Digitimes: Qualcomm to switch 28nm HKMG wafer orders from TSMC to Globalfoundries

[SiliconWars]

http://www.digitimes.com/news/a20130806PD215.html

Qualcomm reportedly will switch 10,000 units, or 20% of its orders for 28nm HKMG wafers, from Taiwan Semiconductor Manufacturing Company (TSMC) to Globalfoundries starting September 2013

That's interesting for a bunch of reasons.

Could this be Qualcomm retaliating against TSMC and Apple? GF's 28nm in much better shape than suggested? Seronx right after all and AMD is going to grab the free capacity at TSMC for Kaveri?

 

[MisterMac]

Well that's a highprofile order - if not massive volume regardless still good for GF.

 

[mrmt]

Could this be Qualcomm retaliating against TSMC and Apple? GF's 28nm in much better shape than suggested? Seronx right after all and AMD is going to grab the free capacity at TSMC for Kaveri?

If GLF 28nm is good enough for Qualcomm to make a switch, why wouldn't AMD use the same process?

 

[MisterMac]

If GLF 28nm is good enough for Qualcomm to make a switch, why wouldn't AMD use the same process?

Better SHP\HP processes while GLF has shifted towards lowpower?

 

[frozentundra123456]

"Global Foundaries is believed to have offered price incentives", seems the obvious reason, or at least a big part of it.

 

[lefty2]

Doesn't that mean that Qualcomm have to tape out their designs again?

 

[Blandge]

I was under the impression that Qualcomm didn't use HKMG on TSMC 28 nm? Or are they saying that now they WILL use HKMG on GF 28 nm?

 

[Exophase]

I was under the impression that Qualcomm didn't use HKMG on TSMC 28 nm? Or are they saying that now they WILL use HKMG on GF 28 nm?

They're using HKMG on S800.

 

[Blandge]

They're using HKMG on S800.

Ahh I guess that's old new then.

 

[Idontcare]

Doesn't that mean that Qualcomm have to tape out their designs again?

Yes, absolutely they will. Gate-first has very different design rules than gate-last, even though they both claim to share the same node label (28nm).

 

[erunion]

http://www.digitimes.com/news/a20130806PD215.html

GF's 28nm in much better shape than suggested?

In what way? 28nm is no longer cutting edge. TSMC already soaked up all the early adoption high margins, now GF is undercutting TSMC on price.

 

[SiliconWars]

28nm is cutting edge, even TSMC won't be on 20nm until 9 months or so, and it won't be anything like the current capacity of 28nm for another 2 years.

 

[mrmt]

28nm is cutting edge, even TSMC won't be on 20nm until 9 months or so, and it won't be anything like the current capacity of 28nm for another 2 years.

28nm is cutting edge in a sense as it's the best available node today, but it's not the node receiving the bulk of new investments. Only if you are in the lagging edge you are investing in 28nm now. If you are bleeding dge, you are gearing up for 20nm.

 

[lefty2]

28nm is cutting edge, even TSMC won't be on 20nm until 9 months or so, and it won't be anything like the current capacity of 28nm for another 2 years.

TSMC say 20nm has been at risk production since the start of this year and will start volumn production next February. Also, claim to be ahead of everybody else (except Intel):
http://www.xbitlabs.com/news/other/...ng_20nm_Process_Technology_to_Early_2014.html

 

[AtenRa]

28nm is the one that receives the bulk of orders right now and will continue for the entire 2014 as well. 20nm may be the cutting edge but it will be 2-3 times more expensive for the early adopters in H2 2014.

 

[Idontcare]

I really hope this news is true. I haven't heard or seen anything to confirm it on the ground, so it may very well be bogus.

But TSMC moves in direct reflection of where they think the competition is moving (you saw that when they announced their gimped 16nm plan only after GloFo announced their 14nmXM plan) so without competition TSMC really will fall back to "rest on laurels" mode.

GloFo needs a nice high profile customer that is well monied and has the product volumes to fill a fab or three. That will really help solidify their own internal R&D plans and galvanize their workforce to double down.

