You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Bear in mind this is the bulked up GF with the acquired IBM semi operations. Reports say that Mubadala is looking to get $15-20B.
I'm always a little concerned when industries consolidate to pretty much 3 players - TSMC, Samsung, and Intel as named. But, I'd also factor in Sanjay Jha and his experience in divestiture after the Motorola Mobility takeout.
I would say that out of the three mentioned none of them would be remotely interested. I aksed friends at TSMC and they said no way. From what I heard in Taiwan this week China's Tsinghua Unigroup is the leading suitor, if Homeland Security allows it:
"If you can't be the top-three giant, it will be very hard to develop your business in the chip industry," Zhao said, citing reports that China imported more chips than crude oil every year.
If they really want to be in the top three GF would be a good start...
Based on last year IBM paid $1.5 billion to Globalfoudries to take over IBM's semi manufacturing division and GF has a between $4 ~ 5 billion annual revenue, then the asking price of $15 ~ 20 billion seems way too high.
We have a lot of semi industry guru here. Do you know if GF has anything that is so valuable and other competitors wish to have?
Based on last year IBM paid $1.5 billion to Globalfoudries to take over IBM's semi manufacturing division and GF has a between $4 ~ 5 billion annual revenue, then the asking price of $15 ~ 20 billion seems way too high.
We have a lot of semi industry guru here. Do you know if GF has anything that is so valuable and other competitors wish to have?
Well, let´s start with the four 300mm FABs and the other 5 at 200mm as pure assets.
FAB8 in Malta completed the qualification of the 14nm, both LPE and LPP and they already have several design wins. They have invested more than 12b$ in this site only, so far.
FAB1 in Germany is the second 28nm best fab in the world for yield and cycle time (you can guess who is the leader) and they just started the 22 FDSOI qual (aka 22FDX). GF is the only company offering this technology at the moment. Killer technology to me, almost 14nm FinFET performance at the 28nm cost.
They also offer ASICs, FX-14 and CU-32 design systems and RF-SOI.
In the foundry business fab utilization is key, especially for the 300mm fabs. My sense is that the 300mm fab in Dresden is under utilized. The 22nm FDSOI process for Dresden is promising for IOT and mobile applications but likely a few years away from generating significant fab fill. In Malta I think the fab is also under utilized. AMD recently announced they will use the 14nm process in Malta but that will also likely take 28nm volume away from Dresden and Malta. The former IBM ASIC group has also recently announced they will design for the 14nm process but that will likely take a few years to generate much volume. The 200mm former IBM fab in Burlington is busy and long term the RF SOI process they run there could migrate to the 300mm Fab in Singapore for 5G applications but that is a few years away. The former IBM East Fishkill fab is old and small.
In one to two years Global could look very different with good execution but I think buyers are going to want to see some performance before they buy. The one exception to this in my view is the Chinese who are trying really hard to increase their footprint in the semiconductor industry. I think the problem with a possible china purchase is the US government is unlikely to allow it.
It would be a first, but I think the powers that be would allow China the same privileges as Abu Dhabi. Production in the US would be OK, production in China not allowed.
Obviously the goal of Tsinghua is to acquire technology for Chinese domestic production, so the deal could never work.
I am worried about elections (US and TW). After TW elections, China may not respond well to an independence-minded administration. And who knows what will happen here. These questions are political as much as economic, and China seems to be more and more impatient for change. And more and more aggressive.
In one to two years Global could look very different with good execution but I think buyers are going to want to see some performance before they buy. The one exception to this in my view is the Chinese who are trying really hard to increase their footprint in the semiconductor industry. I think the problem with a possible china purchase is the US government is unlikely to allow it.
Your latter statement is probably true but I think this points to the most likely outcome of GF being split up with the Chinese buying all the fabs outside the US, and something else being arranged for Malta and Burlingham.
Your latter statement is probably true but I think this points to the most likely outcome of GF being split up with the Chinese buying all the fabs outside the US, and something else being arranged for Malta and Burlingham.
