AMD Splits In Two, Planning New Fab in 2009

Rob Williams

Editor-in-Chief
Staff member
Moderator
From our front-page news:
We first learned about AMD's "Asset Smart" plan a few months ago, but the company today made an official announcement that the company would finally be splitting up into two entities. To help rid some of their debt and also make production easier, AMD has teamed up with an Abu Dhabi company, Advanced Technology Investment Company, who will together create "The Foundry Company" - a temporary name.

Once the new company is established, it will have a total value of $5B, which consists of AMD's current Fab, ATIC's contribution of $1.4B and another $1.2B of debt assumed by 'The Foundry Company'. The only two owners of the new business will be AMD and ATIC, although the latter will own a slightly larger stake at 55.6%. Both companies have equal voting rights, however.

This move is no doubt going to help AMD improve both their business and products over time, and the plans laid out so far look good. Construction will begin in early 2009 in Saratoga County, NY for the first 300mm Fab producing silicon on a 32nm process. This promises to create thousands of jobs in upstate New York and will become the only independently-managed semiconductor manufacturing foundry in the US.

When all said and done, AMD will be part of two companies based in five different locations. Their main HQ will remain in Sunnyvale, California, while their current Fab in Dresden, Germany, will also. Their offices in East Fishkill, New York and Austin, Texas will stay put, with the addition being the upcoming Fab in upstate New York, resulting in two NY-based locations.

In related news, Mubadala, also based in Abu Dhabi, has bought $314 million worth of newly-issued shares and as a result, has bumped its stake in the company to 19.3%, from 8.1%. The stark increase for so little money is due to the fact that the company is now split in two, so the values of each has dropped. They say today is a landmark day for AMD though, and it's hard to disagree. Their debt is being remedied and a new Fab is right around the corner (at 32nm, no less), which should help push AMD towards becoming much more serious competition for Intel. The coming year is going to be an interesting one.

amd_assetlight_100708.jpg

On Oct. 7, 2008, AMD and the Advanced Technology Investment Company announced the intention to create a new global enterprise, The Foundry Company, to address the growing global demand for independent, leading-edge semiconductor manufacturing. There is a strong shift to foundries occurring – particularly to foundries with the capacity to produce devices using leading-edge process technologies.


Source: AMD Global Foundry Page
 

Rory Buszka

Partition Master
The idea of AMD (the chip engineering business unit) going completely 'fabless' makes quite a bit of sense now, since AMD would have access to a great deal of foundry capacity to manufacture its chips, both at TSMC and at The Foundry Company, as well as at their old friends Chartered Semiconductor Manufacturing, who have a technology roadmap out as far as 22nm, and who have made products for AMD in the past. Intel might have a hard time competing, since they've traditionally owned all their own fabs, and their chip process roadmap depends on their ability to develop smaller chipmaking processes as time goes on, which will be a continual drain on Intel's manufacturing budget. Intel could see their process technology lead slip away relatively soon under the new arrangement at AMD.

Of course, as we've recently heard, Intel is going ape over the protection of their X86 intellectual property, but I see three key counterarguments here. First of all, AMD began manufacturing X86 processors all on their own, with no pre-existing cross-licensing agreement between AMD and Intel for Intel's 'trade secrets'. Which means they must not have been very 'secret' to begin with, since AMD figured out how to do it. AMD simply settled out of court back then because they didn't have a whole lot of money to play with to win a court battle against the 'trade secret' status of X86 technology. Secondly, Intel didn't complain one bit when Chartered Semiconductor made AMD Athlon chips (which,of course, were X86) some time ago to help AMD fulfill demand. Thirdly, when AMD announced that the CPU core component of their 'Fusion' CPU would be manufactured by TSMC, we heard no such complaining from Intel. So it looks like Intel is simply searching for possible toeholds in the legal system to prevent the AMD split from going through as planned, and wouldn't otherwise have a problem with AMD using another fab house to manufacture their chips -- except that now, AMD will be able to source massive quantities of their products, driving costs down, instead of being hamstrung by their aging, obsolete chip fabs.

