Carl, what are your thoughts on this: Is this an "Intel" casualty? Intel has been absolutely crushing the competition in regards to process node jumps. Back in 2004 there was a lot of talk from certain informed posters that Sony was right there, if not ahead, with 90nm mass production and was headed on the way to beat them on 65nm. Retrospectively none of that happened and I get the suspecion that Sony would have preferred Cell on 65nm for launch for many reasons. My initial reaction is that as TSMC et al and such are approaching parity in process node technology the cost to have it fabbed elsewhere may be a bigger savings than pushing the process nodes as quickly as necessary to keep pace, and in turn turning over that fab space to chips that could benefit from the advanced manufacturing processes but not quite as sensative as to process delays (like a chip like Cell could be).
I do remember those 65nm thoughts of yore. They stemmed from Toshiba/Sony advancements and claims on the eDRAM front and their joint CMOS process made public through press releases in late late 2002 and 2003. Back then, their (optimistic) roadmaps indicated a move to 90nm in 2004, and 65nm in 2005. Obviously that didn't happen, but I don't fault them for their naivity at that point in time. It was pre-90nm rollout for everyone, and with 90nm came the advent of leakage current as a primary concern, and numerous difficulties with getting the process lines up to speed relative to past process transitions - those sober realities that have set in since color processor design quite heavily in the present.
Sony succeeded in attaining 90nm in 2004 on their and Toshiba's CMOS4 process, but I think the difficulties of 65nm began to make themselves apparent shortly thereafter. Remember also that Sony's own Nagasaki fab was/is an SOI line, and has a lot more in common with IBM fab tech than their traditional Sony/Toshiba CMOS bulk process. I think in that context, Sony's ramp at Nagasaki, - though well behind schedule - is still as good as that of their development partner IBM... and what can you do? Kutaragi knew the score in May 2005 when he indicated that they would stick to 90nm for launch; indeed the prior plan
had been for 65nm. But IBM, AMD, Toshiba... everyone has found 65nm harder to achieve than anticipated.
Intel.
Ultimately, I don't view Intel's clear advantage in fabrication expertise as competition to Sony's own efforts; truthfully, I felt the closest they ever came to that was Intel's abortive LCoS effort in the HDTV space. So I think Sony is sitting pretty in terms of process tech relative to their own direct competitors - granted that's not to say that I don't think the outsource model has points of merit. But just that when you're on the level with IBM, no reason at all for them to be considered 'behind' - Intel is just so clearly far out ahead.
If this is the approach they are taking I wonder how they will keep pace in cutting edge fab technology. If they are outsourcing maybe they won't need to, and being slightly behind the curve could result in significant savings. But once you back off the pace it doesn't seem like it would be very easy to jump back in the game, at least to compete with Intel. On the other hand hadn't Sony announced a number of alliances with NEC, Toshiba, IBM, and AMD as of late in regards to process technologies?
Obviously they can develop Cell2 or whatever they plan to put in the PS4, as well as RSX2 in unison with NV, in conjunction with a company like TSMC and stay relatively up to date and on pace in regards to process technology and then use their own fabs to make a major play to capture huge segments of other markets (like sensors), and in the end coming out on top financially.
Their alliance with Toshiba is ancient and longstanding; NEC is a new addition to their alliance over on the bulk process side. Sony and Toshiba then also have an alliance separately with IBM for both the Cell architecture and related process technologies; IBM has its own separate alliance with AMD and Chartered as well, so lots of threes it seems.
Sony in my opinion can stay as far out ahead as they want on in-house process technology; they run with the big boys and as long as they're willing to invest, they can stay there. Again, I'm treating Intel as above-the-fray in these terms - they simply are in another class. I think it makes
sense to invest in fabs, as I think you never really have a sense of where your R&D is leading without having physical production tied to it. But then again, fab houses like TSMC are attaining such a scale that even if it costs more per chip ultimately, the hedge on risk and the ability to free billions in cap-ex for other purposes definitely in my mind warrants attention on the part of Sony.
It'll be interesting to see what they ultimately decide to do in this regard. I
hope as a hardware company and innovator that Sony chooses to expand on their fab efforts, but Stringer has a duty to the shareholders in the immediate-term as well; it's ultimately up to him and his circle to determine whether new fabs, new acquisitions, research elsewhere, investments, or whatever else will lead to the highest return on investment. In favor of in-sourcing, I would like to highlight that the fact that creating their own 45nm line is implicitly still on the table hints to the clear benefits that having their own fab capacity brings; the question they're asking of themselves today is simply the classic opportunity cost question any rational being or business needs to ask of themselves when limited resources need be commited to achieve few goals from a range of many options.
