450mm Wafers!!

[imported_cinder]

Today, I had the pleasure of seeing a 450mm Wafer hands on. In fact, my professor informed us that we were the 1st college students to have that opportunity.

He works for Sematech in Austin, TX. He brought it in to show us an example of what he deals with at his facility. The class is facilities design at Texas State University for those curious.

My professor explained that the 450mm wafer was not grown, but it was made with powdered silicon. It is not the high quality of the grown silicon wafers. The wafer is not good enough for commercial use, but is used in the development and design of equipment to accommodate a wafer of such size. Once the wafers can be silicon grown, there will be even more need for wafers that size. It will benefit consumers and manufacturers (only large ones at first) to go to 450mm as you can fit more chips on one wafer saving machine times and much more, then the savings trickle down to the customer.

Anyways, I found this pretty exciting. I figured I would post my excitement hah!

 

[Idontcare]

I would be surprised if silicon substrates are the semiconductor material of choice for active elements by circa 2025 when 450mm substrates are being used commonplace.

I would also not be surprised if silicon is used as the underlying substrate material of choice for 450mm wafers while the active semiconductor material is deposited on top (say epitaxial SiGe).

After-all by the time 450mm fabs could become a reality the industry's leading edge would have scaled well beyond the 16nm node. We may be on buckytube transistors by then and there is no real requirement that such exotic transistors be manufactured on silicon.

EUV is pretty much going to do-in (consolidation, not bankruptcy) about 1/3 of the independent semi mfg's out there attempting to remain within 5 years of the leading edge. Hi-k/MG will snag another 1/4 by my estimates, let alone scaling to non-planar CMOS elements.

Economically it is pretty much accepted in the industry that 450mm will be out of reach for all but Samsung and Intel sized-businesses. So existing companies either need to massively increase their company size or they need to consolidate to become equivalent sized.

The economics are brutal and the across-wafer process variations all but kill the effort to yield more sellable product per wafer despite their larger size for the high-performance logic MPU guys (memory is less impacted by this due to industry standardized minimum speed-bins).

At Texas Instruments (3rd largest semi manufacturer behind Samsung and Intel) we built our first and only 300mm fab in 1996...didn't move equipment into it until 2000, started shipping wafers for revenue in 2001.

For the 7 years (2000-2007, KFAB shut down in Dec'07) that we had both 200mm KFAB and 300mm DM6 operating our top-speed bin chips (by volume) for SUN microprocessors always came from the 200mm wafers...the within-wafer process variation was just such a killer at 300mm.

So logic guys will need massive resources to actually optimize their processes to extract the promised benefits of 450mm...and at this time there are only 2 companies even close in revenue to supporting such efforts.

Now the memory guys, your Toshiba's and Micron's...they would be the volume play for 450mm substrates but their margins are so razor thin that the only way they could afford the transition is if there was a pre-existing market and volume production of the required toolsets had reduced equipment prices substantially.

Micron is not about to foot the bill for the world's first 450mm fab. They would probably get into it after say 10 or 15 450mm fabs had already been built worldwide.

TSMC might be big enough in 2025 that they may lead the charge into 450mm fabs. The 300mm fab industry was created off the lucrative profit margins of the logic producers, once the volume was there then the memory guys got to move in and take advantage.

Only Intel has the money to make it happen at 450mm from this standpoint, and even then folks have to ask whether the effort exceeds the diminishing returns. It took nearly a decade for 200mm->300mm to payoff industry-wide. What will the semiconductor industry even look like in 2030 when 450mm reaches that point?

Progress is exciting, 450mm is inevitable if CMOS is still around in 15 years. I'm not betting on it though. 450mm fabs could be the Iridium satellite network of the industry.