[TechEye] GlobalFoundries to buy IBM Semi

[Vesku]

Having a process ready and first product in retail doesn't go together every time. Intel manages to do it at the same time because they are both the Fab manufacturer and the retail product designer.
Also, GloFo had 32nm process and products in retail before TSMCs 28nm. GloFos 28nm is a half node of 32nm when TSMC 28nm is a full node over 40nm.
Comparing time-line of products in retail doesnt show when each process was ready.

Sure, it was "ready" but despite GF basically begging non-AMD clientele to use their foundry it took over a year to pump out the first product. Yes, that's exactly how it went down. Reality check. It doesn't seem very likely, timeline suggests there were yield issues well after GF's claim of being ready. We are talking about an industry that routinely claims their nodes are farther along then they are by simply ignoring the yield aspect:

"GLOBALFOUNDRIES: Ready for 20nm semiconductor designs" 8/30/2011
http://electroiq.com/blog/2011/08/globalfoundries-ready-for-20nm-semiconductor-designs/

"Two flavors of GF 20nm, one for 2013 rollout one for 2014 rollout" 8/30/2011
http://www.brightsideofnews.com/news/2011/8/30/globalfoundries-announces-20nm-process-no-to-soi.aspx

"ARM, TSMC tape-out 20-nm processor" 10/18/2011
http://www.eetimes.com/document.asp?doc_id=1260423


If we go by foundry PR then we should be swimming in 20nm chips.

 

[witeken]

Im sure you have technical data to provide ???

Edit: Not to mention that 28nm SLP (Gate First) got 40% lower power, 30% higher performance and 2x times the density of 40nm LP (Gate Last). Not only that, being a Half Node of 32nm means they can save huge amount of resources, time and money both GloFo and AMD.
Also people always forget that Gate First is 10-20% cheaper than the same process on Gate Last.

I don't want to search for the exact details, but as far as I know gate-first was much more difficult than gate-last, causing delays, and for smaller process nodes, gate-last will be required anyway. So staying 1 more node on gate-first didn't make any sense.

 

[Vesku]

I don't want to search for the exact details, but as far as I know gate-first was much more difficult than gate-last, causing delays, and for smaller process nodes, gate-last will be required anyway. So staying 1 more node on gate-first didn't make any sense.

Yes, difficulty and resulting delays. However, Homeles' claim of a large performance difference does not seem to be supported by TSMC vs GF Kabini.

Not that it's any less damaging to making money for AMD, Kabini is actually very competitive vs Bay Trail outside of the tablet space yet the A4-5000 and A6-5200 seemed to be supply constrained at least for the prices the market wanted.

 

[CHADBOGA]

Like everybody else. How can that be?

Insurance.

 

[CHADBOGA]

Why do you think that? Until recently, Intel didn't have any products to offer Apple for their iPhones and iPads.

Because Apple keep hiring semiconductor designers like they are going out of style.

Apple may eventually go to Intel for their phone & Tablets, but that looks at least a few years away.

 

[CHADBOGA]

Just because a company got billions doesnt mean they wanna waste it

And btw, the stockholders got a first ticket to those billions.

Intel has shown us that companies can spend billions making ridiculous acquisitions & investments(see McAfee) and the stockholders get no say whatsoever.

The stockholders get assurances that the acquisition makes sense and will pay off, but when it doesn't, heads don't roll.

PowerPC so to say will happen all over again with ARM. And they will be x86 in the future.

Probably, but that could still be 5+ years away and in the meantime I think Apple is going to see what they can do, going it alone.

 

[ShintaiDK]

Intel has shown us that companies can spend billions making ridiculous acquisitions & investments(see McAfee) and the stockholders get no say whatsoever.

The stockholders get assurances that the acquisition makes sense and will pay off, but when it doesn't, heads don't roll.



Probably, but that could still be 5+ years away and in the meantime I think Apple is going to see what they can do, going it alone.

Intel pays regular dividends. So their stockholders are happy enough.

It would take more than 5 years to make a foundry business like IBM competitive. And it would cost the first 10B$ in those 5 years just in R&D.

 

[mrmt]

I dont see Intels interest in buying ibm process tech. What should it be? There is nothing imo.

If Intel is really committed to launch a foundry business to third parties, acquiring IBM client portfolio and IBM relationship portfolio would be a significant boost to that business. This is a movement akin to Globalfoundries acquisition of Chartered. Get IBM and IBM clients and move them to Intel processes.

