AMD Amends Wafer Supply Agreement with GLOBALFOUNDRIES

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
AMD today announced that it successfully amended its Wafer Supply Agreement (WSA) with GLOBALFOUNDRIES Inc. The closure of amendment negotiations solidifies AMD’s new operating model as communicated at the third quarter 2012 earnings announcement. To better align with today’s PC market dynamics, AMD and GLOBALFOUNDRIES agreed on purchase commitments for fourth quarter 2012 and established fixed pricing and other terms of the WSA which apply to products AMD will purchase from GLOBALFOUNDRIES through Dec. 31, 2013.
 
Seems AMD is always paying someone a penalty.

one time charge of $165 million this quarter, $200 million loss another quarter, $100 million writedown, excess inventory, and whatever remains of the $320 million penalty.

People already expect AMD to bleed $150 million every quarter.
 
I found this part interesting:
Separately, as AMD moves to standard 28nm process technology, AMD will reduce future reimbursements to GLOBALFOUNDRIES for certain research and development costs.
Click to expand...

I guess that makes it more definite they are planning on moving to 28nm and not skipping a node down to 20nm. Now the question is when that shrink will happen...
 
This could be deadly for AMD:

"AMD estimates that it will purchase $115m worth of wafers from GlobalFoundries this quarter under the amended WSA. The previous agreement between AMD and the company, which it spun off in a deal with the Abu Dhabi–based investment firm ATIC in 2009, had been for $500m worth of chips."

If they now expect to sell less than one third of the chips that they previously expected to sell, they may be in a tail spin that they will never recover from.

I may have to rethink an upcoming AMD build.
 
The title sounds so innocent, but bad business is really, really screwing AMD. Ars has an article this afternoon with all the ugly details, including pretty bad break-even numbers that are over a year out by the interim CEO's rosy projections (around $5 billion a year in sales, compared to around $8 billion a year just a couple of years ago after shedding fabs).
 
Silent.Sin said:
I found this part interesting:


I guess that makes it more definite they are planning on moving to 28nm and not skipping a node down to 20nm. Now the question is when that shrink will happen...
Click to expand...

If going by any official roadmap from AMD, 20nm or smaller won't happen until when AND if Excavator gets released. If you recall, AMD was planning on introducing high-density libraries with that architecture-- basically squeeze more into a smaller space. That would be great if you're going to squeeze GPU cores onto the same die-space to be used as an FPU-replacement, which is what AMD is intending to do in the future.

But, at the moment, with all desktop consumer (non-server) processors delayed, not sure when the next die shrink from 28nm to 20nm or 14nm will happen. Only ones getting attention now are desktop and mobile APUs-- Kaveri, Kabini, Temash, etc.-- and server processors-- Opteron and ARM-Opteron. At best we'll see refreshes of current processors until Excavator.

It'll probably look like: Piledriver -> Piledriver refresh (2013) -> Steamroller -> Steamroller refresh -> Excavator. Something like that to keep AMD barely on its feet until it gets the resources again to work on Excavator or an entirely new architecture.

You have to remember, CEO Read is taking drastic, if not risky, measures to return AMD to break-even or profitability. Laying off employees, delaying processor releases, and now amending wafer supply agreements with GloFlo. Alot of the investments right now it seems are going into mobile and desktop APUs and server processors. I'm not sure if their graphics division is affected by any drastic cuts other than lack of employees. At the least, Radeon 8000-series should still be out by the middle of 2013 or summer of 2013.
 
octoberasian said:
If going by any official roadmap from AMD, 20nm or smaller won't happen until when AND if Excavator gets released. If you recall, AMD was planning on introducing high-density libraries with that architecture-- basically squeeze more into a smaller space. That would be great if you're going to squeeze GPU cores onto the same die-space to be used as an FPU-replacement, which is what AMD is intending to do in the future.

But, at the moment, with all desktop consumer (non-server) processors delayed, not sure when the next die shrink from 28nm to 20nm or 14nm will happen. Only ones getting attention now are desktop and mobile APUs-- Kaveri, Kabini, Temash, etc.-- and server processors-- Opteron and ARM-Opteron. At best we'll see refreshes of current processors until Excavator.

It'll probably look like: Piledriver -> Piledriver refresh (2013) -> Steamroller -> Steamroller refresh -> Excavator. Something like that to keep AMD barely on its feet until it gets the resources again to work on Excavator or an entirely new architecture.

You have to remember, CEO Read is taking drastic, if not risky, measures to return AMD to break-even or profitability. Laying off employees, delaying processor releases, and now amending wafer supply agreements with GloFlo. Alot of the investments right now it seems are going into mobile and desktop APUs and server processors. I'm not sure if their graphics division is affected by any drastic cuts other than lack of employees. At the least, Radeon 8000-series should still be out by the middle of 2013 or summer of 2013.
Click to expand...

