cageymaru

Fully [H]
Joined
Apr 10, 2003
Messages
22,060
Digitimes is reporting that Global Foundries may be sold after its downsizing that started last year. On January 31, 2019 the company sold its 200 mm fab located in Singapore to Vanguard International Semiconductor and reduced its workforce by 5% last year to cut costs. The foundry has fallen behind in its competition with Samsung Electronics and TSMC. Industry insiders think that the fab will be sold to a South Korean company like Samsung Electronics or SK Hynix.

At present, ATIC, a state-run enterprise of the United Arab Emirates, is the largest shareholder in Global Foundries with a shareholding of 90 percent. "Global Foundries is unlikely to be bought by a Chinese company such as SMIC in that the U.S. government is keeping China in check in various industries," said an industry insider, adding, "The most potential candidates include South Korean companies such as Samsung Electronics and SK Hynix, and Samsung Electronics can increase its share in the market to 23 percent at once if it takes over Global Foundries."
 
"200mm fab" - One of three possibilities:
  1. That's a typo.
  2. They're referring to the full die size.
  3. Douglas Adams was right and the earth really IS a processor.
 
Last edited:
I can't imagine anyone wanting to buy a fab that's at least 2 generations behind the curve. It would take $15 billion plus to bring it up to speed.

But I guess if you just want to make chips for toys and washing machines, then its more than ok, for two or three years.
 
Hopefully none of the top manufacturers will gobble it up. There's value in spreading these techs around. GloFo might still have some tech in their portfolio that might be tasty to the major actors though.

Otherwise there's relatively easy money to be made in making chips on existing nodes. Being on the cutting edge is risky business.
 
They already sold one of (not Burlington) their 200mm fabs and are in the process of getting rid of one of the Singapore fabs, the article is referring to the 300mm fabs that are left
 
Last edited:
I can't imagine anyone wanting to buy a fab that's at least 2 generations behind the curve. It would take $15 billion plus to bring it up to speed.

But I guess if you just want to make chips for toys and washing machines, then its more than ok, for two or three years.

but they really aren't.. the enterprise/consumer side is only a small portion of the what TSMC/glofo/samsung/etc produce. they produce a lot of other things like the fab that glofo sold in January which produces parts for microelectromechanical systems and analog/digital audio chips.. hell, almost all the fabs in north america produce commercial/industrial/military hardware applications.
 
As long as AMD and others get their needs met that is what counts at the end of the day in my books.

Hopefully they still fill the void for AMD until someone else comes in to fill the space for what they do still get from them (such as 14nm and 12nm as well as ofc the 14nm chiplet they currently fab, or will AMD move this to TSMC as well?

I am sure at the end of the day whatever meets the needs (or better) when it comes to price per unit, performance, power consumption etc maybe it is quite likely by GF getting out and basically making the current WSA null and void sure hope AMD is working double time on finding a worthy replacement as required.

Woule not be good if GF is out say in a year whereas they generally keep CPU as example on open market a good 3-5 years (needs replacement SKU) are they planning on keeping a "skeleton" foundry for this sole intention, not for new product development but for current product phase out preparation?
 
"200mm fab" - One of three possibilities:
  1. That's a typo.
  2. They're referring to the full die size.
  3. Douglas Adams was right and the earth really IS a processor.
https://en.wikipedia.org/wiki/Wafer_(electronics)#Standard_wafer_sizes
200mm is one of diameters silicon wafers arrive in. When they say 200mm lab they mean it is only equipped for that size wafer, but doesn't specify the size of the transistors it can etch. Don't quote me on this but I was under the impression CPUs and GPUs were primarily derived from 300mm wafers.
 
I can't imagine anyone wanting to buy a fab that's at least 2 generations behind the curve. It would take $15 billion plus to bring it up to speed.

But I guess if you just want to make chips for toys and washing machines, then its more than ok, for two or three years.


It would be far easier to retro fit an existing fab then to completely start from scratch. It's not only about the money saved by not building a new facility but the infrastructure is already done for the existing fab.
 
Basically all advanced process nodes are 300mm, legacy process nodes are likely on smaller wafers. A few years back everyone had planned to move to 450mm, but I think that all fell apart due to the cost of having to replace all their equipment in the fabs. It could be that when fabs start having to consider EUV to advance that they'll re-examine 450mm, as they'll be replacing all their stuff anyway.
 
I can't imagine anyone wanting to buy a fab that's at least 2 generations behind the curve. It would take $15 billion plus to bring it up to speed.

But I guess if you just want to make chips for toys and washing machines, then its more than ok, for two or three years.

Many of the tools can be used without modification. Obviously they would need new scanners for EUV litho, but implanters, etchers, wet etchers, furnaces, etc. are all able to handle smaller nodes with the appropriate recipe modifications.

Intel was cannibalizing their 45nm tools to send to the Irish/Arizona 22nm fabs, and the same tools are now working on 14nm. It's quite viable.
 
Back
Top