I had a bit of an “aha!” moment yesterday at the MEPTEC 2.5D, 3D and Beyond Conference, with regard to the whole supply chain conundrum that seems continues to be part of the 3D commercialization hold-up. After listening to Sunil Patel of GlobalFoundries and Rich Rice of ASE discuss their respective companies’ game plans, it hit me. Perhaps there IS more than one way to skin this particular cat. (God – that’s an awful cliché isn’t it? It’s much harsher in text form than verbal… but I digress).
I don’t know about you, but all along I was waiting for one particular business model to appear out of the mist that will answer all the issues of who does what when, and who owns the liability, etc. But it turns out what’s more likely to happen is that there will be more than one answer to what Rice called the TSV Puzzle, which extends to both 2.5D (TSV silicon interposer) configurations as well as true 3D.
Herb Reiter, who leads up the GSA’s 3D EDA Workgroup, spoke historically of the time it took to establish a 2D ecosystem with a “highly disintegrated supply chain.” “Now that we we’ve got 2D figured out, we’re changing it.” He quipped. It’s not good enough to have a product, we need an application around it, as well as an infrastructure, an ecosystem and open standards to make it all work in collaboration. But does that mean we need one model? Can there be several that function differently made up of various collaborations? There are already a number of collaborations going on around the world with various supply chain approaches. Who’s to say they don’t each figure out on their own which partner is most suited to develop, for instance, the TSV interposer, or who will put in capacity for the middle of the line; or who will handle the backside thinning processes.
Patel described three scenarios he dubbed “Foundry Plus” where the entire responsibility from fab to assembly and test falls to the foundry; OSAT Plus” where the OSAT takes responsibility for middle and back-end processes; and a third party model where an EMS-type company takes ownership of gathering and assembling all the parts. Rice said he sees all three flows happening because different companies have different preferences and different capabilities. He added that in his opinion, IDMs are in an advantageous position, with the means and resources to move forward when the time is right. Some are moving already and some are just getting started; and this poses a challenge to the traditional fabless/foundry/OSAT supply chain model.
While the solution may be in the options available, Rice also expressed concern about what he defined as industry fragmentation. “It’s tough for us to build an infrastructure with this kind of fragmentation in the industry,” he said. He also noted that with regard to liability ownership, it’s his opinion that memory should always be supplied by the product owner. “If someone has their logo on top of that product, they should be responsible for what’s inside the product.” he noted.
In the end, Rice says companies will have to rely on their relationships with the customer. As different fabless/foundries/OSATS collaborations are promoting different models, might that ultimately be the main differentiator? Ultimately, the customer will decide on the business model that works for their product and go with that. Is it a bad thing to have choices in this regards? Definitely food for thought. ~ F.v.T.
PS: You can get a thorough synopsis of this event by reading Ira Feldman’s blog post here.