SOITEC SOI wafer business almost wiped out

[mrmt]

http://www.soitec.com/pdf/pr_soitec_q3_13-14_sales_en.pdf


Sales in our Digital business (i.e. total 300 mm wafer sales) were down by 74.6% at 8.1 million Euros for the third quarter on a year-on-year basis and by 48.8% on a sequential basis.

Looks like that business will close down soon. Given that the only significant line of products using SOI wafers is AMD opteron/FX, it will become a nice parameter of how sales of those chips are going.

 

[norseamd]

Looks like that business will close down soon. Given that the only significant line of products using SOI wafers is AMD opteron/FX, it will become a nice parameter of how sales of those chips are going.

any other plants for opteron or fx

 

[WhoBeDaPlaya]

I think it's still a popular choice for implementing silicon photonics.
Hell, it's what I did for my doctoral research.

 

[mrmt]

I think it's still a popular choice for implementing silicon photonics.
Hell, it's what I did for my doctoral research.

The numbers speak for themselves regarding the popularity of SOI.

 

[Xpage]

sad, i thought FD-SOI has potential for helping voltage characteristics for GloFo to try to compete better vs Intel.

 

[NTMBK]

Between AMD's APUs dumping SOI, IBM's POWER server sales dropping like a stone, and the next gen console chips not going with SOI, I'm not surprised. Are there any new products choosing SOI, other than IBM's 22nm chips? (Actual products, not just tech demos.)

 

[mrmt]

Between AMD's APUs dumping SOI, IBM's POWER server sales dropping like a stone, and the next gen console chips not going with SOI, I'm not surprised. Are there any new products choosing SOI, other than IBM's 22nm chips? (Actual products, not just tech demos.)

IBM is peddling its foundry on the market, and given that the new crop of EP and EX Intel servers is another generational leap in both number of cores and power consumption, I'd be worried about POWER future as a viable product.

 

[krumme]

I will take 100 years before anyone jumps the ibm foundry tech wagon again. Isnt it like 65nm was the last reasonable decision for their part? The rest was just bad. Bad like really really expensive. And bad. Lol

 

[krumme]

The first through the gate was last

 

[mrmt]

I will take 100 years before anyone jumps the ibm foundry tech wagon again. Isnt it like 65nm was the last reasonable decision for their part? The rest was just bad. Bad like really really expensive. And bad. Lol

There will be no IBM foundry tech wagon in a few years, because IBM will either sell the business or fold it.

 

[krumme]

Sure. Its incredible they could get this long. Some healthy profit on serverside in some niches and pretty excellent research capabilities on the technology side.

 

[Ajay]

sad, i thought FD-SOI has potential for helping voltage characteristics for GloFo to try to compete better vs Intel.

Same here, seems like some sane people thought it would have been the equivalent of a whole node decrease in perf/watt - but AMD just didn't have the cash to pay GLF for the extra development costs. There was an ultra low power node on GloFo's node chart for 20nm, but without any interesting in FD-SOI from AMD I doubt GFL will foot the bill even for that (plus the fact that 20nm is running very late).

 

[norseamd]

Same here, seems like some sane people thought it would have been the equivalent of a whole node decrease in perf/watt - but AMD just didn't have the cash to pay GLF for the extra development costs. There was an ultra low power node on GloFo's node chart for 20nm, but without any interesting in FD-SOI from AMD I doubt GFL will foot the bill even for that (plus the fact that 20nm is running very late).

if amd puts the small cores like the mullins and the console chips as well would they be able to have tsmc make their big chips like the potential large x86 cpu they have in development

 

[Ajay]

if amd puts the small cores like the mullins and the console chips as well would they be able to have tsmc make their big chips like the potential large x86 cpu they have in development

I keep hearing of a 'potential' large core x86, but where is AMD going to find the funds to release a new micro architecture??? The CMT module approach is moribund, even though I once thought it could be fixed. I think AMD is stuck with follow ons to Kabini and Kaveri APUs plus ARM microservers for their CPU business (which is another way of saying it is slowly dying out, except for console APUs).

 

[norseamd]

I keep hearing of a 'potential' large core x86, but where is AMD going to find the funds to release a new micro architecture??? The CMT module approach is moribund, even though I once thought it could be fixed. I think AMD is stuck with follow ons to Kabini and Kaveri APUs plus ARM microservers for their CPU business (which is another way of saying it is slowly dying out, except for console APUs).

well if they make a great x86 core for enthusiasts i will be very grateful. there could be a market for it as long as it works more than bulldozer based modules.

do you know if they could use tsmc for it?