 

[erunion]

28nm is cutting edge, even TSMC won't be on 20nm until 9 months or so, and it won't be anything like the current capacity of 28nm for another 2 years.

28nm is no longer earning cutting edge prices. TSMC has reduced 28nm wafer pricing as its capacity improved. GF's entry will bring prices even lower.

28nm is mainstream.

 

[Idontcare]

Its cutting edge for GloFo, and "good enough" for Qualcomm (if there is any truth to the "story" itself), so everyone wins.

Even TSMC's other customers will win from this as the 28nm wafer allocation at TSMC will be under less demand and less pressure.

The only non-winner here is TSMC. Not that they won't still sell every 28nm they can make, but they will have less incentive to expand 28nm capacity to the same degree as they were planning when they were planning on a stable demand picture from Qualcomm.

 

[erunion]

Its cutting edge for GloFo, and "good enough" for Qualcomm (if there is any truth to the "story" itself), so everyone wins.

Even TSMC's other customers will win from this as the 28nm wafer allocation at TSMC will be under less demand and less pressure.

The only non-winner here is TSMC. Not that they won't still sell every 28nm they can make, but they will have less incentive to expand 28nm capacity to the same degree as they were planning when they were planning on a stable demand picture from Qualcomm.

I don't claim to have your insight into the fab business, but mrmt's comment makes a lot of sense.

Anyone currently designing a new product on 28nm now has to be considering that it will be competing against 20nm products during its life.

I wonder if Qualcomm might be bringing more budget products to GF. Does Qualcomm still have products on TSMC 40nm?

 

[wlee15]

I don't claim to have your insight into the fab business, but mrmt's comment makes a lot of sense.

Anyone currently designing a new product on 28nm now has to be considering that it will be competing against 20nm products during its life.

I wonder if Qualcomm might be bringing more budget products to GF. Does Qualcomm still have products on TSMC 40nm?

The first Qualcomm chip that are 28nm HKMG are it's Snapdragon 800 line whose products have just been released.

 

[SiliconWars]

28nm is no longer earning cutting edge prices. TSMC has reduced 28nm wafer pricing as its capacity improved. GF's entry will bring prices even lower.

28nm is mainstream.

28nm is still less than a third of TSMC's revenue, with 40m and below being half in total - http://www.xbitlabs.com/news/other/...Nearly_One_Third_of_TSMC_s_Wafer_Revenue.html

It's still growing for them, or it was before GF got in on the act.

TSMC’s 28nm shipments on track to triple in wafer sales this year.

Clearly it's still a very profitable node, and there's no reason why it won't be for GF also. GF has a history of ramping production far quicker than TSMC does so they'll be up to speed very quickly.

Anyone currently designing a new product on 28nm now has to be considering that it will be competing against 20nm products during its life.

Companies can go bankrupt waiting on the next node, then paying for early access to it. 20nm wafers could cost up to 4x more for the first 6 months-1 year than 28nm is currently. Even at half the die size these chips are uneconomical for a long time, except for those who absolutely need to be on the extreme cutting edge and can command much higher prices because of that. Even at that we're talking low volume.

I wonder if Qualcomm might be bringing more budget products to GF. Does Qualcomm still have products on TSMC 40nm?

Of course, 40nm is only slightly less popular than 28nm now, that's why in 2 years time 28nm will be a mainstream node.

 

[Idontcare]

I don't claim to have your insight into the fab business, but mrmt's comment makes a lot of sense.

Anyone currently designing a new product on 28nm now has to be considering that it will be competing against 20nm products during its life.

I wonder if Qualcomm might be bringing more budget products to GF. Does Qualcomm still have products on TSMC 40nm?

These could definitely be budget designs that simply need to exist for use in the $50 mobile phone segment.

Another thing is that 10,000 wafers is really peanuts in the big picture. This is more about Qualcomm testing GloFo to see how it does as a fab. Dipping its toe into the water so to speak.