All sounds like great plans in a highly technologically cut throat business where you have a new player in ARM coming up to eat away your low end sales. lol yeah right, fire all your talent and work force when you already cant get basic jobs done. We have seen this all before it just wont work. AMD needs to take bigger risks not try to reduce risk. Its like you are being attacked and your solution is to try to wall up in the bomb shelter while your enemy sits outside and waits for you to die of starvation or come out and get shot single file.

The real question is soon going to be when are we going to start seeing global foundries start looking to ARM partners as their major source of production.
 
rudy said:
The real question is soon going to be when are we going to start seeing global foundries start looking to ARM partners as their major source of production.
Click to expand...

That might happen sooner than later, and before AMD can get any benefit from it. GlobalFoundries already previously announced 14nm process with ARM, and 3D transistors for ARM. AMD is not mentioned anywhere regarding those two. So, a company like Apple with their Ax SoCs, or Samsung with their Exynos, Nvidia with Tegra, or Qualcomm may look to GlobalFoundries for smaller, future mobile processors. Outside of Intel, I believe GlobalFoundries is the only other fabricator able to be 14nm ready in the next few years and have 3D transistors. I'm unsure of TSMC will be moving to 14nm anytime soon. I think TSMC will go to 20nm next.

AMD will either a) need a new fabricator for their processors (TSMC?), or b) work with GlobalFoundries and try to benefit from the new smaller process and 3D transistors.
 
The 14nm-XM GloFo node is actually just a revised 20nm node with FinFETs and a 20nm BEOL.

But, yes, it's focused on low power - XM standing for eXtreme Mobility. AMD won't have access to that process for their high power chips until the process is more mature and GloFo puts some work into it. They can take advantage of that node for their low power SoCs, though. I also believe that's it's bulk and not FD-SOI, so it would require AMD designing the chips for bulk process; something they did with Bobcat/Brazos for 40nm at TSMC.
 
pelo said:
The 14nm-XM GloFo node is actually just a revised 20nm node with FinFETs and a 20nm BEOL.

But, yes, it's focused on low power - XM standing for eXtreme Mobility. AMD won't have access to that process for their high power chips until the process is more mature and GloFo puts some work into it. They can take advantage of that node for their low power SoCs, though. I also believe that's it's bulk and not FD-SOI, so it would require AMD designing the chips for bulk process; something they did with Bobcat/Brazos for 40nm at TSMC.
Click to expand...

And, if I recall, I don't think AMD has any resources to change their processor designs for a bulk process or a different process from GloFlo. Wasn't there another article posted in the AMD forum about this?
 
octoberasian said:
And, if I recall, I don't think AMD has any resources to change their processor designs for a bulk process or a different process from GloFlo. Wasn't there another article posted in the AMD forum about this?
Click to expand...

I'm not sure about that. At best case scenario it's going to be at least 2-3 years until AMD reaches that point and they do have experience doing bulk. That also assumes that AMD didn't know what GloFo's future processes would be... which is insane. AMD most definitely knew 14nm-XM would be bulk and they were well aware of what process GloFo would have available for them in the following 3+ years.

The amendment to the WSA isn't due to AMD being unhappy with what GloFo is offering, though that could be true, but it definitely revolves around lower demand. AMD had to restructure the WSA so they wouldn't suffer from oversupply/overstocked issues like they did with Llano, where AMD had to write off a whole lot of Llano APUs that they couldn't sell off. They're selling a lot less chips than they thought they would, so the WSA would need restructuring in order to reflect that.
 
octoberasian said:
And, if I recall, I don't think AMD has any resources to change their processor designs for a bulk process or a different process from GloFlo. Wasn't there another article posted in the AMD forum about this?
Click to expand...
People just get SOI crazy. AMD has supposedly seen the end of PD-SOI and has been working to move to bulk Si processes for a few years. Less knowledgeable fans hear about FD-SOI or other things and "wish" AMD to move to that instead (FD-SOI is not nearly the same as PD-SOI, design-wise).

You can kind of see this already by AMD sticking to 32nm for some products for much longer than usual. When bulk Si products are ready, AMD will move those previously PD-SOI products to smaller process nodes.

Of course, if the bulk move doesn't work out and GLF is no longer an option, there's always IBM's expensive foundry. :p
 
I got kind of depressed when I heard this news. Things are pretty dire for AMD. They're getting destroyed by the market and their own internal operations. Nothing is going good for them. I really rooted for them back in the K6-2-A64 days. They made some good stuff, but this time they don't have anything up their sleeves that looks like it has a hope in hell of turning things around.
 
I think this is a case of cutting out global for more TSMC. Regardless this has limited impact that's spread out over time.

Still, they need to up the performance of those APUs, and they need a shrink. The ARM waters are dark and deep, and since they've chosen to also wade into them there's a lot of architecture floating around for not a lot of engineers.
 
You must log in or register to reply here.
Back
Top