Question [WSJ] Intel in talks to buy GloFo

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
Well this could be troublesome if AMD can't switch to Samsung for IODs.
I am not so sure... I think an acquisition like this would need to go through the usual regulatory review, which usually takes a year or more. AMD is already winding down their reliance on GloFo for their IODs and will be moving to TSMC N6 next year, meaning by the time the acquisition were to go through, AMD may not need to rely on GloFo at all. Also would not be surprised if the AMD WSA contract either gets upheld till 2024 per the terms and agreements or if it has to be nullified prematurely that there be a penalty clause for GloFo.
 

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
I think the more important thing is that GloFo was given a great working 7nm process from the IBM 'gift'. They chose to not buy the equipment to actually make them.
 

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
I think the more important thing is that GloFo was given a great working 7nm process from the IBM 'gift'. They chose to not buy the equipment to actually make them.
It's been awhile since I checked GloFo/IBM's 7nm process, but Intel already has a working "7nm"-class node, which is their 10nm node, so I am not sure what the point of having two 7nm-class nodes would do for their IDM 2.0 strategy. Are you implying they'd keep their 10nm ESF node in house and only offer the IBM 7nm node for the IDM clients if they acquire GloFo with the intent of reviving the IBM 7nm node?
 
  • Like
Reactions: Tlh97

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
It's been awhile since I checked GloFo/IBM's 7nm process, but Intel already has a working "7nm"-class node, which is their 10nm node, so I am not sure what the point of having two 7nm-class nodes would do for their IDM 2.0 strategy. Are you implying they'd keep their 10nm ESF node in house and only offer the IBM 7nm node for the IDM clients if they acquire GloFo with the intent of reviving the IBM 7nm node?
If it performs better then sure. IBM had some of the best process engineers in the business. Their volume was too low though to afford how expensive ever smaller nodes cost. IBM is the only company to release a 5.5 GHz from the factory CPU for example.
 

Mopetar

Diamond Member
Jan 31, 2011
7,797
5,899
136
Well this could be troublesome if AMD can't switch to Samsung for IODs.

Intel would still have to honor any contracts that were set up ahead of time or pay off AMD to break the contract if they really didn't want to do business with them.

I am not so sure... I think an acquisition like this would need to go through the usual regulatory review, which usually takes a year or more. AMD is already winding down their reliance on GloFo for their IODs and will be moving to TSMC N6 next year, meaning by the time the acquisition were to go through, AMD may not need to rely on GloFo at all.

I'm not so sure about all of that. AMD recently committed to $1.6 billion in wafer purchases from Global Foundries through 2024. There's also a lot of demand for TSMC wafers right now so it doesn't make a lot of sense to move IO die production to TSMC while they're constrained on wafers.

There's also a strong possibility that all of this goes nowhere or is completely bunk. The article seems to suggest that it isn't overly serious or anything more than some idea that got brought up in an Intel board meeting:

Any talks don’t appear to include GlobalFoundries itself as a spokeswoman for the company said it isn’t in discussions with Intel.
 

Vattila

Senior member
Oct 22, 2004
799
1,351
136
This could be good news for manufacturing in USA, by creating a formidable foundry competitor on the global stage. But Intel Foundry Services needs to become fully independent for it to work, I think. Then, strangely, we may see AMD manufacture chips at IFS in the future.


PS. Intel Foundry Services' mission statement is to become an independent entity:

"This independent organization will leverage the strength of Intel’s capabilities to offer three key pillars underpinning a world-class foundry [...] IFS has been set up as a standalone business completely dedicated to ensuring foundry customers’ products receive the utmost focus in terms of service, technology enablement and capacity commitments."

Fact Sheet: (intel.com)
 
Last edited:

Vattila

Senior member
Oct 22, 2004
799
1,351
136
The article seems to suggest that it isn't overly serious or anything more than some idea that got brought up in an Intel board meeting

"It’s clear from the WSJ story that the deal isn’t a sure thing, and GlobalFoundries outright denied that it was in talks with Intel. But it’s possible Intel’s negotiating with the investment firm that owns GlobalFoundries instead, as the WSJ points out."

 
  • Like
Reactions: Tlh97 and Saylick

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
I don't know about buying GlobalFoundries. However, they could assist in groundbreaking with Fab8.2/Fab8.3 and get exclusivity for Intel Foundry Service customers. GloFo potentially can also ensure that .2/.3 get Gov-contract compliance thus allowing defense-app customers to produce.

Fab 8.2 = 14nm/10nm DUV Intel processes
Fab 8.3 = 7nm/5nm EUV Intel processes

With Fab8.1 ditching Samsung-derived nodes to follow up on GlobalWafers/SOITEC agreements.
 
Last edited:
  • Like
Reactions: Tlh97 and Vattila

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
I'm not so sure about all of that. AMD recently committed to $1.6 billion in wafer purchases from Global Foundries through 2024. There's also a lot of demand for TSMC wafers right now so it doesn't make a lot of sense to move IO die production to TSMC while they're constrained on wafers.
Not sure what volume of IODs will move to TSMC N6, but all the rumors point to a shrinking of the IOD, with the fab being TSMC. I highly doubt next gen AMD chiplet parts will stick with GF 14nm for the IOD while moving forward with TSMC N5 for the compute dies, so the 14nm wafers will be to provide a continuous supply of last gen products (i.e. Zen 3-based products). I get that IO doesn't scale well, but in its current form the 14nm IODs take up more power than we'd all like. AMD knows that they need to address the power as well, so moving the IOD to the former cutting edge node makes the most financial sense.
 
