You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Or, in the words of Roger Kay, how do IBM lemons become GF lemonade?
I can see a bridge for GF to reach emerging giants like Huawei (and HiSilicon), who have always had strong ties to IBM. Will ARM-or-Power based servers be in the play, with HiSi for Huawei? And then on the IoT client side, perhaps with the strong RF IPs, GF can build a kind of IoT sweet spot like TSMC/UMC/ARM's 55nm LP?
Might there be a Global UniChip kid of subsidiary/affiliate for GF, by way of generating higher-end demand and compete effectively in APAC?
Very interesting question. Design enablement is critical for the foundry business. In fact, one key component missing from the Intel Custom Foundry and Samsung Foundry business model is a dedicated design services business. UMC has Faraday, SMIC has Verisilicon, TSMC has GUC, and everybody has eSilicon since they are the largest independent ASIC provider.
GlobalFoundries has Invecas, which is a joint venture launched last year. I had lunch with Invecas founder Dasaradha R. Gude (DG) a while back and found him to be very engaging. It would be worth your time to Google him and read his story. I also saw him at the GSA awards last year. He won "Start-up of the year" or something like that.
In my opinion IBM Semiconductor has three things to offer that stand out for design enablement: ASIC Business, IP, and EDA tools. From what I understand Invecas will be taking control over all three which makes for a powerful combination. I know the IBM SRAM team from a previous engagement and they are now at Invecas. In fact John Barth (26 year IBMer) is listed as VP of Memory Design at Invecas. Everyone in SRAM knows John and SRAM is a critical part of the semiconductor IP business, absolutely.
I will do some more digging here...... It really is an interesting story that is evolving pretty quickly.
Great post, Dan. I have heard of Invecas but wasn't too clear about their business model in association w/ GF.
The interesting twist to the ASIC model here is, what if the foundry OWNS (rather than licenses) the IP collaterals. How would the cost accounting work differently? And, since GF owns not only the manufacturing and design services but also IPs and tools - unprecedented for a foundry - how would this fact affect the IC product ownership? Say, can a deal be structured to do an ASIC, with a pre-agreement that this will in time evolve into an ASSP of which GF has a stake? Might this open the possibility of a foundry/IDM hybrid model? Will this level of flexibility be positive or negative for system house customers?
Of course one can say that, hey, K.I.S.S., folks; where the IPs come from can be treated separately, and let's not mess with the pure-play model. Fair enough. But GF does have an opportunity to structure something bolder. For example, with IoT, it could be better positioned than TSMC/UMC (55LP) and SMIC. GF can offer something cheaper than a full ASIC yet more customizable than an SP, by selecting assembling IPs and do a partial frontend design for a certain application domain (automotive, etc), making it easier and faster for customers to complete the netlist, then back to GF for the backend design and everything else.
Serend and Daniel: strong interaction between the design services and design-kit teams sounds a great idea to me. I think it's why IBM's design kits were (from my viewpoint) vastly superior to TSMC's. I think GF could be onto a winner here if they can lose the built-in over-conservatism of IBM's foundry model and also leverage volume to get the pricing fully competitive (and not just for their IBM-derived processes either).
It was all part of the old school ASIC model. IBM used internally developed IP as did the Japanese ASIC companies. eSilcion and Open-Silicon came out as IP agnostic relying on commercial IP.
Today it is more of a hybrid model. Faraday develops IP for UMC but also uses commercial IP when a customer requires it. GUC uses TSMC IP and commercial IP. eSilicon has internal IP now but uses mostly commercial IP. Same with VeriSilicon.
Hopefully GF can leverage Invecas for design services, IP, and design kits based on the IBM Semiconductor assets to increase their competitive advantage.
