Sony to leave three-way microchip development pact with NEC and Toshiba

[one]

Another revision for the 45nm process development for Sony, and Tosh/NEC have to seek a new joint to reduce the increased cost burden.
http://www.reuters.com/article/technologyNews/idUST27285720070407

TOKYO (Reuters) - Sony Corp (6758.T: Quote, Profile, Research) said on Saturday it has no plan to pursue further development of advanced microchips in a three-way pact with NEC Electronics and Toshiba, in a potential blow to the Japanese chip sector's bid to join hands to better compete with overseas rivals.

Sony, Toshiba Corp (6502.T: Quote, Profile, Research) and NEC Electronics Corp (6723.T: Quote, Profile, Research) said in February 2006 they would co-develop technologies to make system chips with circuitry width of 45 nanometers, sharing hefty costs and pooling technological expertise.

They announced in December they had developed technologies for volume production of the advanced chips, and the contract between the three expired in March.

 

[Carl B]

Great find One - it's hard to interpret; Sony's disbanding the alliance, but does the associated December announcement more or less mean 'mission accomplished' in terms of researching the 45nm bulk process, and they're now going it alone... or is this a sign that they are indeed scaling back from internal fab expansion... or is it a sign that they are content enough in their IBM process alliance (of which Toshiba is also a part)?

EDIT: After reading the full article, seems to be more along the lines of the scaling back. As previously I question the wisdom of ceding ground in this space - what seems like a good idea today might be a source of regret tomorrow - but it is what it is.

 

[Deusp]

Probably a rational move. Only Intel, IBM, AMD, and the foundry companies really have the ability to make decent 45nm chips a few years from now. Sony and co. have very little need for such advanced fab technology in-house and should out-source to the foundries and IBM anyways.

 

[Carl B]

Probably a rational move. Only Intel, IBM, AMD, and the foundry companies really have the ability to make decent 45nm chips a few years from now. Sony and co. have very little need for such advanced fab technology in-house and should out-source to the foundries and IBM anyways.

Deusp your take on this is a little simplistic - of those stated companies, truthfully only Intel has what it takes to go it alone. The rest of them are all in development alliances, and in that regard Sony is no better or worse off than the rest. IBM is the research engine for a lot of these alliances, but their R&D funds would be greatly diminished if not for the partners that rely on the relationship. In terms of the fabs themselves, they are simply a capital-intensive proposition. Sony is right there with IBM and AMD in getting their 65nm process up for example, so I don't think that if they wanted 45nm would come any later either - rather it seems that Sony is hesitant to pull the trigger on the investment required.

EDIT: For more of a visual aid, see here two different Top-25 lists, one with foundries included (IC Insights) and one without (iSuppli); you'll note that Sony is actually larger than several prominent firms, and not too much smaller than a player like AMD ultimately (and in fact larger by IC's estimates). Substantially larger than IBM themselves in terms of silicon sales...

Anyway, but this is just how things seem to be trending these days - Texas Instruments recently announced that past 45nm it was all outsourcing for them, and this came as a shock to the industry. We're truly at a point where soon it will be Intel, maybe Samsung also, vs everyone else as a collective research front. The more players drop out of active participation, the more absolute Intel's victory can be if they ever manage to get a full generation spread in between themselves and everyone else. I for one would like to see Sony (and TI, and every other big player) stay engaged.

 

[avaya]

I would have guessed that the best way to deal with the current costs are JV’s. So this move is highly surprising. The horror-story scenario for the industry looks like coming true at the moment, a very oligopolistic marketplace looks set to arise with just a few huge players (best case) or Intel reaches Microsoft/GE levels where only the US Govt/EU/OEDC governments could threaten their existence. An effective monopsony is not out of the question either.

Consolidation amongst the Japanese CE’s could happen to offset this, but if Sony are pulling out of JV’s what signal does that send? Nippon Electric and Toshiba must be scrambling for a partner – who will dig? MEI?

Same in the US, could IBM end up with AMD?

 

[Shifty Geezer]

Consolidation amongst the Japanese CE’s could happen to offset this, but if Sony are pulling out of JV’s what signal does that send?

They're not pulling out of joint ventures. They've pulled out of one venture focussed on fabrication technology, which will be relevant a few years from now. Then they'll have to outsource their chip production - chips developed in partnership with other companies. I can only guess Sony see the economics of running and developing their on fabrication facilities as inefficient compared to outsourcing. I find that surprising and woud have thought chips at cost would be better than chips with profit to 3rd parties. Those fab experts must just be that danged efficient at what they do, it's not in Sony's favour. Or at least, it's not clear to Sony to be in the favour.

