AMD and GloFo change WSA terms

[kiska3]

Quote

Under the terms of the five-year deal, which sets annual wafer-purchase targets along with wafer prices for this year and a framework for yearly wafer pricing in future years, AMD will pay Globalfoundries $100 million in installments. AMD also agreed to pay Globalfoundries quarterly, beginning in 2017, based on the volume of certain wafers bought from another foundry.

GlobalFoundries and AMD sign a five year WSA(Wafer Supply Agreement). For the price of $100 Million in a one time payment and $235 Million to Mubadala development Co for 75 million shares at a price of $5.98. The warrant is exercisable by 2020, Feb 29th. The Agreement is from 2016, Jan 1st til 2020, Dec 31st. This move will allow AMD to use GloFo 7nm when it comes out(that is). Wonder how that $335 million charge is going to look for investors

 

AMD investors relations release: http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=2198716

 

EDIT:

Looks like the deal has some terms that might make it costly for AMD to change foundries if GloFo is not as good as it is

Quote

GlobalFoundries however, is being very careful with this new amendment and a number of terms in the deal preventing AMD from jumping ship or make it very costly to do so.

This is very dangerous for AMD

Source

http://www.nasdaq.com/article/amd-globalfoundries-change-terms-of-business-agreement-20160831-01108

http://www.custompcreview.com/news/amd-globalfoundries-amend-multi-year-wafer-supply-agreement/32194/

Edited September 1, 2016 by kiska3

[Citadelen]

AMD's lucky that they recently got a large cash infusion.

[Prysin]

10 minutes ago, Citadelen said:

AMD's lucky that they recently got a large cash infusion.

of roughly 350m

[kiska3]

1 minute ago, Prysin said:

of roughly 350m

That's barely leaves enough for that WSA change  

[zMeul]

make it break it

GloFo hasn't been inspired with their current 14nm FinFet

[patrickjp93]

And this is why buying ATI, not merging with Nvidia, and selling its foundries were all BAD things!

[kiska3]

1 minute ago, patrickjp93 said:

And this is why buying ATI, not merging with Nvidia, and selling its foundries were all BAD things!

For one merging with Nvidia for AMD means giving up its x86 licence that intel has 'granted' due to changes in the agreement

[patrickjp93]

1 minute ago, kiska3 said:

For one merging with Nvidia for AMD means giving up its x86 licence that intel has 'granted' due to changes in the agreement

And it would mean Intel loses x86_64. They'd have to renegotiate, but they'd get it done.

[kiska3]

Interesting note is that Mubadala already has 18% of AMD shares, and that once a charge has been issued, it can not be reversed

[Notional]

1 hour ago, kiska3 said:

This is very dangerous for AMD

Why? This is literally what AMD has been doing ever since they sold off GloFo. Rather now that GloFo has a node agreement with Samsung, as well as IBM's tech for 7nm, I really don't see this as an issue at all. In fact, if Vega and ZEN becomes successes, AMD might need to go to Samsung's foundries as well, if GloFo's capacity is not enough.

[rattacko123]

So AMD is paying $100m to GlobalFoundaries?? AMD must have a very good reason to pay to stick with Global Foundaries rather than TSMC or other Fabs.
 

[patrickjp93]

51 minutes ago, Notional said:

Why? This is literally what AMD has been doing ever since they sold off GloFo. Rather now that GloFo has a node agreement with Samsung, as well as IBM's tech for 7nm, I really don't see this as an issue at all. In fact, if Vega and ZEN becomes successes, AMD might need to go to Samsung's foundries as well, if GloFo's capacity is not enough.

GloFo is on its own for 10nm though, and IBM's 7nm demo used quad-level patterning and full EUV. The cost of such a process at scale will be astronomical.

 

10 minutes ago, rattacko123 said:

So AMD is paying $100m to GlobalFoundaries?? AMD must have a very good reason to pay to stick with Global Foundaries rather than TSMC or other Fabs.
 

Mudabala owns significant stakes in both companies. It's just an effort to keep them both alive.

[carzacc]

I don't see it as a positive or negative thing, it's a risk that AMD is taking

[Notional]

1 hour ago, patrickjp93 said:

GloFo is on its own for 10nm though, and IBM's 7nm demo used quad-level patterning and full EUV. The cost of such a process at scale will be astronomical.

 

Mudabala owns significant stakes in both companies. It's just an effort to keep them both alive.

Probably why AMD said they would skip 10nm entirely. If the 7nm node is in conjunction with Samsung's node, I don't see a problem as such.

[patrickjp93]

1 hour ago, Notional said:

Probably why AMD said they would skip 10nm entirely. If the 7nm node is in conjunction with Samsung's node, I don't see a problem as such.

