
So AMD are paying $435M (including the termination fee) for $115M worth of chips, rather than $500M for presumably $500M worth of chips? Huh. With financial genuises like that in charge, is it any wonder they're doing so badly...
The slumping PC market has put the hurt on AMD to the extent that the struggling chipmaker has sharply reduced its Wafer Supply Agreement (WSA) with its chip-baking partner, GlobalFoundries. AMD estimates that it will purchase $115m worth of wafers from GlobalFoundries this quarter under the amended WSA. The previous agreement …
Rather than pay $500m this quarter to fill the warehouses with stock they can't move, they're paying $195m this quarter and not having to deal with the overstocking problem. Then there's a definite $40m next quarter and a further $200m that isn't due until at least 2014.
I guess you've got to assume that the overstock problem would be a huge cost (with depreciating assets they can't shift) and/or that liquidity is a problem they're hoping will prove to be short term...
Slightly different.
When the channel knows there is surplus inventory they can and will twist your arm into discounting it. When the channel knows that WYSYG they have considerably less leverage to force you to discount.
So they buy 115$ worth of wafers to make chips from but will sell them at near list price. That is better then buying 500M of chips and selling them for 115$ and establishing a "permanently discounted" low price in the channel so they expect that you will continue to do so with your next chips and so on.
Long overdue if you ask me. They should have done that long ago. Disclaimer - I have managed to use the glut time to get a whole raft of HP Microservers for development as well as a few Fusion based desktops and laptops for test/day to day work. The discounts of the last half a year were beyond stupid. Laptops were being discounted into ~210£ territory and Microservers down to ~80£. That had to stop - it was unsustainable.
...because, in the field of MPUs, at this point in time, there are two distinct growth curves to be considered; those of application processing and storage processing.
Intel's high-performance chips are ideally suited to application processing because application processing is still largely linear and intrinsically unsuited to parallel processing. AMD's new low-power chips, on the other hand, are ideally suited to parallel processing jobs like handling storage.
Right now, and until there's a big breakthrough in AI, the growth in demand for application processing will decline whilst the demand for storage processing will grow.
In this context, it's interesting that both companies are heavily promoting future low-power MPUs; Intel with its Atom and AMD with its ARM Opterons, and what's significant here is x86. x86 is ubiquitous and the architecture that everything else is compared with but there's no way that x86 will be the architecture in use in 20 years time; the problem for Intel is that they only want do x86, something that will eventually be supplanted.
AMD had a contract with GloFo to buy $500M of chips in Q4. With the world wide economic meltdown, they elected to buy out of the contract for $320M plus processing fees, spread over 2013 instead of spending $500M in Q4 of 2012. This allows them to bring newer chip designs forward for '13 and not have $500M in chips that may be difficult to sell when the new models are released in '13. It also aids cash flow which is important in an economic depression.
AMD is going thru a complete overhaul and many things will change. If Rory Read gets it right then AMD's future will be even brighter. If he doesn't get it right, his replacement will have even more work to do.
You people must be new to Biz... AMD like other GloFo customers executes a contract a year or more in advance to purchase X amount of chip wafers per year and a specific quantitiy per calendar or fiscal quarter. In normal times when there isn't a worldwide economic meltdown, consumers of chip wafers can reasonably predict how many wafers they will need over the coming year. In the current economic environment it's difficult for most businesses to predict their sales for the next month, let alone a quarter.
In AMD's and many other company's situation, sales are down and thus it is economically better to buy out one quarter's chip wafer production and save $500 million dollars in cash and then spread that charge over 2013 when revenue should increase via new products that are scheduled for sale under existing contracts with PC makers.
Couldn't they just move some of their chipset and gpu designs over to GloFo to use more capacity in Dresden and Singapore? I know they couldn't produce enough chips a little over a year ago when demand was high, but spread over all GLOBALFOUNDIRES plants including the soon opened NY plant, just outside of Albany which makes it three 300mm plants in all. It's not a solution to all their problems but it's starting to feel like they need to make up their minds, their CPU, APU and ARM chips will be on GloFo process. Many already design against them. Feels like they need the flexibility to fab some of their other products there when they need to, capacity is available or whatever.
At least AMD see's that Piledriver plus AMD gpu isn't good enough for the mainstream notebook market where they really need to get a foothold in. Hopefully they use the cash flow for development for products that can compete we will see soon enough. The books doesn't like that quiet awful yet.
The answer to your question is: NO
You don't just move some chip production from TSMC to GloFo or other GloFo locations using a different node or Fab process. These companies use different chip wafer designs and to switch would take 12-18 months of redesign of a CPU/GPU by AMD to switch Fabs.
AMD's Trinity laptop chips are excellent and the GPU performance far exceeds anything Intel has offered to date. AMD is not having issues selling Trinity laptop, Trinity desktop, Vishera or the new Piledriver based Opterons - even in a terrible world wide economic depression.
AMD is positioning itself for future growth and streamlining product processes to be more cost efficient and able to adjust faster to market dynamics which are going to be changing more quickly and more frequently over the next 5-10 years.