  • Like
Reactions: Tlh97

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
The last thing this world needs is even MORE semiconductor consolidation. As it sits right now we only have a few huge players. This is very bad for the industry.
Consolidation has been going on for sometime and will continue. TSMC is so dominate now that other semiconductor manufacturers will need to merge to compete. Oh, and the Chinese will eventually get their act together/get out from under the US ban and become a player with deep pockets.
 

RanFodar

Junior Member
May 27, 2021
19
17
51
The last thing this world needs is even MORE semiconductor consolidation. As it sits right now we only have a few huge players. This is very bad for the industry.

Consolidation may be a huge concern, but geopolitics is more and more relevant nowadays. As I understand it now, the United States (esp. the government) desires to take their historical semiconductor hegemony back to the Western nations, and Intel may play a huge role for that. We also need a worthy competitor to TSMC.

This will ONLY make sense if Intel spins off its Foundry+GF into an independent foundry business with an arm's length. Until then, I still have my reservations.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
Could be good for GF, secure financial backing and an eye on heavy infrastructure upgrades across the board to boost production and run a lot of active foundries.

An issue I see in the semiconductor business as a whole is lack of capacity and perhaps premature abandonment of older gen nodes. Given the immense real estate area available in Arizona for example, I could foresee perhaps three or four gens of production running simultaneously to meet different needs. We've reached the point where a great number of things don't really need or see substantial benefits to being pushed to cutting edge fabs.

This goes from wise to downright sage if the CCP makes a move on Taiwan, which would lead to a devastating pacific war and that entire tech ecosystem being annihilated. Modern urban total warfare would be a terrifying prospect that is only a small error in judgment or system failure away from becoming a reality.
 
  • Like
Reactions: Tlh97 and Vattila

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
If it performs better then sure. IBM had some of the best process engineers in the business. Their volume was too low though to afford how expensive ever smaller nodes cost. IBM is the only company to release a 5.5 GHz from the factory CPU for example.

You forget the Zec was also used as a multi chip package, which was also impressive for its time.
 
  • Like
Reactions: Tlh97 and scannall

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Probably not until IDF can capture at least 15% of the market. Investment on a future node isn't going to sustainable without Intel's core business for a while.
IDF 2.0 or whatever it's being hailed as won't matter much when the majority of GF's clientele utilizes older nodes. Their work in photonics won't mean much either if Intel or GF as an independent subsidiary can't get things going. If this story is true, Gelsinger has bigger you know whats than prior Intel CEOs in the last 20 years. It goes without saying that even if Intel can manage to hit 3 or 2 'nanometer' within the next 3-4 years, it doesn't mean the product will be good. A slide down the scales of nodes, regardless of how it's being measures, means little if the company designing their processors can't come up with a good design that can be extrapolated over multiple generations.
 

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Not sure what volume of IODs will move to TSMC N6, but all the rumors point to a shrinking of the IOD, with the fab being TSMC. I highly doubt next gen AMD chiplet parts will stick with GF 14nm for the IOD while moving forward with TSMC N5 for the compute dies, so the 14nm wafers will be to provide a continuous supply of last gen products (i.e. Zen 3-based products). I get that IO doesn't scale well, but in its current form the 14nm IODs take up more power than we'd all like. AMD knows that they need to address the power as well, so moving the IOD to the former cutting edge node makes the most financial sense.
The GF contract is for older and lower end product and some semi-custom work AMD still does. IOD going to N6 makes a lot of sense because it's the same design rules, plus N7 would have more fab space as Zen 4 begins manufacturing some time in April-June 2022. No idea what node a console refresh will use, but possibly N6 to keep costs down. Also no idea if by then the IOD would be shrunk again or another approach would be crafted and thus N5 be used while processors would be on either 3 or 2 nm.
 

sandorski

No Lifer
Oct 10, 1999
70,085
5,618
126
Unless GloFo has some talented Engineers that would benefit Intel, this feels like a consolidation of the Has-Been Foundries. Not sure how this helps anyone.
 

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
Sounds like Intel is looking for customers for their IDM 2.0 strategy. Let's see how this plays out.

$30 billion to buy GF's customer base? Somehow I doubt that makes any financial sense.

Unless GloFo has some talented Engineers that would benefit Intel, this feels like a consolidation of the Has-Been Foundries. Not sure how this helps anyone.

GF has (or at least had) a bunch of ex-IBM people, and Intel is currently collaborating with IBM's remaining semicon research division. Might make some sense to bring them all together.

Well this could be troublesome if AMD can't switch to Samsung for IODs.

I agree with @Mopetar that Intel will be forced to honor the existing WSA along with extension clauses.

All that aside, it's a bit stunning that Intel sees GF as a better use of $30 billion than investing in their own fabs/fab research. What do they really want out of this deal? 12FDX? Access to some old modified Samsung nodes? Does GF have any outstanding orders for EUV equipment we don't know about yet?
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
All that aside, it's a bit stunning that Intel sees GF as a better use of $30 billion than investing in their own fabs/fab research. What do they really want out of this deal? 12FDX? Access to some old modified Samsung nodes? Does GF have any outstanding orders for EUV equipment we don't know about yet?
Intel is heavily lobbying both US and EU as the knight in shining armor that could restore balance to global manufacturing.

nfo.png

The first thing the knight does is... make sure no other local adventurers present themselves before the court.