ChipWatch: Will The Industry Shift To 450mm?
By Matt Grimshaw
This article all started earlier in the week; news reached my eyes that France & Germany were "officially out of recession", followed by Hong Kong today. Then closer to my own world the first analyst has broken ranks and claimed that the recession was over for the battered & beleaguered Semiconductor Tool & Equipment vendors, although it may be a little too early to cry Hallelujah and cease ritualistically sacrificing of 300mm test wafers to the great god of productivity, there is light at the end of the tunnel. Downturns are never nice, but this one has been the ugliest I have personally experienced, with hundreds of my friends in the chip sector either floating across to the Photovoltaic industry or becoming 'consultants'. Some have even gone running to the hills to begin lives as professional hermits.
However, downturns do serve a purpose and although painful they do create the space for those companies and technologies that survive to grow and so they form an essential part of the Semiconductor industry's continuing cycles. This fact got me thinking about the challenges the next cycle may bring, which will be along far faster then we'd all like. The chip industry is inexorably entering the end game of scaling (Dr. Moore's Law), and there are a number of challenges that divide the industry like an Iron Curtain from the days of the Cold War - everyone in the Chip industry knows what they are, indeed it's well trodden ground, but I feel their fates will be determined by the fall out of this downturn and how bad the next one is... and it's an interesting, if slightly controversial, topic genre within which to delve.
The Move to 450mm.
If you even mention 450mm in a room of chip & tool vendor execs, sporadic fighting breaks out followed by the parties having to be pulled apart whilst both groups continue shouting & finger pointing about who paid for the previous wafer size switches. The logical argument for increasing wafer sizes is simple; in order to keep profit margins in line with the cost per function equations (part of the aforementioned law), larger wafer sizes are needed in order to reduce both the time it takes to make chips (bigger wafers equals more chips per wafer) and to make the individual chips cheaper to produce (more chips per wafer reduces the cost per device due to numerous factors relating to material and manufacturing costs).
The trouble is that transitioning to 450mm means all the vendors that want to play have to produce bigger versions of their tools and deal with the arising mechanical issues like dishing & bowing of the wafers themselves as they are so thin. This will cost anywhere from hundreds of billions to over a trillion dollars, and no one wants to pay for it... This is in no small part due to what happened in the transition from 200mm to 300mm in the late 1990's. The manufacturers wanted 300mm equipment ready for the 180nm technology cycle, the tool vendors obliged at great cost, then when the ramp up was supposed to happen a down-cycle hit and manufacturers cancelled their orders leaving the tool vendors with a whole heap of shiny, expensive but ultimately useless equipment that they'd developed.
Understandably this did not exactly endear the manufacturers to their supplier base who had to absorb the losses and start again for the 130nm technology cycle which did eventually see the 300mm switch. Now the tool vendors are being asked to develop 450mm tools and they have feet colder then the Arctic Circle for fear of the same thing happening again, needless to say they'll not be doing anything without cash upfront as this time the risks are also much, much higher. If a similar situation to the 300mm false start happened at 450mm, no tool vendor would be safe. Even the mighty giants like AMAT, ASML, TEL, KLA-Tencor, Lam and Nikon could get crushed under the weight of losses so large they'd make Everest look like an overachieving molehill.
To exacerbate this list of problems there really only seem to be three companies in the world that can afford Fabrication plants that are in the range of $10bn a shot without feeding their respective Boards of Directors & shareholders a steady diet of Prozac & Xanex in order to counter the fear of killing their companies through crippling capital investment....Of the three; one's the biggest Processor manufacturer, the second is the biggest DRAM producer and the last is the biggest Foundry: All are blindingly obvious so I'll mention no names. Everyone else seems to be happy with 200mm (still accounts for 80+% of total industry manufacturing capacity) and 300mm factories.
You see the problem is that it takes about 2-3 months of processing to make a chip from start to finish, and if there's a sudden downturn after chips have been started you may as well finish them because it'd cost too much to just trash the materials. 450mm wafers are 2.25 times bigger in surface area than 300mm wafers so that means that after you've hit the 'off' switch, the fab will pump out 2.25 times the devices (before an analyst slaps my wrist I'm speaking in rough terms here). That's 2.25 times the 'ouch' factor when the proverbial excrement hits the fan and 2.25 times the chips sitting around depreciating in value like an SUV in a fuel crisis.
So the facts are; no one wants to pay for it, it's incredibly risky and only a few stand to make any benefit. Does this mean I think it'll not happen? Well I am not a gambling man so I'd never say never...but I do question how many 450mm fabs the world's demand could sustain.
[Matt Grimshaw is the Editorial Director of the Semiconductor Technical Journal; Future Fab International (www.future-fab.com).]