Or maybe it's just a matter of a financial break? Leave investment in 45 nm fabrication, save millions now to offset expsenive endeavours like PS3, take a hit on profits when using 45 nm parts (will they be content to use 65nm Cell for some years in CE? Perhaps 45nm isn't on their roadmap because costs will be low enough??), and maybe pick up fabrication tech again when they've more moey to spare. Or leave it be, because by that point we're getting near the absolute limit and will need whole new technologies I believe.

 

[Kryton]

I would have guessed that the best way to deal with the current costs are JV’s. So this move is highly surprising. The horror-story scenario for the industry looks like coming true at the moment, a very oligopolistic marketplace looks set to arise with just a few huge players (best case) or Intel reaches Microsoft/GE levels where only the US Govt/EU/OEDC governments could threaten their existence. An effective monopsony is not out of the question either.

Consolidation amongst the Japanese CE’s could happen to offset this, but if Sony are pulling out of JV’s what signal does that send? Nippon Electric and Toshiba must be scrambling for a partner – who will dig? MEI?

Same in the US, could IBM end up with AMD?

I doubt you'll see an oogopoly. It's a case of best performance/fabrication wins, and someone will always want that for their crazy new idea (64Mb L3 cache!!!11). If you can provide that, you are likely ahead of the game and able to build further off your research. So everyone else must catch up, then someone else gets the chance to make a breakthrough. There may be partnerships, but the competition for new business should prevent any collusion.

 

[Carl B]

I would have guessed that the best way to deal with the current costs are JV’s. So this move is highly surprising. The horror-story scenario for the industry looks like coming true at the moment, a very oligopolistic marketplace looks set to arise with just a few huge players (best case) or Intel reaches Microsoft/GE levels where only the US Govt/EU/OEDC governments could threaten their existence. An effective monopsony is not out of the question either.

Consolidation amongst the Japanese CE’s could happen to offset this, but if Sony are pulling out of JV’s what signal does that send? Nippon Electric and Toshiba must be scrambling for a partner – who will dig? MEI?

Same in the US, could IBM end up with AMD?

Well, the trend among Japanese CE companies has already seemed to favor their rolling semi operations into collective efforts, such as Elpida and Renesas. The Sony/Toshiba partnership seemed a strong one (hell they have joint-fabs), and NEC in the mix just made it all the stronger. It wasn't an all-out JV like the above cases, but it was significant in its own right.

Stringer, and through him Nakagawa, are looking to generate increased profits at Sony semi, but I think this is a dangerous path to walk; you have to see the forest through the trees, and realize where slim margins at semi can benefit the company elsewhere. Once you leave the process development road, you are never to return in a competetive state. They'll save billions in capital expenditure, but the price may be high as they'll be captive to external capacity and pricing concerns - and ironically as more frms go down that path, those concerns only increase for everyone. I can understand why the transition recently to 65nm would make anyone groan at the thought of having to repeat it for 45nm, but I don't think running and hiding in the arms of TSMC is the answer.

Toshiba, Samsung, and Matsushita will continue their own internal development afterall, and forgetting about consoles entirely, Sony will be taking a risk in the CE space by banking on internal design alone as being in their favor and differentiated enough to enable market-supported price-premiums on their goods vs those of their otherwise more 'covered' competitors.

I personally think Stringer's going too far in looking to achieve his 5% corporate margin goal, but more than that, for every large player like Sony and TI that drops out of fabrication R&D, that's one less solid player for the industry in general to draw upon to further overall advancement. Sony may (or may not) ultimately benefit from this on 45nm, but come 32nm and 22nm, there's going to be a difficulty of transition that's absolutely going to put a lot of pressure on the outsource crowd if the foundries have trouble keeping up. Not that you can bank on your own efforts in that scenario either, but there at least the foundries can serve as a safety net, whereas in the reverse no one's building out a fab on the fly. Of course the feared expense of those nodes is exactly why companies outside of Intel are getting squeamish to begin with.

 

[Deusp]

Deusp your take on this is a little simplistic - of those stated companies, truthfully only Intel has what it takes to go it alone. The rest of them are all in development alliances, and in that regard Sony is no better or worse off than the rest. IBM is the research engine for a lot of these alliances, but their R&D funds would be greatly diminished if not for the partners that rely on the relationship. In terms of the fabs themselves, they are simply a capital-intensive proposition. Sony is right there with IBM and AMD in getting their 65nm process up for example, so I don't think that if they wanted 45nm would come any later either - rather it seems that Sony is hesitant to pull the trigger on the investment required.

