Number of 300mm IC Wafer Fabs Expected to Reach 121 in 2019

[Daniel Nenni]

Nine new 300mm wafer fabs scheduled to open in 2019, five of them in China.

IC Insights recently released its Global Wafer Capacity 2019-2023 report that provides in-depth analyses and forecasts of IC industry capacity by wafer size, by process geometry, by region, and by product type through 2023. The newest edition of the Global Wafer Capacity report shows that 300mm wafers took over as the industry’s primary wafer size in terms of total surface area used in 2008. Furthermore, the number of 300mm wafer fabrication facilities in operation continues to increase. With nine new 300mm wafer fabs scheduled to open in 2019, the worldwide number of operational 300mm wafer fabs is expected to climb to 121 this year (Figure 1) and grow to a total of 138 fabs at the end of the forecast period.

Figure 1​

Some highlights regarding 300mm wafer fabs are shown below.

• At the end of 2018, there were 112 production-class IC fabs utilizing 300mm wafers (there are R&D fabs and several high-volume fabs around the globe that make “non-IC” products using 300mm wafers, but these are not included in the count).

• There are nine 300mm wafer fabs scheduled to open in 2019 (five of them located in China) following seven that opened in 2018. Nine new fabs in 2019 would be the most opened in a single year since 12 were opened in 2007. Another six are scheduled to open in 2020. All of the new fabs coming in 2019 and 2020 will be for DRAM and flash memory or for foundry capacity.

• The number of active volume-production 300mm fabs declined for the first time in 2013 when ProMOS closed two large fabs and two other fabs that were scheduled to open in 2013 where delayed until 2014. The quantity of 300mm fabs has risen every year since.

• By the end of 2023 there are expected to be 26 more fabs in operation than in 2018, bringing the total number of 300mm fabs used for IC production to 138. For comparison, at the end of 2018 there were 150 volume-production 200mm wafer fabs in operation (the peak number of 200mm fabs was 210).

PDF Version of This Bulletin
A PDF version of this Research Bulletin can be downloaded from our website at http://www.icinsights.com/news/bulletins/

 

[Tanj]

I suspect the curve would be tilting up if it were wafer capacity. Newer fabs produce more than old.
Another insight would be to separate logic, DRAM, and Flash capacities. These days they are not interchangeable fabs so the variations between them tell a story.

 

[simoncc2]

is 450mm fab idea is completly dead ?
does it become more and more financially viable as we move to smaller technology nodes ?

 

[Daniel Nenni]

is 450mm fab idea is completly dead ?
does it become more and more financially viable as we move to smaller technology nodes ?

Good question, let me look into that. My guess would be no since the equipment guys now have a very limited customer base.

 

[Tanj]

Last I heard 450 was shelved. 450mm ▸ News and Analysis | 450mm.com
450mm | SEMI.ORG
450mm, EUV, Scaling, 3D NAND and FD-SOI

It looks like death by a thousand cuts. So much equipment would need to transition, while still supporting 300. EUV. CMP. Thinning. TSV. Etc. The momentum of 300mm and diversity of equipment makes the supposed 10 to 15% improvement a rather low benefit for all the risks involved, and there are plenty of other things to bet on.

 

[Tanj]

I also found this: Semiconductor Engineering .:. 450mm Silicon Wafer Issues Emerge

One thing that looks alarming is the 450 wafers are even thicker than 300, while the chips being produced are thinned to 10% of that.

The 156mm squares used in the solar cell industry are mass produced in thin form by exfoliation with buried oxide, seems like a fundamentally cheaper way to go. And easier handling as things now need to be thin. I wonder if bigger really makes sense, or if it all could be rethought on a more compact footprint optimized for thin silicon. Those guys aim to crank out a billion squares per year and would use the Si 10x more efficiently (indeed, that is why they focus on that size and thickness).

 

[Daniel Nenni]

So why is TSMC building 8" fabs? Cheaper equipment?

New TSMC 8-inch fab to fulfill robust automotive chip orders
Monica Chen, Taipei; Willis Ke, DIGITIMES Friday 22 March 2019
TSMC's new 8-inch wafer fab to be built in Tainan will have its capacity mostly fulfill robust orders for automotive chips from STMicroelectronics and other dedicated chipmakers, according to industry sources.

A major reason behind TSMC's decision, announced in December 2018 by its CEO CC Wei, to built an 8-inch fab - 15 years after it last built one - is that a new plant is needed to meet substantial orders the company has secured from major vendors of automotive chips, the sources indicated.

At the moment, many leading makers of automotive chips have placed orders with TSMC, and even Nvidia and Qualcomm, long-term clients of the foundry giant, are also aggressively venturing into the automotive chip segment, the sources noted. For TSMC, automotive chip foundry orders will become one of the firm's major growth drivers after 2020.

The sources continued that vendors of fingerprint ID ICs, power management ICs , MOSFET chips and IGBT chips have shifted from 6-inch wafer fabs to 8-inch ones for foundry support, while automotive chips and IoT chips see significant lower production costs at 8-inch fabs than at 12-inch ones.

Accordingly, all 8-inch fabs in Taiwan and China are running at full capacity, prompting foundry houses to carry out capacity expansion plans. Besides TSMC, Taiwan's Vanguard International Semiconductor (VIS) and United Microelectronics (UMC), as well as China's Semiconductor Manufacturing International Corp (SMIC) have also announced plans to expand their 8-inch foundry capacities.

SMIC has decided to expand equipment purchases from ASML for its 12-inch wafer fabs in Shanghai and Shenzhen in exchange for the latter renewing production of 8-inch equipment to support capacity expansion at its plant in Tianjin, said the sources.

 

[Tanj]

Some of those chips might not be Si? GaN for power, for example? Not sure what the IGBT use, but it points to there being a segment of equipment optimized for 8 inch. If you are making small devices the wafer size is not so important.

The 450mm wafer is double the area but the claims for throughput were only 15% at best, which shows that the size of wafer is marginal in many steps. The 8 sq cm reticle size may be the main reason to want a larger wafer, minimizing edge effects both of lost chips and wasted exposure time.

 

[ConnerVT]

Perhaps because no one really needs a 4GHz processor in their microwave oven.

The processes and the tooling are available and well tested for the older nodes processed on 200mm wafers. MTTF for automotive is extremely rigid, and doesn't require bleeding edge technology nodes.

Sticking with mature processes, with folks who have been working with them for years, vs. reinventing it all again (and then proving it bulletproof), is much more cost effective.