Whatever Happened to 450mm Wafers?

When I started at VLSI Technology in the early 1980s, our fab used 5" wafers (150mm, also called 6" sometimes). At some point, VLSI converted to 8" (200mm) wafers. This is not easy to do without taking the fab down completely since you can only introduce new equipment one machine at a time, and you can't use any of the 8" equipment until at least one of everything required for the process is in place. I think you can run 5" masks and wafers on 8" equipment, which simplifies the transition a little.

Since about 2002 or so, most new fabs have used 12" wafers (300mm). The criterion for whether the investment in 300mm wafers is justified is not the process generation but the volume. For example, Bosch's Dresden fab is 300mm even though it will be mainly serving the automotive industry, where almost all chips are in what are somewhat disparagingly usually called legacy nodes.

The next change of wafer size was meant to be 450mm or 18". It was meant to have happened by now, so why didn't it?

Here's a picture of me at SEMICON West in 2015 holding an 18" wafer. They already existed seven years ago. But the big question is whether I will ever hold another.

In my post the following year, 200mm Fabs Awaken, I wrote a summary of the history of wafer sizes.

Modern fabs use 300mm (12") wafers. Older fabs have used 200mm (8") wafers since the early '90s, starting with IBM Fishkill's pilot with Siemens to build 64Mb DRAM in 1990. Before that, we had 5" wafers (150mm), and if you go back far enough in history, you'll find 4", 2", and even 1" when integrated circuits started. 300mm started in 2000 with Siemens/Motorola (and two beta lines in the US at l300l and Japan at Selete).

For several years there has been a program called G450C, working on the feasibility of 450mm (18") wafers. There seems to be a full equipment flow for this, but there is no enthusiasm on the part of the IDMs and foundries to make the massive investment necessary to build a 450mm fab. So it looks like 300mm will be the standard for some time. The big issue is that the equipment industry is unclear how they will ever recover their investment if the transition takes place since it takes many process generations. Fabs for 10nm are being or have already been built, so the earliest process that could use 450mm would be 7nm. An old SEMATECH study from 1997 showed that a wafer size remains in production for approximately 24 years so that everyone can recover their investment. That would take us to 2040 or beyond, way past any visibility we have on semiconductor processes.

I mentioned G450C in that quote from my old post, full name Global 450mm Consortium. This was created in 2012 as a consortium of New York State's College of Nanoscale Science and Engineering (CNSE), Intel, TSMC, Samsung, IBM, GlobalFounries, and Nikon. Nikon is the only semiconductor equipment vendor on that list. In general, the semiconductor manufacturers want 450mm wafers since it has the potential to reduce their costs by perhaps, 30% per die. The equipment manufacturers are much less keen since they will need to put a huge investment in up-front R&D that they will only recover over time if they sell enough tools. That, in turn, requires enough 450mm fabs to be constructed. Of course, there are other funding models. When ASML ran into financial difficulties over EUV scanner development, the semiconductor manufacturers who most required the technology made massive equity investments into ASML. G450C was rolled up at the end of its 5-year initial program when some of the members of the consortium pulled out. As far as I know, there is no R&D going on today towards producing tools for 450mm wafers.

My go-to person for anything to do with fabs is Scotten Jones, my old colleague from Wikipedia days. His company, IC Knowledge, creates cost models for fabs. His cost models showed a 20-25% saving from 450mm. In a post several months ago, The Lost Opportunity for 450mm, Scotten gave his version of the end of 450mm:

Unfortunately, the efforts to develop 450mm have ended, and the only 450mm wafer fab has been decommissioned. The 450mm effort was different than past wafer size conversions. At 150mm, Intel was the company that led the transition and paid for a lot of the work, and at 200mm, it was IBM. At 300mm, a lot of the cost was pushed onto the equipment companies, and they were left with a long time to recover their investments. At 450mm, once again, the costs were being pushed onto the equipment companies, and they were very reluctant to accept this situation. In 2014 Intel (one of the main drivers of 450mm) had low utilization rates and an empty fab 42 shell, and they pulled their resources off 450mm, TSMC backed off, equipment companies put their development efforts on hold, and 450mm died.

I mentioned how switching from 5" to 8" wafers are VLSI was complicated since you need to switch enough tools to get a whole manufacturing flow. Well, switching an industry from 300mm to 450mm is like that too. It is no good if only, say, the lithography vendors switch. You also need plasma etchers, CMP machines, diffusion furnaces, metrology, and more. Every piece of equipment needs to be available in 450mm form before a manufacturer can build the first 450mm fab. Basically, it's all or nothing.

And for now, and perhaps forever, it seems to be nothing.

The 300mm Transition

I wrote this post before I went away to UK for a week. The very day I was leaving, Asianometry published a video on the last transition, from 200mm to 300mm (with a passing reference to 450mm):

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

.