EDIT: For more of a visual aid, see here two different Top-25 lists, one with foundries included (IC Insights) and one without (iSuppli); you'll note that Sony is actually larger than several prominent firms, and not too much smaller than a player like AMD ultimately (and in fact larger by IC's estimates). Substantially larger than IBM themselves in terms of silicon sales...

Anyway, but this is just how things seem to be trending these days - Texas Instruments recently announced that past 45nm it was all outsourcing for them, and this came as a shock to the industry. We're truly at a point where soon it will be Intel, maybe Samsung also, vs everyone else as a collective research front. The more players drop out of active participation, the more absolute Intel's victory can be if they ever manage to get a full generation spread in between themselves and everyone else. I for one would like to see Sony (and TI, and every other big player) stay engaged.

My point was that only AMD, IBM, Intel had the justifiable need to have advanced fabs. Other companies make lots of lower end stuff and thus don't need those high end fabs.

 

[Carl B]

My point was that only AMD, IBM, Intel had the justifiable need to have advanced fabs. Other companies make lots of lower end stuff and thus don't need those high end fabs.

That's just simply not true.

CMOS sensors, cell phone chipsets, memory modules, integrated system ICs... all exist in a highly competitive space where smaller process nodes lead to higher margins, lower power consumption, smaller footprints, etc etc... and these have very real effects in product differentiation.

I guess another way of approaching your post would be to ask, what is the "justifiable need" you see Intel, AMD, and IBM as having that the other manufacturers seemingly don't?

 

[Capeta]

Low-end chips can be outsourced to foundries who have different nodes, they don't need to be fabbed inhouse with state of the art fab tech. As an example chips inside the upcoming iPhone don't need to be fabbed on 65nm fabs, they will more likely be fabbed on 90nm process. I'm willing to bet the majority of CE ICs are fabbed on 90nm or larger nodes.

 

[Deusp]

That's just simply not true.

CMOS sensors, cell phone chipsets, memory modules, integrated system ICs... all exist in a highly competitive space where smaller process nodes lead to higher margins, lower power consumption, smaller footprints, etc etc... and these have very real effects in product differentiation.

Like Capeta said, all of these can be outsourced and usually don't need the latest and greatest fabs. Memory requires a special fab different from the ones Sony needs and hence don't factor into the problem.

I guess another way of approaching your post would be to ask, what is the "justifiable need" you see Intel, AMD, and IBM as having that the other manufacturers seemingly don't?

If you don't need world-beating performance, or in IBM's case world-beat fab technology, you don't need your own fab. Look at it this way: A fab cost >$3 billion, and needs several hundred million dollars of R&D to get a worthwhile product on it. And the fab is only state of the art for 2 years or less, so the cycle repeats every other year. Not many companies can justify owning one of these in-house, especially considering how much performance GPUs can get using exclusively outsourced fabs.

 

[Carl B]

If you don't need world-beating performance, or in IBM's case world-beat fab technology, you don't need your own fab. Look at it this way: A fab cost >$3 billion, and needs several hundred million dollars of R&D to get a worthwhile product on it. And the fab is only state of the art for 2 years or less, so the cycle repeats every other year. Not many companies can justify owning one of these in-house, especially considering how much performance GPUs can get using exclusively outsourced fabs.

Any company that has the internal and/or external demand to occupy the capacity of said fab can 'justify' it, and its as simple as that. When you have a higher chip volume to begin with than a foundry like UMC, how can you say that UMC is someone you should be outsourcing to - and your own capabilities? - well who cares...

You and Capeta are viewing things in way too PC-centric a manner, as if the rest of the world of semiconductors is 'meh' because we don't read about it on tech sites every day. You haven't answered my question yet Deusp as to why companies like AMD and IBM need advanced fabs, so I'm still waiting on that - personally I don't think there's any logic that can be applied there that will be exclusive to those companies. Why shouldn't AMD just outsource everything to Chartered and stop making their own fabs? Indeed, it will be interesting to see what node the chips inside the iPhone do indeed use - Capeta guesses 90nm... perhaps. But it's a lot more likely to be 65nm than 130nm for instance, agreed?

There are arguments to be made for outsourcing vs insourcing, but the ones centering on "they don't need the capacity" are not it. Rather, capital investment dollars spent on product/chip R&D, other investments that might yield a higher return over time, etc etc... it's about alternatives to the ways that $3 billion can be spent, not because that $3 billion itself would prove a negative investment over time - because there's no question that it wouldn't. And this is an ongoing dialogue at Sony no doubt, with pros and cons on each side. Obviously Nakagawa feels this is the way to go to meet the corporate margin goals, but his predecessor was of the mind to go ahead with internal 45nm buildout. Is one crazy and the other sane? They just both are pinning their hopes on different philosophies.

