Where do fabs go when they die?

[DrMrLordX]

I'm assuming that most chip fabs, foundries, etc. retool for new runs of chips when the product in production is EoLed. However, I also have to assume that smooth transitions from one production run to the next do not always happen.

For example, I remember Sony blowing over $2 billion on fabs for Cell. They had planned on using Cell in many, many products which never really happened (at least not the way they had hoped). To the best of my knowledge, Cell is not used in anything anymore, so . . . what happened to those facilities? Do they still crank out Cell chips for some applications? Did they sell off all the equipment and shut down the facilities?

 

[BUnit1701]

The 'tooling' and 'transitions' happen from node to node, i.e. 45nm->32nm, and given product has to be designed to work on that node. In the specific case of Cell, Sony spent the money on the design, not on fabs. In the fab world, $2B is pocket change.

 

[Tsavo]

Around here, they become Dollar Marts.

 

[Lonyo]

Some get used for a long long time for older products, e.g. Intel and TSMC. Some do eventually get retooled.

Intel produce chipsets on their older nodes. TSMC have a lot of customers who don't need bleeding edge fabs for their products.

http://download.intel.com/newsroom/kits/22nm/pdfs/Global-Intel-Manufacturing_FactSheet.pdf

 

[ShintaiDK]

Retooled until its not economicly sensible anymore. Then they get demolished or reused if possible as something entirely different.

That list is also outdated. Its made the 4th May 2011.

The 200mm Fab 17 is also either closed already or in the process of being closed.

D1X is 2 fabs, M1 and M2. D1XM1, D1XM2 and Fab42 are 450mm ready.

 

[Ajay]

Retooled until its not economicly sensible anymore. Then they get demolished or reused if possible as something entirely different.

That list is also outdated. Its made the 4th May 2011.

The 200mm Fab 17 is also either closed already or in the process of being closed.

Yep, Fab17 (bought from DEC, IIRC) won't work with the larger floor plans of Intel's modern fabs because of it's copy-exact manufacturing methodology.
http://www.bostonglobe.com/metro/20...ant-lay-off/wBpsOao3dJp6wbHxgcTmiO/story.html

 

[Roland00Address]

Pretty much if the fab was created in the last 8 years it is advanced enough that it can be retooled or it can still make something out there that is economically feasible. I am talking things like chipsets or embedded that do not need to be on the latest process.

Well after 8 years (going 10+) well then it just does not make sense anymore for the electricity and the wafers costs are not worth the old fashioned process you are still running.

 

[Ayah]

Around here, they become Dollar Marts.

In a sense, they kinda do. You can buy a PSOC4200 for 1$, which is Cortex-M0 based.

AFAIK, everything from 14nm to 180nm is running.

 

[Idontcare]

I'm assuming that most chip fabs, foundries, etc. retool for new runs of chips when the product in production is EoLed. However, I also have to assume that smooth transitions from one production run to the next do not always happen.

For example, I remember Sony blowing over $2 billion on fabs for Cell. They had planned on using Cell in many, many products which never really happened (at least not the way they had hoped). To the best of my knowledge, Cell is not used in anything anymore, so . . . what happened to those facilities? Do they still crank out Cell chips for some applications? Did they sell off all the equipment and shut down the facilities?

In most cases the shell (the building) will continue to be used in some capacity for its clean-room facilities alone.

Plenty of demand for older clean-room facilities within the semiconductor industry (analog chips, hard-drives, packaging, etc) as well as within other industries such as medical and aerospace.

The equipment itself can easily be sold off at about 40-50% new purchase price in the second-hand market. Lots of demand of used tools in the foundry and Asian semiconductor markets.

Not too unlike second-hand cars, at TI we use to buy old tools simply to use them as spare parts for keeping our own dilapidated tools running in the fab.

Only in very rare cases does an outdated fab (the shell) truly serve no good purpose - be it from haz-mat contamination or simple lacking of economic scale and location - such that it gets converted into office space or out right demolished.

 

[ninaholic37]

They don't go to heaven where the angels fly. They go to the lake of fire and fry...

 

[dave_the_nerd]

They don't go to heaven where the angels fly. They go to the lake of fire and fry...

That song was stuck in my head before I read your comment.

You just made it worse. Thanks.