 

[CHADBOGA]

Intel pays regular dividends. So their stockholders are happy enough.

I doubt that in what has been a bullmarket, any Intel shareholders are particularly happy with how Intel has performed, dividend or not, in the last 3+ years.

And why would any shareholder be happy with a company squandering billions? That makes no sense whatsoever.

Shareholders don't have to be rioting in the streets, to be unhappy with decisions made by companies they have invested in.

It would take more than 5 years to make a foundry business like IBM competitive. And it would cost the first 10B$ in those 5 years just in R&D.

This may be so, but none of that is compelling enough to make me think Apple definitely won't buy its own fab.

Just to be clear, I'm not saying Apple is definitely going to buy its own fab, just that it wouldn't surprise me in the least, if they did.

 

[mrmt]

Judging by A6-5200 vs GF 5350, TSMC's 28nm seems to be a bit more efficient but it's not drastically so. Hard to get an exact comparison since the TSMC kabini is on prebuilt boards and GF is socketed but the difference appears to be relatively minor. The main impact of following IBM's lead seems to be delays, though, for whatever reason time to production doesn't seem to be near the top of IBM's priority list when planning node transitions.

With the caveat that Globalfoundries is launching 28nm high volume chips two years after the first TSMC 28nm high volume chip. 2 years is almost the bleeding edge life of a given node, and Globalfoundries had to spend exactly this amount of time to get their 28nm to the point where TSMC's 28nm was two years ago.

I can't stress enough how bad this situation is.

 

[mrmt]

It would take more than 5 years to make a foundry business like IBM competitive. And it would cost the first 10B$ in those 5 years just in R&D.

R&D isn't the problem for IBM, production volumes and CAPEX are. Each new node you have to invest MORE in equipment and each wafer will yield MORE candidates, meaning that you need a higher volume to justify a fab. Does IBM have enough volume to warrant a 14nm or 10nm fab? With their foundry sales stalled and POWER sales crashing, they should not have a viable foundry business in the next 5 years.

 

[OCGuy]

While GF may eventually be the winning bidder, the title of this thread/story is currently inaccurate.

 

[witeken]

Because Apple keep hiring semiconductor designers like they are going out of style.

Hiring people to build your own SoCs is one thing, and even small companies like ARM can do that. But building a leading edge foundry is something completely different. It would costs multiple 10B USDs on R&D and capex, just to save a few bucks, that would otherwise go to Samsung or TSMC, to keep the foundry margin like Intel does. I think Apple rather wants to spend that money to create new products to sell at high margins to the iSheeps.

 

[ShintaiDK]

R&D isn't the problem for IBM, production volumes and CAPEX are. Each new node you have to invest MORE in equipment and each wafer will yield MORE candidates, meaning that you need a higher volume to justify a fab. Does IBM have enough volume to warrant a 14nm or 10nm fab? With their foundry sales stalled and POWER sales crashing, they should not have a viable foundry business in the next 5 years.

With 14nm its already questionable if GloFo can make money. And with 10nm its likely that Samsung will also be out of the race.

This may be so, but none of that is compelling enough to make me think Apple definitely won't buy its own fab.

See above. Apple would need to drasticly increase its volume or be a major foundry to ever get its money back if they tried. But again, its against Apples business plan. Just like they outsource all production for everything else.

And currently Apple is the loser in the grand game. So its not exactly good business to invest in something you are stuck on.

Its the same thing over and over like with GloFo. For some reason people thing someone outside will just throw massive amounts of money headless into the game.

Apple could just as well buy Intel or TSMC rather than last place IBM. IBM is getting out for the exact same reasons you want Apple in.

 

[R0H1T]

Hiring people to build your own SoCs is one thing, and even small companies like ARM can do that. But building a leading edge foundry is something completely different. It would costs multiple 10B USDs on R&D and capex, just to save a few bucks, that would otherwise go to Samsung or TSMC, to keep the foundry margin like Intel does. I think Apple rather wants to spend that money to create new products to sell at high margins to the iSheeps.

You're ignoring the money Apple can make out of that one foundry, by using it for their own products but more importantly allowing their competitors, unlike Intel to make use of the given foundry.