 

[Deusp]

Any company that has the internal and/or external demand to occupy the capacity of said fab can 'justify' it, and its as simple as that. When you have a higher chip volume to begin with than a foundry like UMC, how can you say that UMC is someone you should be outsourcing to - and your own capabilities? - well who cares...

Even if you could profitably make chips with your own $3 billion fab, there's the opportunity costs to deal with. Can you still say that you're better off make slightly better chips than having that $3 billion dollars spent elsewhere? Excluding the foundries, probably only AMD, IBM and Intel can say yes. Even among these three, only Intel has a genuine need for them. AMD only needs them to be competitive with Intel and IBM only needs them for the tech in order to facilitate relationships with other companies.

You and Capeta are viewing things in way too PC-centric a manner, as if the rest of the world of semiconductors is 'meh' because we don't read about it on tech sites every day. You haven't answered my question yet Deusp as to why companies like AMD and IBM need advanced fabs, so I'm still waiting on that - personally I don't think there's any logic that can be applied there that will be exclusive to those companies. Why shouldn't AMD just outsource everything to Chartered and stop making their own fabs? Indeed, it will be interesting to see what node the chips inside the iPhone do indeed use - Capeta guesses 90nm... perhaps. But it's a lot more likely to be 65nm than 130nm for instance, agreed?

AMD is a special case because Intel will crush them with fab technology if they didn't have their own fabs. For nearly the entire rest of the industry, a foundry will suffice. Very few companies have the R&D to even make a chip on the latest process, nor do the chips they make need to be so advanced. For most companies, it's a waste of billions of dollars.

There are arguments to be made for outsourcing vs insourcing, but the ones centering on "they don't need the capacity" are not it. Rather, capital investment dollars spent on product/chip R&D, other investments that might yield a higher return over time, etc etc... it's about alternatives to the ways that $3 billion can be spent, not because that $3 billion itself would prove a negative investment over time - because there's no question that it wouldn't. And this is an ongoing dialogue at Sony no doubt, with pros and cons on each side. Obviously Nakagawa feels this is the way to go to meet the corporate margin goals, but his predecessor was of the mind to go ahead with internal 45nm buildout. Is one crazy and the other sane? They just both are pinning their hopes on different philosophies.

This I agree. I never claimed they don't need to capacity, but other reasons, similar to what you've laid out here. With Sony, the most advanced stuff they will make is the Cell processor, which probably will become a low-end device by the 45nm process. Hence, there's very little need to have their own fabs by then.

 

[Carl B]

Deusp your Intel/AMD/IBM view of the universe is such that this conversation cannot progress; you basically consign 80% of the semiconductor industry into some sort of basket of non-competition. Of that list of Top-25 semi companies, which of those are pursuing internal 45nm fab efforts? Certainly a lot more than your chosen 3 champions.

45nm production is not about making 'advanced' chips, it is about putting more transistors into a smaller space. It is a fundamentally different concept, and it benefits chips of any size and/or complexity. The reason you associate it with 'advanced' chips is because without advanced processes', such chips aren't feasible at all. You say that AMD risks elimination at Intel's hands if it doesn't pursue its own fabs; how is it different for a CE company that falls behind its own competitors?

Here's an article that spoke to the risks TI was taking on when they announced their own shift in January: http://www.electronicsweekly.com/bl...og/2007/01/nxp-ti-stmicro-and-the-end-of.html

 

[Deusp]

Deusp your Intel/AMD/IBM view of the universe is such that this conversation cannot progress; you basically consign 80% of the semiconductor industry into some sort of basket of non-competition. Of that list of Top-25 semi companies, which of those are pursuing internal 45nm fab efforts? Certainly a lot more than your chosen 3 champions.

45nm production is not about making 'advanced' chips, it is about putting more transistors into a smaller space. It is a fundamentally different concept, and it benefits chips of any size and/or complexity. The reason you associate it with 'advanced' chips is because without advanced processes', such chips aren't feasible at all. You say that AMD risks elimination at Intel's hands if it doesn't pursue its own fabs; how is it different for a CE company that falls behind its own competitors?

My point is not an argument against Moore's Law, but the practically of going internal. The foundries are more than capable of taking to former forward, but only three companies have need of the latter. Today, the cost of simply developing a chip that utilizes the latest and greatest fab process is so high that the design rule of the process is only one concern among many. It's fully possible to design a chip that's better than its competitor even if its on the same or lesser process by the same foundry (see R300 vs. NV30, possibly G80 vs. R600). For the majority of CE companies, especially if they are making high-end chips, the cost of chip development far outweigh the relatively minor benefit of having a better process. Since having a internal fab is so expensive, going with foundries saves a lot of money and has very few downsides. Only AMD's case is special, since Intel pushes its fabs to the break point, any competitor needs similarly advanced fabs just to keep up.

