Will GlobalFoundries Inc. or some other manifestation of the oil-rich state of Abu Dhabi, be buying the chip R&D and business interests of IBM any time soon? And what would be the price?
Future Horizons Ltd. (Sevenoaks, England) has said it included the comment "We assume GlobalFoundries will purchase IBM’s semiconductor division and that Hynix/Micron will buy up the remaining smaller memory firms," in the conclusions of a report prepared by Future Horizons and Decision SA for the European Commission on the future of 450-mm wafer processing in Europe.
The idea that GlobalFoundries – a foundry owned by an Abu Dhabi sovereign wealth vehicle – could buy significant chunk of American technology leadership is likely to be controversial.
When asked why he thought the move was a valid assumption, Future Horizons analyst Mike Bryant, said he had heard rumors that discussions are taking place from enough reliable sources to consider the outcome likely.
And the move would make sense in some ways. It is in-line with IBM's strategic retreat away from hardware and towards software and consultancy and GlobalFoundries looks like a natural inheritor of IBM chip interests in New York state.
Indeed IBM's role as the paternalistic overseer of the Common Platform Alliance on process technology, alongside collaborators GlobalFoundries and Samsung Electronics Co. Ltd., looks increasingly anachronistic.
It was over a decade ago that IBM established the model of sharing R&D costs to develop chip technology. IBM is still a manufacturer of chips at East Fishkill, New York, and contributes a great deal of advanced research to various semiconductor-related consortia and initiatives it is involved in. Since it started on the 90-nm node more than a decade ago it has fostered out a number of manufacturing processes and innovations including SOI. It has also been a regular source of presentations at leading conferences such as IEDM.
But IBM does not sell chips on the open market and since it began its semiconductor collaborations it has sold off its PC business to Lenovo, as a key part of its strategic transformation.
So why does IBM need to pay for research into manufacturing processes, extreme ultraviolet lithography, and 450-mm diameter wafers?
I think the foundry business needs a lot of capitol to survive. If you can't survive, any desire to put "america first" or "new york first" is irrelevant. A company I worked for made explicit efforts to put "eurozone first" and went out of business. I would have preferred they made survival priority #1, #2 and #3.
I am really happy to read this very detailed article. We did some analysis in our recent post that detects GF getting really close to #2 (UMC)
click here to see the graph:
"Two other sticking points might be U.S. national security and national pride."
The first half of that is probably a point worthy of more attention than given here. I'm I mistaken that IBM fabs a lot of chips destined for military equipment?
Yes, IBM is a "secure foundry" and as such are permitted to manufacture ICs for classified military use. This is a small niche business, not sure if they would spin off or kill this activity if they merged with GF.
As Peter pointed out here, I, too, always wondered about what's in it for IBM in paying for research into manufacturing processes, extreme ultraviolet lithography, and 450-mm diameter wafers... i don't think we've ever gotten straight answers from IBM on that either...
I don't believe Global can buy IBM's chip entity. If it ever happends, It should be the other way around. Just like Chartered folded under Global. Abu Dhabi can buy IBM's chip entity and Global can be folded under. That makes sense from technology R&D & management skills perspective. It's too much for Global to manage IBM's chip unit. GF lacks management skills and experience to manage such a heavy unit.
By the way IBM chip unit was rumored to be on sales sometime ago and I believe that TSMC and Samsung looked at it. The price tag was rumored to double digit $B then.
Blog Doing Math in FPGAs Tom Burke 2 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...