A few years down the line only two or three major foundries will remain, Intel being one, now the node shrink costs at that point in time would be astronomical even for Intel so it would make sense to cash in on an opportunity like that & kill two birds with one stone. All speculation, as CHADBOGA says, atm but it's certainly something worth pondering over considering Apple may just have some real intentions to move their Mac line away from x86

 

[bullzz]

@R0H1T - buying a foundry for mac line which sells <10m units doesnt make sense. like intel, apple will have to keep selling more chips with every node. and this is possible if they can only fab for others..

 

[R0H1T]

@R0H1T - buying a foundry for mac line which sells <10m units doesnt make sense. like intel, apple will have to keep selling more chips with every node. and this is possible if they can only fab for others..

Yes that was my primary argument but we're just speculating here, so if Apple does intend to move away from x86 & Intel with their own custom ARM SoC then it would make sense for them to buy a foundry. The iphone/ipad line would certainly eat up a huge chunk of the capacity while the bigger chips would be there for Macs, of course the likes of Qualcomm or Nvidia & even AMD would be no direct threat to them so fabbing their chips would certainly be very real possibility at that point in time.

 

[ElFenix]

I doubt that in what has been a bullmarket, any Intel shareholders are particularly happy with how Intel has performed, dividend or not, in the last 3+ years.

And why would any shareholder be happy with a company squandering billions? That makes no sense whatsoever.

This is moving far off topic now, but if the company believes its shareholders can invest better than it can (for example, if the company is over capacity), then the company should be cashing out shareholders. Though I think share buy-back programs are commonly preferred to dividends nowadays.

 

[NTMBK]

While GF may eventually be the winning bidder, the title of this thread/story is currently inaccurate.

I was just going with the title of the TechEye story, but I agree that it is inaccurate. If a mod wants to change it then I have no objection.

 

[krumme]

Insurance.

Agree and
Insurance = money

You pay for safety and keeping risk low. Samsung and Apple wants to protect what makes money. Phones and tablets. They pay to protect the cash cows and the rising stars.

 

[krumme]

Because Apple keep hiring semiconductor designers like they are going out of style.

Apple may eventually go to Intel for their phone & Tablets, but that looks at least a few years away.

Go ask Otellini. The train left years back.
What we see of results now shows Apple actually made a brilliant decicion. Who would have thought it panned out so well. It is damn impressing and surpricing performance they get now in so few years imho.

 

[krumme]

Intel has shown us that companies can spend billions making ridiculous acquisitions & investments(see McAfee) and the stockholders get no say whatsoever.

The stockholders get assurances that the acquisition makes sense and will pay off, but when it doesn't, heads don't roll.



Probably, but that could still be 5+ years away and in the meantime I think Apple is going to see what they can do, going it alone.

Intel seems to be very unsuccessful untill now in investments in their non core business. But there are some very innovative people working there. Outside of mcafee that was the most expensive bummer ever i think its difficult to judge their initiatives - but they try a lot.

Everytime they do something new it takes ton of flak here. But doing new business is damn hard. Perhaps they should look inside (!) And see if there is any culture whatever damping or skewing innovation.

Perhaps they should just realize they are just not the most innovative guy in the class and just make sure they can follow far faster than what they have managed until today. Looking at their competences and history that might be the most realistic strategy. Lol.

 

[krumme]

If Intel is really committed to launch a foundry business to third parties, acquiring IBM client portfolio and IBM relationship portfolio would be a significant boost to that business. This is a movement akin to Globalfoundries acquisition of Chartered. Get IBM and IBM clients and move them to Intel processes.

Very good argument. As beeing tied to a wsa and mubadala is a hell Intel is in good position to take the deal if they want. It means we can then know how serious they are about the third party strategy if or if not they take ibm process tech.

 

[krumme]

With the caveat that Globalfoundries is launching 28nm high volume chips two years after the first TSMC 28nm high volume chip. 2 years is almost the bleeding edge life of a given node, and Globalfoundries had to spend exactly this amount of time to get their 28nm to the point where TSMC's 28nm was two years ago.

I can't stress enough how bad this situation is.

This. It is a huge mess. What is the purpose of gf if its 1-2 years behind tsmc?

 

[AtenRa]

This. It is a huge mess. What is the purpose of gf if its 1-2 years behind tsmc?

You forgetting 32nm SOI HKMG, GloFo's 28nm HKMG Bulk is half-Node of the 32nm process. GloFo is not 1-2 years behind TSMC, their 20/14nm are on par with TSMC.