Sony to cut semiconductor capital expenditure; 45nm Cell in FY08/09

[Carl B]

I can't quite understand how this thread has gone in the direction of people thinking that Sony might *skip* 65nm, or that their own fabs, even if they were to outsource Cell at 45nm, would go unused. Sony's fabs will always be used to capacity, so long as there are IC's to be fabbed there. It should be noted as well that their CMOS sensors for cameras and such are indeed sold to other companies as well. Clearly at least for 65nm, Sony will be fabbing in-house at Nagasaki, and in fact has a claim on a good bit of IBM's 65nm output through their investments in the East Fishkill line.

The move to 45nm isn't anything that should be treated as something out of the ordinary; if they outsource it, they outsource it. But to say that their 'dedicated' Cell fabs will then be un-utlized is off the mark. They have a 65nm SOI line at Nagasaki - what were their plans for it anyway when Cell went 45nm? Whatever their plans were before, those plans can remain. In the meantime Sony is involved in several research alliances with Toshiba, IBM, and NEC for immersion lithography, High-K dialectrics, and the move to 45nm and 32nm... so we'll see what happens as the decisions are made going forward.

It's not an issue of outsourcing 'saving money,' because it doesn't. What it does is free up cash to spend in other, potentially more lucrative ways. I think there are strng arguments to be made from both angles, and I think Sony thinks so too, which is why I feel there will be a lot of soul-searching between now and 45nm.

 

[TheChefO]

which is why I feel there will be a lot of soul-searching between now and 45nm.

Not to mention the future is very questionable for cutting edge future process shrinks beyond 32nm. I'm sure Sony does not want to be caught investing heavily in this risky environment.

 

[avaya]

Over 66% of group revenues comes from hardware. Game divison is next highest with 12%, Pictures third, not too far behind (and catching).

Sony's been slow with software integration, that's what Stringer means when he talks about software. Their music software is a joke, and they even bloat their VAIO's with bundles of crap. Their software can no longer afford to be sub-par.

Hardware is bread and butter, they will never stop being a hardware company at heart and for a company of their size they will likely never stop fabbing their own chips to some degree or another. There are several product lines that benefit from in-house production as identified by xbd. For the big three CE's: MEI, Sony and Samsung along with Toshiba fabbing in-house is not going to be stopped any time soon.

 

[inefficient]

Informative wikipedia entry: Semiconductor sales leaders by year

http://en.wikipedia.org/wiki/Worldwide_Top_20_Semiconductor_Sales_Leaders

As you can see it is a multi billion dollar revenue stream for Sony. It's not something they are going to abandon. They just want to improve their margins.

 

[Titanio]

Forget about that Inq article - the original story is translated via the Marketwatch article linked to in the OP, complete with Sony executive statements. The Inq article highlights Cell R&D specifically, whereas all other sources indicate cap-ex in general, framed in terms of fab capacity; those are two different things, and until further clarification (if forthcoming) should be viewed as an in-house inference by the Inq. Perhaps 'One' will chime in with a more complete take on the article.

Further clarity:

Nakagawa said that Sony has no intention of stopping R&D efforts. But continuing to maintain volume production facilities is another story.

http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=197006176

There's been some poor reporting and misunderstanding of the original news on that front, but I guess it's to be expected from the likes of The Inq at least.

 

[Capeta]

I can't quite understand how this thread has gone in the direction of people thinking that Sony might *skip* 65nm, or that their own fabs, even if they were to outsource Cell at 45nm, would go unused. Sony's fabs will always be used to capacity, so long as there are IC's to be fabbed there. It should be noted as well that their CMOS sensors for cameras and such are indeed sold to other companies as well. Clearly at least for 65nm, Sony will be fabbing in-house at Nagasaki, and in fact has a claim on a good bit of IBM's 65nm output through their investments in the East Fishkill line.

The move to 45nm isn't anything that should be treated as something out of the ordinary; if they outsource it, they outsource it. But to say that their 'dedicated' Cell fabs will then be un-utlized is off the mark.

