News TSMC and GlobalFoundries Dismiss All Patent Lawsuits, Announce 10-Year Cross-License - Tom's

[UsandThem]

https://www.tomshardware.com/news/t...-all-litigation-announce-patent-cross-license

Today GlobalFoundries and TSMC issued separate, but identical, press releases stating the companies have resolved all of their global patent disputes, resulting in the dismissal of all patent lawsuits between the two companies.

I'm actually really surprised it ended this quickly. When GlobalFoundries filed their lawsuit, and TSMC countered with one of their own, it appeared this would likely drag out for years.

 

[bbhaag]

Yeah I'm surprised it ended so quickly to and the settlement appears to be a solid advantage for GloFo. TSMC must have seen something that they knew they couldn't win and it would have cost them more in the long term. Either that or their business partners/contracts put them under immense pressure to settle this quickly and put it behind them.

 

[NostaSeronx]

It's more advantageous for TSMC as they get all the GloFo high mobility patents. That GlobalFoundries would only implement ever exclusively for IBM/AMD only.

So, its a win for TSMC as they can now implement custom processes for AMD w/ a new yet similar WSA as GloFo. Think 12FFN, but N7A/N5A/N3A and without GlobalFoundries awful production drive and upper management.

 

[Abwx]

It's more advantageous for TSMC as they get all the GloFo high mobility patents. That GlobalFoundries would only implement ever exclusively for IBM/AMD only.

So, its a win for TSMC as they can now implement custom processes for AMD w/ a new yet similar WSA as GloFo. Think 12FFN, but N7A/N5A/N3A and without GlobalFoundries awful production drive and upper management.

It s the other way around, TSMC will access to nothing, actually they likely agreed to pay royalties to GF to keep using advanced processes and their future derived iterations.

FTR GF did characterise the infringment by citing explicitely the relevant patents, on the other hand TSMC, in their counter suing, only did some sabre rattling without providing any patent as infriged by GF, and was just using this move as a mean to save face.

"We are pleased to have quickly reached this settlement that acknowledges the strength of our respective intellectual property. Today's announcement enables both of our companies to focus on innovation and to better serve our clients around the world," said Thomas Caulfield, CEO of GF, as cited in a press statement. "This agreement between GF and TSMC secures GF's ability to grow and is a win for the entire semiconductor industry which is at the core of today's global economy."

TSMC and Globalfoundries end legal disputes through patent cross-licensing

TSMC and Globalfoundries (GF) have announced they are dismissing all litigation between them as well as those that involve any of their customers. The companies said they have agreed to a broad life-of-patents cross-license to each other's worldwide existing semiconductor patents as well as...

www.digitimes.com

 

[DrMrLordX]

Hmmm.

Broad cross-licensing of patents? That's interesting. I wonder what will be the future effect of that agreement.

 

[VirtualLarry]

So, I assume that GlobalFoundries now has access to TSMC's process technology, and can live to "die another day" (James Bond reference).

 

[Asterox]

So, I assume that GlobalFoundries now has access to TSMC's process technology, and can live to "die another day" (James Bond reference).

Yes something like that , or you can conclude in future even more serious problems for Intel.

 

[Tuna-Fish]

So, I assume that GlobalFoundries now has access to TSMC's process technology, and can live to "die another day" (James Bond reference).

Just because you can legally use the patents does not mean you can fully use the technology. Intel and AMD have a full cross-licensing agreement about x86, yet there are still substantial differences between their cpu designs.

 

[Mopetar]

Just because you can legally use the patents does not mean you can fully use the technology. Intel and AMD have a full cross-licensing agreement about x86, yet there are still substantial differences between their cpu designs.

Yup. If it were just a matter of having access to a patent, countries like China could have incredibly advanced fabs of their own. It turns you that you need a lot of expertise and know-how in order to successfully implement all of that technology.

 

[thesmokingman]

Okay, maybe going fabless for Intel is the better route moving forward. There's no way they gonna catch up on process if these two are not at war.

 

[thesmokingman]

Just because you can legally use the patents does not mean you can fully use the technology. Intel and AMD have a full cross-licensing agreement about x86, yet there are still substantial differences between their cpu designs.

Ha, try telling that to Nvidia whom wishes they had a license to x86, which REALLY doesn't matter since they will never get one to AMD64, most likely.

 

[soresu]

Hmmm.