 

[MisterMac]

Since mubadala likes burning money for fun - what's the chance further down the line once IBM gets more desperate to offload the foundry... that they court to GF to offload?

Wouldn't that be an interesting strategic perspective for GF to acquire that expertise and do something with mubadala's money?

They seem like a good match.

 

[norseamd]

Since mubadala likes burning money for fun - what's the chance further down the line once IBM gets more desperate to offload the foundry... that they court to GF to offload? Wouldn't that be an interesting strategic perspective for GF to acquire that expertise and do something with mubadala's money? They seem like a good match.

interesting

would that make global foundries much better?

 

[MisterMac]

interesting

would that make global foundries much better?

IBM has the Tech - lack the execution money.

GF has the money to execute but not TECH.


Doesn't seem like a bad alliance - but maybe some wiser technology EE people can fill us in.

 

[norseamd]

surprised global foundries has that much money

 

[norseamd]

ibm has some really advanced stuff at least i think so this might be good if global foundries can get some of it

 

[podspi]

surprised global foundries has that much money

They don't, their owners do. The big question for GF has always been how much they're willing to dump into the thing, and for how long.

 

[mrmt]

IBM has the Tech - lack the execution money.

GF has the money to execute but not TECH.

IBM surely doesn't lack the money to build a foundry business, but they don't want to invest their money in a business earning less than it should.

ATIC might want to swallow the losses and the huge CAPEX bill to build a foundry business, but the question is, for how long. It's been almost 5 years after the spin off, enough to run an entire R&D cycle and Globalfoundries doesn't have much to show. I don't think they can afford to spend the next 5 years chasing customers, footing huge CAPEX bills and eating losses as they are now.

 

[NTMBK]

IBM surely doesn't lack the money to build a foundry business, but they don't want to invest their money in a business earning less than it should.

ATIC might want to swallow the losses and the huge CAPEX bill to build a foundry business, but the question is, for how long. It's been almost 5 years after the spin off, enough to run an entire R&D cycle and Globalfoundries doesn't have much to show. I don't think they can afford to spend the next 5 years chasing customers, footing huge CAPEX bills and eating losses as they are now.

They are at least starting to get some wins from customers who aren't AMD- Mediatek, Rockchip, and I think Qualcomm all have some parts there, which is a start at least. Obviously it's no TSMC, but they have to start somewhere.

How much do GlobalFoundries get out of the "fab club" at the moment? They've diverged significantly from IBM processes, getting rid of SOI, no 22nm process- do they actually piggyback on much IBM R&D these days?

 

[mrmt]

How much do GlobalFoundries get out of the "fab club" at the moment? They've diverged significantly from IBM processes, getting rid of SOI, no 22nm process- do they actually piggyback on much IBM R&D these days?

Globalfoundries is trying to develop R&D inside the company, especially after the 32nm snafu, when IBM engineers refused to listen to GLF engineers pointing out problems in the process. But given their track record with the 28nm process, I don't think they are even near of what they should be to become competitive against TSMC, their true target in the foundry market.

As for the orders, Rockchip is too small to be meaningful. The Mediatek and Qualcomm orders are the relevant ones. Albeit small the orders show that these companies want to at least give Globalfoundries a shot in their production runs, but between dipping the toe and making Globalfoundries a second source for their lagging edge products there's a huge gulf. As much as these orders shouldn't make Globalfoundries much money, because the foundry should have been very aggressive on prices, they might boost Globalfoundries credibility, something they sorely need after their debacles with AMD.

I think it's fair to say now that Globalfoundries lost the battle for 20nm. The question now is whether they will deliver their "16nm" Finfet node in a time frame competitive enough and in a shape good enough to compete against TSMC on price, and whether they will get their "10nm" node in time to compete against TSMC for bleeding edge orders. If they don't, I don't think they will have any other choice but shrink down the business.

In that scenario, it's easy to see Globalfoundries acquiring IBM foundry business: It's an engineering team they are already accustomed to work with and it's a top notch team that should have a path for 14/16nm and 10nm already in the pipeline. They would have to just tune the project targets for these nodes. Instead of uber power hogs SOI chips, they should target small, power sipping, bulk mobile chips.