I see little evidence to suggest their fabs will be used to full capacity if they outsource. The CMOS sensors argument doesn't hold water. Where is the CMOS being fabbed now? Why would they shift it to the idle fabs? Are their CMOS sensors currently supplied constrained? Unless their ICs need more fab capacity there is no need to shift them to the dedicated CELL fabs otherwise it's just a solution looking for a problem.

 

[one]

I see little evidence to suggest their fabs will be used to full capacity if they outsource. The CMOS sensors argument doesn't hold water. Where is the CMOS being fabbed now? Why would they shift it to the idle fabs? Are their CMOS sensors currently supplied constrained? Unless their ICs need more fab capacity there is no need to shift them to the dedicated CELL fabs otherwise it's just a solution looking for a problem.

What's the source of your speculation? From the EE Times article Titanio posted above yours:

Under the new policy, Sony intends to focus on three areas—imagers, game LSIs and system-on-chip LSIs for audio/video devices. Sony has about 60 percent market share in the CCD imager market and is now reinforcing CMOS sensors as well. One hundred and fifty engineers were shifted to the imager section from audio/video sections to boost the development.

 

[Carl B]

I see little evidence to suggest their fabs will be used to full capacity if they outsource. The CMOS sensors argument doesn't hold water. Where is the CMOS being fabbed now? Why would they shift it to the idle fabs? Are their CMOS sensors currently supplied constrained? Unless their ICs need more fab capacity there is no need to shift them to the dedicated CELL fabs otherwise it's just a solution looking for a problem.

Whatever the case, you can be sure that fabbing the ICs at 65nm would be less expensive than at 130nm, and would have power benefits as well. So whatever your views on excess fab capacity and supplies, it makes all the sense in the world to utilize more advanced lines to produce these ICs if they would otherwise go under-utilized - assuming here a transition to 45nm on Cell and the 65nm lines being freed up.

And Capeta, where do you get this idea of 'dedicated Cell fabs?' - that is a fantasy of your own creation, nothing more, nothing less. No fab is 'dedicated' to any chip save in a temporary sense. A fab is a fab - it makes chips. What those chips are change throughout the life of the fab.

 

[Capeta]

Sony intends to focus on three areas—imagers, game LSIs and system-on-chip LSIs for audio/video devices. Sony has about 60 percent market share in the CCD imager market and is now reinforcing CMOS sensors as well. One hundred and fifty engineers were shifted to the imager section from audio/video sections to boost the development.

more development =! more production

And Capeta, where do you get this idea of 'dedicated Cell fabs?' - that is a fantasy of your own creation, nothing more, nothing less. No fab is 'dedicated' to any chip save in a temporary sense. A fab is a fab - it makes chips. What those chips are change throughout the life of the fab.

SONY spent more than a $billion to build the fab which was to be used solely for production of console chips, PS3, PSTwo, PSP.

Whatever the case, you can be sure that fabbing the ICs at 65nm would be less expensive than at 130nm, and would have power benefits as well. So whatever your views on excess fab capacity and supplies, it makes all the sense in the world to utilize more advanced lines to produce these ICs if they would otherwise go under-utilized - assuming here a transition to 45nm on Cell and the 65nm lines being freed up.

And I already told you why that doesn't always makes sense. You don't simply move to a smaller process if the demand isn't there. You also need to redesign a lot of the PCBs to use these new small chips. It made sense for stuff like EE+GS because the demand WAS there. I don't think demand for PSTwo or PSP will rise in the next few years. It's a solution looking for a problem. If they're supplied constrained sure, but I dont' see any evidence of that.

 

[Carl B]

more development =! more production

It's written like this: !=

SONY spent more than a $billion to build the fab which was to be used solely for production of console chips, PS3, PSTwo, PSP.