Broad cross-licensing of patents? That's interesting. I wonder what will be the future effect of that agreement.

Hopefully more capacity at smaller nodes, a reduction to 2 foundries from the previous 3 meant less total capacity for this generation.

 

[soresu]

Ha, try telling that to Nvidia whom wishes they had a license to x86

Their ARM core isn't so great, I doubt they would fair better with x86.

 

[Topweasel]

Just because you can legally use the patents does not mean you can fully use the technology. Intel and AMD have a full cross-licensing agreement about x86, yet there are still substantial differences between their cpu designs.

Generally they can use the same technology. There is a lot of negotiations that went on between AMD and Intel when the cross licensing deal went down. AMD and Intel could make the exact same CPU's in theory, just the only thing AMD doesn't get is the actual blueprints to Intel's CPU and vice versa. Way back when AMD was in a rough spot they also made an agreement on top of the cross licensing that they had to stop making Intel Platform compatible CPU's (death of Socket 7). They couldn't do even that now if they wanted to, even in the BD era, Intel is using to many AMD patents to break away from the agreement. As far as fabs go the tech is a lot more straight forward. They can't clone a TSMC fab (and at the 30 Billion TSMC is spending on a 3nm lab, they probably don't want to). But it gives them all the access to all the patented techniques that again are going to be pretty straight forward to move to each process.

For Goflo it keeps them in the game at a low cost. They can do slow transitions for their limited customers as equipment costs go down and for TSMC, Goflo is no longer a competitor in the lucrative bleeding edge contracts (where AMD or Apple are willing to take big hits on yields for a product right away, requiring them to purchase more wafers). Anything Goflo uses will be like 5 years behind them.

 

[EXCellR8]

I don't see the consumer base being affected by this much, if at all.

 

[darkswordsman17]

Yeah I'm surprised it ended so quickly to and the settlement appears to be a solid advantage for GloFo. TSMC must have seen something that they knew they couldn't win and it would have cost them more in the long term. Either that or their business partners/contracts put them under immense pressure to settle this quickly and put it behind them.

I was too. Well less that it ended quickly after TSMC responded in kind (I think they both likely have a case, it was one of those where the overall industry kinda shared stuff as it benefited everyone to operate like that). More that it ended how it did with a 10 year broad cross-licensing agreement.

I think they're seeing political stuff will start to play a bigger role, and likely sees value in potentially being able to produce chips in America and Europe that GF would enable. Plus this kinda enables TSMC to increase production (they seem to be struggling with that right now, and if their 7+ and 5nm are as popular as 7nm, then they'll need all they can get). GF likely gets some use out of some of their advanced fabs that they let languish after abandoning their 7nm. TSMC gets extra fab space for relatively cheap (compared to building new fabs). Keep in mind I'm not saying that's absolutely what is happening (as I feel like we don't know that yet), but that's the potential.

To me its a win all around. By the time the agreement is over we'll likely be looking at some big changes to the computer market, as that'll probably be getting to 1nm and tapping out what they can achieve via the usual means.

 

[NostaSeronx]

GlobalFoundries is still indefinite on FinFETs.

I fully expect September 2020 to be when they announce 7FDX. As FDSOI beyond 64CPP is faster than FinFETs and cheaper than Nanosheets. Monolithic 3D complexity is also lower with 7FDX compared to any other process. So, 5FDX can be two stacks of 7FDX, and 3FDX can be four stacks of 7FDX.

Skywater's stacked 90nm(3DSoC) is faster than bleeding edge, so GloFo definitely needs to make the jump to stacked logic.

 

[darkswordsman17]

I don't see the consumer base being affected by this much, if at all.

I do simply in that this should alleviate issues like tariffs and production capacity constraints. TSMC could move fabs to newer processes quicker, while GF lags a bit behind (so as TSMC moves to 5nm, GF would be taking over 7nm production, letting TSMC move fabs that otherwise might would have to have stuck at 7nm to keep up with demand). They could work licensing or sub-contracting deals or something so that they both benefit.

I don't think it'll drive prices down exactly, but it will help keep some of that stuff from driving prices up as much.

There's other potential aspects that I think we haven't even begun to see yet. Stuff like interposer tech and other that will start to play a big role, and will need a variety of process tech to implement.

 

[fkoehler]

So, I assume that GlobalFoundries now has access to TSMC's process technology, and can live to "die another day" (James Bond reference).