In my option it won't be broken up and sold in pieces, I think turning it into viable separate businesses would be very complex and not workable. When you look at selling non US fabs to China and US fabs to someone else you have to split up products, product sales, R&D, etc. Just my opinion.
In my option it won't be broken up and sold in pieces, I think turning it into viable separate businesses would be very complex and not workable. When you look at selling non US fabs to China and US fabs to someone else you have to split up products, product sales, R&D, etc. Just my opinion.
I think this is showing that the semiconductor industry is maturing just as the pharma industry did about 25 years ago. For example NXP acquired Freescale but then sold off NXP's RF power business, co-incidentally to the Chinese. As the industry consolidates there are always going to be odd bits which don't fit either for logistic or regulatory issues and those will be sold off. Eventually there will be around three players in all the key markets just as happened with memory.
Also in the specific case of GF, there is little commonality between the three main fabs in any case.
The value of IBM semiconductor in my opinion is the IP and the people. The design IP and the process IP (10nm and 7nm) and that is something you can't really break up.
As Scott said, all of the pieces are there to do something great. Right now it is all about implementation. Breaking it up would be a mistake.
Usually a sale would include an amount for 'goodwill' but unfortunately I think in GF's case the parts are worth more than the sum of those parts. Malta is state of the art, Dresden is a bit of a compromise as it's two fabs joined but it has been recently re-equiped, and Singapore is older but still performs well. Add in the other locations and the assets of GF are virtually the whole of the rumoured asking price.
Probably the only people with an interest in buying the whole company would be the Chinese but obviously there are issues in the US on this. But Mubadala are an investment company and their interest is in getting the most return on their investment so I'm struggling to see an alternative for them other than splitting the company in two.
But then again Sanjay Jha has a habit of pulling rabbits out of his hat so let's see what he can come up with this time.
Usually a sale would include an amount for 'goodwill' but unfortunately I think in GF's case the parts are worth more than the sum of those parts. Malta is state of the art, Dresden is a bit of a compromise as it's two fabs joined but it has been recently re-equiped, and Singapore is older but still performs well. Add in the other locations and the assets of GF are virtually the whole of the rumoured asking price.
Probably the only people with an interest in buying the whole company would be the Chinese but obviously there are issues in the US on this. But Mubadala are an investment company and their interest is in getting the most return on their investment so I'm struggling to see an alternative for them other than splitting the company in two.
But then again Sanjay Jha has a habit of pulling rabbits out of his hat so let's see what he can come up with this time.
I think Mubadala will end up having to hold onto them for a few years.
Selling the parts would be really hard. Dresden is underutilized and the plan to fill it is 22nm FDSOI that is being developed in Malta presumably utilizing a lot of the IBM researchers. If Malta is sold to one buyer and Dresden to another, what happens to 22nm FDSOI.
If someone buys Malta they would need to take on the Samsung 14nm license and they would still have an under utilized fab. They would need to keep AMD and they would need the former IBM ASIC team to keep designing for them, etc.
The bottom line is selling fabs you get 30 to 40 cents on the dollar if you can even find someone. The only companies that want fabs, are companies that already have fabs and need more and I just don't see that right now, plus it would be a huge discount.
As a whole Global would likely be valued based one earnings and my guess is they still lose money. Broken into parts they would sell at a big discount on the assets if you could even find buyers. In either case the value would be way below the $15 to $20 billion dollar level.
Global could be a valuable company in a few years, but I don't see it yet.
Obviously we will see what happens but I would be very surprised to see them sold for anything other than a fire sale price in the next two years.
I think Mubadala will end up having to hold onto them for a few years.
Selling the parts would be really hard. Dresden is underutilized and the plan to fill it is 22nm FDSOI that is being developed in Malta presumably utilizing a lot of the IBM researchers. If Malta is sold to one buyer and Dresden to another, what happens to 22nm FDSOI.
If someone buys Malta they would need to take on the Samsung 14nm license and they would still have an under utilized fab. They would need to keep AMD and they would need the former IBM ASIC team to keep designing for them, etc.
The bottom line is selling fabs you get 30 to 40 cents on the dollar if you can even find someone. The only companies that want fabs, are companies that already have fabs and need more and I just don't see that right now, plus it would be a huge discount.
