While I agree with you on the outcome (making AMD look good artificially), I believe GF is charging an acceptable foundry rate, except that it isn't sustainable with only one customer, large underutilisation and massive amounts of R&D/fab construction spending for upcoming processes.
The loss proves that AMD alone can't sustain GF's business model. Of course when GF has 28nm bulk they will actually have more customers which will i) Increase GF revenue, ii) Reduce underutilisation and iii) give a good reason for all the fab-building, bulk process R&D and the foundry-like services (design libraries) that GF had to spend money on to be a
foundry rather than AMD's fabs.
In other words when GF has the number of customers it is designed for, its cost structure will make sense. In particular the loss now could just represent the extra R&D that is going on the bulk processes and fab capacity.
If by the end of 2011 this accounting situation continues then it will obviously be a trick like that. But right now is too early to tell.
What you are arguing here is really that in order for AMD to function at profits it requires a wafer-foundery that is willing to operate at a loss for the AMD account while subsidizing their operating expenses with the profits generated from the rest of their non-SOI non-SHP customer base.
I understand this business model well, it was exactly the same business model we had at TI for our foundry business that handled the SUN account.
We (me, it was actually my job to be part of that team) built specially crafted 100% dedicated versions of our nodes solely for SUN's products. We could only do it because of the gross margins we sustained on our non-SUN process nodes and products.
And despite our willingness to operate in this fashion (which ended after 65nm) you can see where SUN ended up...
The parallels to what GloFo is doing (and suffering for doing) for AMD run deep. I see a lot of people talk the talk here in the forums but I've never met anyone here who actually walked the walk with me in the business world of this industry.
Its really my only motivation to bother taking the time to post the posts I do, I see a lot of ignorance bread from inexperience and I know I have the knowledge and experience to shed some form of education on the matter and I feel a sense of responsibility to do it. I'm sure I could do a better job, but this is pro-bono after all so I'm not worried about delivering professional quality posts.
Don't get me wrong, glofo represents AMD's best chances at remaining competitive with Intel as well as possibly gaining a process-tech advantage over Nvidia as TSMC and GloFo diverge on their process integration. But don't let the accounting shenanigans fool you into believing something that really isn't true.
I kinda feel for GloFo's other customers to be honest, just as I felt a bit sorry for our non-SUN customers at TI, because basically Glofo is going to those other customers saying "we need your business and money so we can keep subsidizing AMD's cost-structure by developing process nodes that simply will never apply to you".
TSMC doesn't do that, it is very much a "pay as you go" attitude about node specialization...which is why SUN used TI instead of TSMC or UMC or SMIC for a couple decades and when TI said no more at 45nm then SUN kinda imploded thereafter.
Facebook
Twitter