So, even if GF has access to TSMC's IP does that mean much if they can't afford the top-o-line ASML machinery to utilize it?
Sort of like giving Chevy the blueprints and go ahead to make a Bugatti.
Sure they can probably crib some stuff to improve their existing cars, however that doesn't mean they can actually start producing Bugatti's on their existing manf. lines and start competing with TSMC.
At least thats how I understand it.

 

[Kocicak]

We do not know the exact terms of the contract. The only real current impact is that they are not sueing each other any more.

Now it seems that the GF accusation had some substance behind it.

 

[Abwx]

Guess that TSMC has agreed to pay royalties till 2029, cross licencing is a vague term that allow the taiwanese firm to not lose face, a thing that is very important in chinese culture and wich cost nothing to GF to add to the agreement terms.

That being said it s likely that GF realised that they made a big mistake by not ramping their 7nm since it would had required 2-4bn to do so, wich they thought wouldnt worth it considering AMD difficult start with Zen 1/2, but if Q4 numbers are confirmed then AMD is in the course of more than 10bn revenue for 2020 with a lot of products still under their hood.

Renoir, next consoles semi custom chips, new GPUs, and of course existing products, are about all 7nm, and in hindsight GF would had cashed at least 4bn out of those expected 10bn and soaring, enough to amortize the ramping in a matter of barely three years.

 

[DrMrLordX]

For Goflo it keeps them in the game at a low cost. They can do slow transitions for their limited customers as equipment costs go down and for TSMC, Goflo is no longer a competitor in the lucrative bleeding edge contracts (where AMD or Apple are willing to take big hits on yields for a product right away, requiring them to purchase more wafers). Anything Goflo uses will be like 5 years behind them.

Maybe. It's more like . . . Glofo can improve some of their own processes, or toss out ones that are maybe not doing so well and replace them with adapted TSMC nodes.

We do not know the exact terms of the contract. The only real current impact is that they are not sueing each other any more.

That is basically true. We won't know until GF changes their process roadmap, if at all. Right now it's, what, 12nm+ and very low-key FDX rollouts?

@Abwx

I'm wondering if AMD will be able to use GF as a second source for some products. 7nm I/O dice perhaps? GF may be able to fix their "broken"/delayed 7nm node faster now.

 

[NostaSeronx]

Right now it's, what, 12nm+ and very low-key FDX rollouts?

12LP+ and 22FDX to 75K~100,000 wafers per month, after that it is 12FDX time, and sometime by mid-2022 or mid-2023, Dresden is doing 7FDX.

GF may be able to fix their "broken"/delayed 7nm node faster now.

7LP was never broken, they shipped good dies back. GlobalFoundries didn't keep IBM or AMD. No IBM or AMD means no process node. A node at GlobalFoundries "leading edge" => anything beyond 28nm; existed to only serve AMD or IBM.

 

[Abwx]

@Abwx

I'm wondering if AMD will be able to use GF as a second source for some products. 7nm I/O dice perhaps? GF may be able to fix their "broken"/delayed 7nm node faster now.

Their process is neither broken nor delayed, it s just that they knew that they wouldnt have the necessary volume to amortize the ramping cost, and in this respect they are more or less right looking at current AMD sales.

Being relegated as a second source or I/O provider is of the same wood, that is, not financiarly rewarding.

What s left for them is evantUALLY getting back orders from AMD as a primary source, knowing that TSMC has currently captured most of AMD orders.

Dunno what they are going to do, but for sure that at some point they ll have to get to a smaller node than 12nm if they want to keep being relevant, and so far the only readily available one is their 7nm, at said costs...

 

[DrMrLordX]

Their process is neither broken nor delayed, it s just that they knew that they wouldnt have the necessary volume to amortize the ramping cost, and in this respect they are more or less right looking at current AMD sales.

Let's not put on rose-colored glasses here. Their 7nm schedule was delayed even before they threw in the towel.

Being relegated as a second source or I/O provider is of the same wood, that is, not financiarly rewarding.

. . . they did it for Apple. You don't think they'd do it for AMD, especially if it meant keeping their chipset + I/O die business and maybe some of their GPU business as well?

What s left for them is evantUALLY getting back orders from AMD as a primary source, knowing that TSMC has currently captured most of AMD orders.

Doubtful. They will not have nodes in the future to compete with TSMC.