Obviously we will see what happens but I would be very surprised to see them sold for anything other than a fire sale price in the next two years.
The problem is the oil price is so low Mubadala have to liquidate some assets or cancel some serious orders such as aircraft which they actually make a profit on. So indeed it may have a be a fire sale of parts of the company.
Also I believe the GF FD-SOI process was developed in Dresden - certainly Gerd Teepe was the lead director on it. But despite this I really don't see Dresden becoming that full with FD-SOI. Europe has at least one too many FD-SOI processes, and at the moment I believe neither has passed automotive qualification which is the main market nowadays in Europe. Smaller companies doing IoT type products are all directed by European investors to use TSMC so little hope there for GF so they really need to persuade large corporates like Philips Healthcare to switch from TSMC.
How's the chance to have a group of "smaller" IDMs to start a joint venture with Mubadala? By doing so those IDMs might get the capacity and technology they need and Mubadala can gradually exit out.
Or it's just too complicated? Or there isn't enough trust to build such business model?
How's the chance to have a group of "smaller" IDMs to start a joint venture with Mubadala? By doing so those IDMs might get the capacity and technology they need and Mubadala can gradually exit out.
Or it's just too complicated? Or there isn't enough trust to build such business model?
But despite this I really don't see Dresden becoming that full with FD-SOI. Europe has at least one too many FD-SOI processes, and at the moment I believe neither has passed automotive qualification which is the main market nowadays in Europe. Smaller companies doing IoT type products are all directed by European investors to use TSMC so little hope there for GF so they really need to persuade large corporates like Philips Healthcare to switch from TSMC.
There already many customers interested in the 22FDX technology and a couple of them are big ones (the full FDSOI site capacity for the next few years is at least theoretically fully booked if they will deliver what they have claimed. Moreover I see no FDSOI competition there. The TSMC's answer so far is the 16FFC. Time will tell which one will be the best in terms of power reduction per dollar.
There already many customers interested in the 22FDX technology and a couple of them are big ones (the full FDSOI site capacity for the next few years is at least theoretically fully booked if they will deliver what they have claimed. Moreover I see no FDSOI competition there. The TSMC's answer so far is the 16FFC. Time will tell which one will be the best in terms of power reduction per dollar.
The competition I was referring to is the ST FD-SOI process. GF started off aiming to be compatible with this as Samsung are, but then went off and did their own thing. The FD-SOI ecosystem is small enough as it is without this sort of complication. As for FD-SOI capacity being booked up for three years, it must be a pretty small part of the fab then.
The fact is we keep hearing of 'big customers' interested in FD-SOI but nothing actually appears in production units that ChipWorks can take apart.
The competition I was referring to is the ST FD-SOI process. GF started off aiming to be compatible with this as Samsung are, but then went off and did their own thing. The FD-SOI ecosystem is small enough as it is without this sort of complication. As for FD-SOI capacity being booked up for three years, it must be a pretty small part of the fab then.
The fact is we keep hearing of 'big customers' interested in FD-SOI but nothing actually appears in production units that ChipWorks can take apart.
That's what I really meant Mike. To date I haven't seen anything from ST (so, there is no serious competition/competitor yet).
About GF, first volume parts are expected in early 2017, so of course you cannot see anything yet (already took apart by ChipWorks).
About capacity, 22FDX is going to be the biggest portion of the site, absolutely. Not asking you to trust me anyway, time will tell. GF is not public, so there is no speculation here.
That's what I really meant Mike. To date I haven't seen anything from ST (so, there is no serious competition/competitor yet).
About GF, first volume parts are expected in early 2017, so of course you cannot see anything yet (already took apart by ChipWorks).
About capacity, 22FDX is going to be the biggest portion of the site, absolutely. Not asking you to trust me anyway, time will tell. GF is not public, so there is no speculation here.
I am at IEDM right now and I got a briefing from Global Foundries on 22FDX on Monday. I will be blogging about it in the near future. it was a very interesting briefing with lots of details.