That is incorrect. I linked you to a slide presentation earlier in this thread that outlines their policies on these matters; the lines that once made gaming chips in 2000, today make CE components. And such will it always be; a cascading fab capacity favoring the more complicated designs.

 

[one]

more development =! more production

Are you serious or joking? Shifting resources to speed up and augment the production of things that make money is too basic to be in Business 101.

 

[Capeta]

Are you serious or joking? Shifting resources to speed up and augment the production of things that make money is too basic to be in Business 101.

You've never heard of companies boosting R&Development spending while not ever producing a single product? What kind of Business 101 are you talking about? Sorry but more development doesn't = more production.

 

[one]

You've never heard of companies boosting R&Development spending while not ever producing a single product? What kind of Business 101 are you talking about? Sorry but more development doesn't = more production.

Then you should show me how the imager market is shrinking now. After that, I'll hear you again.

 

[Capeta]

the lines that once made gaming chips in 2000, today make CE components. And such will it always be; a cascading fab capacity favoring the more complicated designs.

Uh yeah but this is 2007 and PS3 was just released and they're already thinking about outsourcing.

 

[Carl B]

And I already told you why that doesn't always makes sense. You don't simply move to a smaller process if the demand isn't there.

Capeta, you seem to be under the illusion that smaller process nodes are pursued only when there's demand for the chip in question. I don't know how you came to view things like that, but you're putting the cart before the horse. A smaller process node is *always* favorable to a larger one, in general terms. The only reason why 'demand' factors in, is that to justify the capital expenditures these days for advanced fab equipment, minimum volume assurances must be had. Cell was the catalyst that warranted the 65nm fab line buildout at Nagasaki, but certainly its uses will not be limited to PS3 componentry in its life. Whatever the 'demand' of the other ICs, it doesn't matter; a smaller process node has been made available to them, and it will be utilized when the capacity frees up to do so. Of that, there can be no question.

You also need to redesign a lot of the PCBs to use these new small chips. It made sense for stuff like EE+GS because the demand WAS there. I don't think demand for PSTwo or PSP will rise in the next few years. It's a solution looking for a problem. If they're supplied constrained sure, but I dont' see any evidence of that.

Again, supply-constraints have nothing to do with it, and we're not even talking about PS2 or PSP here to begin with. Redesigning the PCB for a digital camera for instance is a trivial matter, and certainly worth it - let alone for something more complicated like a console. But focusing on the camera, what you gain is the potential for a smaller form factor, higher battery life, and cheaper set of componentry. If you feel that the move to 65nm from 130nm wouldn't be acted on if the opportunity presented itself readily, then honestly, you need to view this from an expanded angle. The business of semiconductors does not favor inertia, as you suggest in your 'why bother moving' example; on the contrary, it favors movement.

 

[Capeta]

Then you should show me how the imager market is shrinking now. After that, I'll hear you again.

Or how about you show evidence that demand for SONY's CMOS imagers are increasing at a significant pace to need more fab capacity? You made the claim so back it up.

The market doesn't need to shrink for the supply to be the same.

Capeta, you seem to be under the illusion that smaller process nodes are pursued only when there's demand for the chip in question. I don't know how you came to view things like that, but you're putting the cart before the horse. A smaller process node is *always* favorable to a larger one, in general terms. The only reason why 'demand' factors in, is that to justify the capital expenditures these days for advanced fab equipment, minimum volume assurances must be had. Cell was the catalyst that warranted the 65nm fab line buildout at Nagasaki, but certainly its uses will not be limited to PS3 componentry in its life. Whatever the 'demand' of the other ICs, it doesn't matter; a smaller process node has been made available to them, and it will be utilized when the capacity frees up to do so. Of that, there can be no question.

Again, supply-constraints have nothing to do with it, and we're not even talking about PS2 or PSP here to begin with. Redesigning the PCB for a digital camera for instance is a trivial matter, and certainly worth it - let alone for something more complicated like a console. But focusing on the camera, what you gain is the potential for a smaller form factor, higher battery life, and cheaper set of componentry. If you feel that the move to 65nm from 130nm wouldn't be acted on if the opportunity presented itself readily, then honestly, you need to view this from an expanded angle. The business of semiconductors does not favor inertia, as you suggest in your 'why bother moving' example; on the contrary, it favors movement.