It will be interesting to see how AMD's fortunes turn around with this influx of cash from the UAE, and freed of the burdensome task of figuring out how to manufacture chips profitably at smaller and smaller process sizes, not just design them.
 
Last edited:

Rob Williams

Editor-in-Chief
Staff member
Moderator
I tend to agree a bit with the thought that Intel could find themselves a little behind the curve when sticking to their own fabs, but Intel as a whole has a lot more money kicking around than AMD, and they are not going to be too hurt by investing billions into their fabs. Though, whoever gets a product to 22nm first deserves intense praise.

Regarding the Intel vs. AMD IP issue, it's an interesting one, but I don't know how successful Intel is going to be there. Doesn't Intel license x86-64 from AMD? That would make for one odd situation. Also, as far as I'm aware, one of the main caveats was that AMD had to always retain a 50% voting right, and with the "Foundry" in place, that's exactly the case. I'm no legal major though, so I'm simply going to have to sit back and wait to see what happens. I don't expect much to...
 

Kougar

Techgage Staff
Staff member
I am not 100% sure, but the licensing issues should result from AMD using Intel's microcode extensions... SSE, SSE2, SSE3, etc. Just look at a CPUz screenie of Deneb, it uses SSE4a which is what Penryn introduced for the first time. Intel certainly won't give those away for AMD to use for free, and those would be considered propriatary Intel tech. ;)

Anyway, I did find this AMD/Intel licensing agreement dating from 2001: http://contracts.corporate.findlaw.com/agreements/amd/intel.license.2001.01.01.html


I would agree with Rory that this is just Intel making a verbal fuss over AMD suddenly springing back to life with a promising potential future, and I don't think Intel can do anything about it and knows as much. After all The Foundry Company is just producing the chips, nothing else. Such provisions were specifically made in that old 2001 cross-licensing agreement, and I'm sure they exist in whatever current agreement exists.
 
Last edited:

Rob Williams

Editor-in-Chief
Staff member
Moderator
The SSE4 thing is a huge mess... I'm not even sure exactly how it works. As far as I know, AMD uses SSE4 instruction sets, but not the extra SSE4.1 instruction sets. Likewise, SSE4a is a group of instruction sets produced by AMD that are only found on their own processors... not Intel's. Still, AMD does license a lot of SSE technology from Intel, so that could be one of the most important factors here.
 

Kougar

Techgage Staff
Staff member
Huh, what a ripe mess. The wiki makes sense thogh, it follows what I knew about it. http://en.wikipedia.org/wiki/SSE4

As the story goes Penryn was expected to introduce SSE4... at some point Intel made the call to do a 180 and instead released what amounts to an incremental update to SSE3, and saved the actual SSE4 for Nehalem. That's what I knew, but at least the wiki seems to set the story straight. I remember hearing "SSSE3" tossed around before.

So officially, Penryn debuted SSE4.1, and Nehalem will debut SSE4.2. Phenom debuted SSE4.1a instead, which =! SSE4.1.

That would still mean AMD must license SSE3, SSE2, etc though....
 

Rob Williams

Editor-in-Chief
Staff member
Moderator
Well, Penryn did include a sweet set of SSE4 instructions, so I'm not sure Intel had plans to gimp the CPU for the sake of holding off the good stuff until Nehalem. I do know that SSE4.2 is supposed to add a few killer instructions though for media buffs, although for some reason, none of those instructions listed at that Wikipedia article resemble what I remember discussing.

As for AMD, they would need to license SSE, and that's what makes this thing a huge mess. It's not as though they can just ignore them either, because there is so much out there that's optimized using the SSE instructions. If AMD doesn't support them, then certain tasks on Intel's CPUs would be tremendously faster.

Still, it's hard to see where this Intel whining will get them. They might have a good case and reason to speak up, but I'm not well-versed enough in law to even assume, heh.
 

Kougar

Techgage Staff
Staff member
IMO it's just whining because they want to whine about it. Corporate execs at multi-million giants do it all the time, often when it servers no direct purpose. I always figured it had a bit of PR behind it though.

Anyway, thanks for pointing out SSE4a and SSE4.1 don't match. I'd goofed on that one, and kinda important to know those things... :)
 
Top