Really? Intel is struggling with 10nm, and you don't see 7 (9nm) being a problem for Samsung and GloFo? Where do they dig you people up? Even IBM could only achieve single-core 7nm chips using full EUV and quad-level patterning. That cost alone will make scale production unaffordable. Samsung is delusional enough to think it can integrate an immature lithography tech partway at 9nm when the only real competition Intel ever had said it couldn't be done. Samsung has scale. It lacks brilliance.

[VanayadGaming]

From what I read the deal is in favour of AMD. They were buying wafers from GloFo anyway, now if they can't deliver, they can buy from other manufacturers as well. Which is nice. 

[Misanthrope]

It's all in with Zen I see. Hope it delivers. 

[Curufinwe_wins]

Anandtech has a pretty good article about this story that goes into detail about each of the parts....

 

It's a pretty richly GloFo weighted document....

 

Also, looks like AMD bought a record low worth of wafers from GloFo in q2 2016 (75M, compared to 180M Q1), and this is with Polaris launch...

Quote

"Total wafer purchases from GLOBALFOUNDRIES in the second quarter was $75 million, and year-to-date, we have purchased $259 million."

 

AMD really needs to get availability down for it's chips or peeps in the US will just be defacto pushed into Nvidia's hands (it's already happened quite a bit on the 480 v 1060 mark).

 

I was honestly expecting some sort of admission about GloFo having yield issues or something and that they needed to make targets better, and while there is a stipulation in that, the gist of this document reflects the bottleneck being on AMD's side.

 

http://www.anandtech.com/show/10631/amd-amends-globalfoundries-wafer-supply-agreement-through-2020

 

[Notional]

17 minutes ago, patrickjp93 said:

Really? Intel is struggling with 10nm, and you don't see 7 (9nm) being a problem for Samsung and GloFo? Where do they dig you people up? Even IBM could only achieve single-core 7nm chips using full EUV and quad-level patterning. That cost alone will make scale production unaffordable. Samsung is delusional enough to think it can integrate an immature lithography tech partway at 9nm when the only real competition Intel ever had said it couldn't be done. Samsung has scale. It lacks brilliance.

And GloFo/TSMC had massive issues with 20nm as well, but 14/16nm is just fine. It's not like we will see any of these nodes any time soon. And of course we can always discuss the definition of these 10/9/7nm sizes anyways. Intel will probably be first, but GloFo/Samsung should not be far behind. I mostly wonder about TSMC though.

[Notional]

17 minutes ago, patrickjp93 said:

Really? Intel is struggling with 10nm, and you don't see 7 (9nm) being a problem for Samsung and GloFo? Where do they dig you people up? Even IBM could only achieve single-core 7nm chips using full EUV and quad-level patterning. That cost alone will make scale production unaffordable. Samsung is delusional enough to think it can integrate an immature lithography tech partway at 9nm when the only real competition Intel ever had said it couldn't be done. Samsung has scale. It lacks brilliance.

And GloFo/TSMC had massive issues with 20nm as well, but 14/16nm is just fine. It's not like we will see any of these nodes any time soon. And of course we can always discuss the definition of these 10/9/7nm sizes anyways. Intel will probably be first, but GloFo/Samsung should not be far behind. I mostly wonder about TSMC though.

[patrickjp93]

5 minutes ago, Notional said:

And GloFo/TSMC had massive issues with 20nm as well, but 14/16nm is just fine. It's not like we will see any of these nodes any time soon. And of course we can always discuss the definition of these 10/9/7nm sizes anyways. Intel will probably be first, but GloFo/Samsung should not be far behind. I mostly wonder about TSMC though.

Because 14/16 is just a slightly tightened 20nm with FinFET added on for them. The hard work had already been done with 20nm.

[Notional]

Just now, patrickjp93 said:

Because 14/16 is just a slightly tightened 20nm with FinFET added on for them. The hard work had already been done with 20nm.

I doubt that, as FinFets are very different from planar. 20nm would have worked fine too, if it had FinFets instead of planar. After all, the reason why Intel could get their 22nm to work, was because of FinFet (Trigate). At the end though, it's pure speculation where Intel/GloFo/Samsung will be in a year or two.

 

As for the topic at hand, I still see this is a good deal, so idk why it's being portrayed as something bad.

[Coaxialgamer]

I actually don't see this as a bad move.  It secures massive amounts of manufacturing power for AMD... 

 

Gloflo's tech just has to be good. 

[Curufinwe_wins]

forum on mobiles is still fucked up.... sigh

[Curufinwe_wins]

1 minute ago, Coaxialgamer said:

I actually don't see this as a bad move.  It secures massive amounts of manufacturing power for AMD... 

 

Gloflo's tech just has to be good. 

You don't get it though.... it was already secured.... now AMD also has to down pay 335M dollars and face a new swath of potential penalties (although the severity of which is unknown).

 

This isn't bad in a vacuum, but it's terrible compared to the status quo.