 

[Nothingness]

Retooled until its not economicly sensible anymore. Then they get demolished or reused if possible as something entirely different.

And it seems this can result in environmental issues: Google Buildings Polluted With Toxic Vapors From Chemical Spills. I guess modern fabs demolition is now better handled.

 

[ShintaiDK]

And it seems this can result in environmental issues: Google Buildings Polluted With Toxic Vapors From Chemical Spills. I guess modern fabs demolition is now better handled.

Yep. Its a classic of the mill. You can more or less replace Intel with any company that works with chemicals and declare their ground polluted if they have been there for more than 15-20 years. Its only a matter of how much. Back in the old days chemical companies in all western countries could do as they wish. Toxic barrel dumps, direct dumping in lakes and seas etc. Back then it was considered good ethics to let money rule over getting some businsses. Today, countries still cleanup after it. And some places are just deemed polluted beyond cleanup for an unforeseeable time. Sometimes you see places where they just put a membrane over and simply hope for the best.

I must comment how scary it is to read the article tho. The "journalist" feels compelled to mention and use half the article to defend fracking that may end up as the biggest environmental disaster is history.

 

[Nothingness]

Yep. Its a classic of the mill. You can more or less replace Intel with any company that works with chemicals and declare their ground polluted if they have been there for more than 15-20 years.

Agreed that it could have been anyone and it looks like Intel is now being much greener than many other companies.

Its only a matter of how much. Back in the old days chemical companies in all western countries could do as they wish. Toxic barrel dumps, direct dumping in lakes and seas etc. Back then it was considered good ethics to let money rule over getting some businsses. Today, countries still cleanup after it. And some places are just deemed polluted beyond cleanup for an unforeseeable time. Sometimes you see places where they just put a membrane over and simply hope for the best.

And now western countries dump their electronic mess in Africa.

I must comment how scary it is to read the article tho. The "journalist" feels compelled to mention and use half the article to defend fracking that may end up as the biggest environmental disaster is history.

I also found that completely misplaced, and also feel the same as you do about the future impact of fracking.

 

[DrMrLordX]

Hmm! Xylene, tasty. So I guess some old fabs never die, they just come back up out of the Earth to take revenge.

In all seriousness though, thank you all for the responses. Makes me wonder what happens to all the fab equipment in the long run. Does it all get cannibalized to repair newer equipment as Idontcare indicated? You'd think people would find a use for the stuff, considering how much it cost new.

It would also be interesting to know who still used, say, equipment Intel used for their p54 or p55 chips from the 90s (for example).

 

[Idontcare]

Hmm! Xylene, tasty. So I guess some old fabs never die, they just come back up out of the Earth to take revenge.

In all seriousness though, thank you all for the responses. Makes me wonder what happens to all the fab equipment in the long run. Does it all get cannibalized to repair newer equipment as Idontcare indicated? You'd think people would find a use for the stuff, considering how much it cost new.

It would also be interesting to know who still used, say, equipment Intel used for their p54 or p55 chips from the 90s (for example).

Tools take a tremendous amount of routine maintenance, and require a lot of replacement parts that are considered "consumable".

Think about an injet printer or a laser printer these days, you spend nearly as much in consumables (ink/toner/fuser) as you do in purchasing the tool.

Now fast forward 10yrs after purchase and you find the original equipment manufacturer doesn't support or manufacture those parts anymore.

No where else to get the spare parts except to try and make them yourself in a machine shop or by purchasing used tools and chop them for parts.

 

[microAmp]

Some collect "dust". Unused after being shutdown. Here is an overhead view of Freescale, former Motorola SPS before being spun off. I worked in MOS 8 and transferred out to another site back in 2001

BAT 1 was shutdown during that time. MOS 2 & MOS 3 were shutdown about a year or two earlier. All that remains is MOS 13, which use to make G4 processors for Apple, no idea what their focus is now. There are also some offices and auditorium between BAT 1 and MOS 8. 3 story office and a very nice cafeteria by MOS 8, shutdown and it a new building put next to MOS 13.

I don't know much about the area of MOS 2 & MOS 3.

Some of that site last I heard was going to be bought by University of Texas they backed out for their reasons. Perhaps clean up is too expensive?

http://pics.bbzzdd.com/users/motoamd/Freescale.png

 

[soccerballtux]

I'm assuming that most chip fabs, foundries, etc. retool for new runs of chips when the product in production is EoLed. However, I also have to assume that smooth transitions from one production run to the next do not always happen.