Here's an article that spoke to the risks TI was taking on when they announced their own shift in January: http://www.electronicsweekly.com/bl...og/2007/01/nxp-ti-stmicro-and-the-end-of.html

It seems to me he's speculating on potential downsides of going with foundries, which has yet to be realized. Could foundries raise prices in the future? It is not a certainty like the article claims assuming capitalism still works. Could chips all be nearly identical in terms of performance among all chip makers in a single market? A possibility, but historically this has been proven incorrect. The quality of the design of the chip are just as big of an influence as the process itself.

 

[Panajev2001a]

Like Capeta said, all of these can be outsourced and usually don't need the latest and greatest fabs. Memory requires a special fab different from the ones Sony needs and hence don't factor into the problem.



If you don't need world-beating performance, or in IBM's case world-beat fab technology, you don't need your own fab. Look at it this way: A fab cost >$3 billion, and needs several hundred million dollars of R&D to get a worthwhile product on it. And the fab is only state of the art for 2 years or less, so the cycle repeats every other year. Not many companies can justify owning one of these in-house, especially considering how much performance GPUs can get using exclusively outsourced fabs.

Good reasons, it shows how Samsung is so unsuccessful with their vertically integrated strategy... oh wait, maybe GOOD management is indeed the KEY here!

 

[Deusp]

Good reasons, it shows how Samsung is so unsuccessful with their vertically integrated strategy... oh wait, maybe GOOD management is indeed the KEY here!

We've ignored Samsung here since they make memory chips and not ASIC. Unlike ASICs, designing memory chip cost very little and hence process is nearly everything. They also need specialized fabs too. Samsung can justify it's internal fabs but most CE companies can't.

 

[Panajev2001a]

This I agree. I never claimed they don't need to capacity, but other reasons, similar to what you've laid out here. With Sony, the most advanced stuff they will make is the Cell processor, which probably will become a low-end device by the 45nm process. Hence, there's very little need to have their own fabs by then.

Cost control, close integration between chip design teams and circuit engineers and other people involved with manufacturing process R&D, IP/trade secrets protection, competition with vertically integrated giants (surely buying all chips and components from Samsung and Toshiba will not make Sony's ability to compete against them that much stronger... quite the contrary IMHO), volume of production prediction and adjustments...

we need more and wider alliances as it is going to be difficult that foundries like UMC and TSMC will always be able to advance AND expand and expand as other players (that also sell at a high premium what they make themselves, like Intel and IBM and AMD...) do as it will be a big risk for them... "will we have enough players ordering chips from us and how much volume should we target?" not as easy of a question as some people might think...

Also, IMHO, if someone thinking all we have said here matters more than Stringer moving Sony to a Content+IP+Software and heavy outsourcing driven company then we might as well believe that the #1 reason for switching Machintosh PC's from PowerPC to x86 was performance as Steve Jobs said...

 

[Deusp]

Cost control, close integration between chip design teams and circuit engineers and other people involved with manufacturing process R&D, IP/trade secrets protection, competition with vertically integrated giants (surely buying all chips and components from Samsung and Toshiba will not make Sony's ability to compete against them that much stronger... quite the contrary IMHO), volume of production prediction and adjustments...

These use to be the reasons to go internal. Problem is, all of these reasons are going by the wayside. Internal fabs don't really save money anymore, nor does need for that last mile of performance a big deal either. No one would go with foundries if trade secrets are at a major risk of being leaked. Competition from vertically integrated companies are also a non-threat, since Intel is the only true vertically integrated company today (excluding Samsung).

we need more and wider alliances as it is going to be difficult that foundries like UMC and TSMC will always be able to advance AND expand and expand as other players (that also sell at a high premium what they make themselves, like Intel and IBM and AMD...) do as it will be a big risk for them... "will we have enough players ordering chips from us and how much volume should we target?" not as easy of a question as some people might think...

That's a vastly overplayed threat. Foundries are not bad at what they do, and it's much more likely that an internal fab company will run into trouble with its own fab than a foundry would.

Also, IMHO, if someone thinking all we have said here matters more than Stringer moving Sony to a Content+IP+Software and heavy outsourcing driven company then we might as well believe that the #1 reason for switching Machintosh PC's from PowerPC to x86 was performance as Steve Jobs said...

You've lost me here.