If that were the case every chip made nowadays would be using 90nm or 65nm to save costs...I dont' think so. Again you're just SHIFTING from one fab to another, you're still going to have idling fabs unless you have more demand. If I have a small box that's filled with stuff and move this stuff into a bigger box I'd still have an empty box unless I fill the empty one with more stuff (demand).

 

[Carl B]

Uh yeah but this is 2007 and PS3 was just released and they're already thinking about outsourcing.

They're thinking about farming out Cell production on the 45nm node, because as mentioned earlier in the thread (by myself and others), there is a cap-ex decision to be made on their part as to whether the billions might carry a higher rate of return elsewhere. But as Panajev ventures, whatever they do, I'm pretty sure they'll opt for at least some production capacity under their direct control, whether via JV or otherwise.

Two years on 65nm before a move to 45nm would be decent enough - I don't view it as "already" thinking about farming out Cell production in a fatalistic sense. They're thinking about it now, because the decisions on 45nm have to happen now; a fab build/upgrade is a multi-year project.

But, again... the point is whenever the Nagasaki fab stops building chips for the PS3, it will immediately switch over to building chips for something else. You don't retire your newer fabs, you retire your older fabs.

 

[Carl B]

If that were the case every chip made nowadays would be using 90nm to save costs...I dont' think so.

My answer to this is contained in the post you quoted. Suffice to say that any company that can afford to move production to a smaller process node, whether via capacity free-up or otherwise, does so at the earliest opportunity.

You seem a fan of Intel Capeta, so let me use them as an example for you. Chipsets are of secondary importance to them relative to their primary CPU production; CPUs are being fabbed at 65nm. Is there enough chipset "demand" to warrant leaving 130nm and going to 90nm? No, 130nm was up to the challenge on a macro level. Yet they have moved chipset production to 90nm nontheless, as it saves them money, and enhances the performance profile of such components. You use your most advanced fab equipment whenever possible, and you use it because you have it to use.

Companies that are stuck fabbing their primary products on 180nm, for example, are companies that are in a bad bad place right now competetively speaking.

 

[Capeta]

My answer to this is contained in the post you quoted. Suffice to say that any company that can afford to move production to a smaller process node, whether via capacity free-up or otherwise, does so at the earliest opportunity.

You seem a fan of Intel Capeta, so let me use them as an example for you. Chipsets are of secondary importance to them relative to their primary CPU production; CPUs are being fabbed at 65nm. Is there enough chipset "demand" to warrant leaving 130nm and going to 90nm? No, 130nm was up to the challenge on a macro level. Yet they have moved chipset production to 90nm nontheless, as it saves them money, and enhances the performance profile of such components. You use your most advanced fab equipment whenever possible, and you use it because you have it to use.

Companies that are stuck fabbing their primary products on 180nm, for example, are companies that are in a bad bad place right now competetively speaking.

Intel does it because they always have new bigger sized chips coming out and they end up selling millions of chips so yes the demand is there. I don't see SONY bulding any new big chips nor do I see great demand for them especially if they outsource.

 

[Carl B]

Capeta, I am being honest when I say what you just wrote makes no sense. Again, what does dmeand and 'new bigger chips' have to do with anything? Nothing - it doesn't. Why are you so fixated on that?

Explain to me again the economic reasoning against utlizing a newer process for older designs - PCB redesign, is that what you're telling us?

Let me phrase it as a question for you, and if this doesn't trigger a lightbulb, then well... you can believe what you want to believe.

If *you* were Sony, and you have now moved Cell production to 45nm - yet you have a state-of-the-art 65nm fab line and several existing products being built on 90nm, 130nm, and above... what would *you* do with that 65nm fab capacity?