For example, I remember Sony blowing over $2 billion on fabs for Cell. They had planned on using Cell in many, many products which never really happened (at least not the way they had hoped). To the best of my knowledge, Cell is not used in anything anymore, so . . . what happened to those facilities? Do they still crank out Cell chips for some applications? Did they sell off all the equipment and shut down the facilities?

yeah cell was supposed to be their content rendering system...why else all the SPEs / truncated FPUs ... perfect for decoding bluray

 

[Borealis7]

They go to Florida?

 

[aigomorla]

Tools take a tremendous amount of routine maintenance, and require a lot of replacement parts that are considered "consumable".

Think about an injet printer or a laser printer these days, you spend nearly as much in consumables (ink/toner/fuser) as you do in purchasing the tool.

Now fast forward 10yrs after purchase and you find the original equipment manufacturer doesn't support or manufacture those parts anymore.

No where else to get the spare parts except to try and make them yourself in a machine shop or by purchasing used tools and chop them for parts.

This^

sometimes it costs more for intel to refab a exisiting node so they dont even bother and set it up for demolish, or they sell it.

it would need to be run though a actuary and see if profits can be produced.
Ie... revamping the node vs moving it out of country.

 

[Nothingness]

A nice way to recycle a fab

 

[Topweasel]

I'm assuming that most chip fabs, foundries, etc. retool for new runs of chips when the product in production is EoLed. However, I also have to assume that smooth transitions from one production run to the next do not always happen.

For example, I remember Sony blowing over $2 billion on fabs for Cell. They had planned on using Cell in many, many products which never really happened (at least not the way they had hoped). To the best of my knowledge, Cell is not used in anything anymore, so . . . what happened to those facilities? Do they still crank out Cell chips for some applications? Did they sell off all the equipment and shut down the facilities?

For your example Sony sold the factory dirt cheap to Toshiba as a peace offering probably to help convince Toshiba to bow out of the HD race (I have a feeling the immediacy and completeness of HD-DVD's death came from this transaction). Which Toshiba and Sandisk probably use most of this factory at this point to do nand work. Really just about any fab that isn't doing semi modern CPU or SoC work is probably doing Nand. Now that Samsung has shown that older processes work just as well with a lot more leg room with 3D Nand.

I am guessing we won't see to many retirements or reconverts. We will just see new ones put up. Specially if 3d nand allows for the complete replacement of spindle drives like the projections like Anands article on the 850 suggests.

 

[DrMrLordX]

For your example Sony sold the factory dirt cheap to Toshiba as a peace offering probably to help convince Toshiba to bow out of the HD race (I have a feeling the immediacy and completeness of HD-DVD's death came from this transaction). Which Toshiba and Sandisk probably use most of this factory at this point to do nand work. Really just about any fab that isn't doing semi modern CPU or SoC work is probably doing Nand. Now that Samsung has shown that older processes work just as well with a lot more leg room with 3D Nand.

I am guessing we won't see to many retirements or reconverts. We will just see new ones put up. Specially if 3d nand allows for the complete replacement of spindle drives like the projections like Anands article on the 850 suggests.

Ahh, so that's what Sony did with it. I know they had several runs of Cell CPUs, some on newer processes (I think there was 45 nm Cell at one point in updated PS3s), but after the PS3 and the compute cards/blade servers, I stopped hearing about Cell.

 

[jpiniero]

A nice way to recycle a fab

Lettuce, huh? :hmm:

 

[coloumb]

Usually converted to another process.

Hynix fab in Eugene closed down and was rumored to be a new data center back in 2012. I visited the fab once to service an optical tool before it was closed down:

http://www.datacenterknowledge.com/...uge-oregon-chip-plant-may-become-data-center/

Japan's Komatsu Group building [microchip plant] in Hillsboro was "closed down" and eventually purchased by Solarworld.

Intel - older generation tools [10+ years old] are either torn down and recycled or sent off to other Intel fabs [Hillsboro Intel moved a lot of the older process tools to F68 in China - which is way under capacity right now. I believe Intel would like to close up shop but can't in order to save face with the Chinese Government].