analysts, as the recent recession appears to have delayed 450-mm fab and tool investments. As reported, Intel, Samsung and TSMC have talked about having 450-mm prototype fabs by 2012. Intel Corp. is readying D1X, a 450-mm ''capable’’ fab. The new development fab in Oregon is slated for R&D startup in 2013.
Before the recent recession hit the industry, 450-mm production was slated for 2015 or 2016 at the earliest, said Trevor Yancey, an analyst with IC Insights Inc., at the firm's recent forecast event.
The recession caused fab tool vendors to de-accelerate their investments and programs in 450-mm. As a result, 450-mm production has been delayed at least to 2017 or 2018, Yancey said.
''While 450-mm would significantly lower the cost per unit and increase the output of a fab, the equipment suppliers are not making the investments needed to move to larger diameter wafers,’’ said Gus Richard, an analyst with Piper Jaffray, in a recent report.
''The equipment suppliers did not realize an acceptable ROI on their R&D investment in 300-mm tools and they are hesitant to fund development of 450-mm wafer capable tools,’’ he said. ‘’While we believe that 450-mm will eventually move to production, this is unlikely to occur in the next five years.’’
Still, fab tool vendors are quietly ramping up their 450-mm efforts. ''There’s a lot of activity going on in the back channel around 450-mm to indicate that some big news will break next year,'' VLSI Research CEO G. Dan Hutcheson, noted in a recent report.
''The equipment suppliers have stopped resisting it and most have some level of effort underway,’’ he wrote. ‘’Moreover, those that don’t are no longer being painted as defiant realists. The thinking has moved from ‘Over my dead body’ or ‘I’ll retire before 450’ to either ‘We’ll have to do it if our competitors move’ or ‘You’re falling behind if you’re not doing it.’ This is a big change.''
Progress is also be reported by wafer makers. German wafer specialist Wacker and others say they have developed a new technique called "Planetary Pad Grinding," which combines grinding and lapping methods to produce extremely flat "jumbo wafers."
The transition to 450 mm has been a tough sell for sure, but I can foresee a time where once a couple of early adopters such as Intel and Samsung make the move to 450 mm other companies will be forced to follow or they will no longer be competitive - especially the commodity memory producers. At that point we may see more consolidation and joint ventures as smaller players struggle to survive.
I agree with Dave and @docdivakar and think this transition will continue to get pushed out...who can really afford it? Its a huge investment requiring new tools and new fabs. Intel and Samsung for sure but I doubt there will be many more takers...there is much more excitement in semi industry in growing vertically (3-D) than horizontally! I bet in the future there will be more silicon cubes than large silicon pizzas! Kris
This is not too surprising given the state of the world's economies during the past couple of years. If you've not yet been able to get adequate returns on the current state of the art, you really aren't very willing to move on to the next, especially if you're still able to make money at the current level.
I suspect we'll see a few large players making the move because they have seen the returns and can afford the investment and some slightly smaller players holding back and hoping to leap-frog to the next level when it becomes available.
The analyst from Piper Jaffray makes a valid point: until the equipment makers (I would include backend and test equipment as well) see value-adding opportunity, the 450mm launch will keep getting pushed out. Chances are, by 2015 and beyond, there will be fewer fabs that can afford the $Billions needed to upgrade.
This is definitely a quandary for the 300mm fabs; on one hand, we see fewer fabless companies with products off the drawing board (we hear any where from $5M to $50M for a new ASIC development), more consolidation that seems to be coming in fabless companies, while on the other, emerging 3D chips may still use 200mm wafers, or worse, put more pricing pressure on the 300mm fabs.
Blog Doing Math in FPGAs Tom Burke 2 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...