The AMD Execution Thread [2007 - 2017]

[DavidGraham]

Points to consider:

Since total shipments are often down in Q2 anyway, Q2 is the least important quarter for long term trends. It affects the one with largest share more than the one with the smallest.

JPR didn't explicitly give the numbers for YoY which is more important from a unit shipment POV as you then can compare it with as little seasonal shifts as possible.

Still doesn't account for product launch, lack of new products, supply limited, price cuts ..etc and you are still limiting your comparison to quarters, a true year over year scale means having to measure the whole year not just a single quarter to try to account for all possible variable factors.

And there is nothing remarkable about this. That's how AMD operated in every Q2 since 2011, sometimes they remain flat, sometimes they fall in shipments, this is one of those times where they gained nothing. They virtually remained flat compared to previous quarter.

Which brings me to the crux of the matter, people will disagree on which is important, you value YoY more, I do not. I consider QoQ much more important since we are measuring market share in a given select period, If you want to look more broadly, then why not look at the whole year since a given quarter? for example Q2 2015: Q 2016, market share is still 80% 20% split.

The point of these numbers is to paint a picture of how a company is doing (well/bad). AMD didn't gain marketshare in Q2 because they operated exceptionally well, historically they remain flat in Q2 or dipped a notch. Which what they did this time, they remained flat. Only variable here is NV dipping sharply due to clearing stocks.

 

[Anarchist4000]

A few years ago the entire discrete market contracted resulting in Nvidia selling roughly the same number of cards and AMD less. That's when we saw this huge skew in marketshare. So your argument would be that Nvidia's 80% share is irrelevant because AMD is just clearing inventory I guess?

 

[Razor1]

You are mixing two different aspects of the market together which have to be taken apart and looked at separately. Both parts separately you see the same trend occurring for each individual IHV, but place them together you get a spike in marketshare uptake by AMD.

This is exactly why I stated in this very thread I can't see AMD's numbers like they stated in their conference call unless I mixed and matched JPR with Mercury.

This is an anomalous data point when you place them together because you can't see what really happened. Now lets say since Q3 is a strong Q, nV regains its momentum (this is what most likely will happen) and goes above the lost shipments it had of the 3 million and AMD sells just a bit more because of Polaris. At the end AMD ends up loosing marketshare from this quarter.

How many people do you think will think Pascal "out shipped" Polaris? The reality might be Polaris did just as well as Pascal in total volume growth though.

 

[CSI PC]

AIBs will need to start shipments well over a week before a launch in order to reach the various retailers around the world in time for launch. That is unless they want to ship via air, which no AIB is going to do unless its a dire emergency. So Polaris shipments in Q2 2016 will likely represent somewhere around 2-3 weeks worth of shipments for Rx 480 and less for Rx 470/460.

When compared to Mercury, which tracks shipments from AMD to partners, Q2 2016 is going to include over a months worth of Polaris shipments.

Regards,
SB

I agree, but that initial batch dried up as quickly as the 1080 and 1070s from Nvidia, albeit that AMD had around double the number of 480s to 1080s at some retailers on launch day (critical point), and Nvidia was drip feeding better than AMD in the 1st 3 weeks of launch, and after that we see more custom AIB from Nvidia than we did from AMD for the Polaris/Pascal comparison.
I am allowing for some leeway where both companies had some AIB models and not others, some being bought asap, and quite a few consumers' pre-order above say another AIB partner/etc.

Putting that 2-3 weeks worth of shipments into perspective; the largest available stock of 480s in the UK was from Overclockers and they had 2000 units, that was slow to be repopulated and that next batch was after June.
On launch day Overclockers had 500 units of 1080 and that was followed up by another 1000 units with 10 days and then giving a total around 2000 units up to one week left in June, which then continued with constant feed of models and then expanded with quite a few options for custom models.
For 1070 they had more units and started selling 11th June, 18 days before the 480.
The problem though is that many wanted a very specific custom model and reluctant to switch, you could say the same about 480 as the 1st custom model was the Sapphire Nitro versions and not everyone wanted that but still sold strongly as there was little other options.
Overclockers said that for launch (over 1st 2-3 weeks the Nvidia cards sold better than 480 and probably down to more selection even if it was slower for the 1st 5 days).
https://forums.overclockers.co.uk/showpost.php?p=29653512&postcount=4754
For context Gibbo is more of an AMD fan than Nvidia, and I did used to have later posts backing up that one I linked.
I like to use Overclockers for this comparison because they use multiple distributor/sales channels to ensure they have the largest stock at launch for both AMD and Nvidia.

So allowing for 2-3 weeks worth of shipments that made up the initial AMD launch batch for June, it still would not influence too greatly those market share figures IMO due to being balanced by Pascal sales that were limited on launch day but picked up after that and in good flow by end of June for both 1070 and 1080.
Key context is relying upon the larger retailers that use multiple distributors to gain models even if it costs them more; Overclockers bought their launch models also from international distributors/channels while some other very large UK retailers would not do that, and I assume that happens in all regions.

Expanding upon your tracked shipments thoughts, could it also be based upon booked orders (manufacturer not consumer), which would be very large for 480/470?
If it is solely shipments, then I still think it is the older models that heavily influenced the market shift possibly due to the low level API narrative and DX12 performance for upper mid-enthusiast 390/390x compared to 970/980 when all of these models started to drop prices.
This is also impacted by the massive drop in consumer purchasing for GPUs for that period, which will become clearer in Q3 IMO with the transition to Pascal/Polaris becoming more complete.
I would think there was about the same amount of 1080s shipped/sold to 480s up to the end of June (bearing in mind that only allows for the initial launch shipment for the 480), but next quarter should let us know how it worked out.

Cheers

 

[Anarchist4000]

It wouldn't be surprising if the 470 was dominant on OEM sales. Considering we're just getting through back to school, I'd have to imagine that's where a lot of volume has been going lately. OEMs would likely be purchasing in Q2 (more towards the end) and probably reflected in the current share numbers we've seen.

 

[Silent_Buddha]

Points to consider:

Since total shipments are often down in Q2 anyway, Q2 is the least important quarter for long term trends. It affects the one with largest share more than the one with the smallest.


Still doesn't account for product launch, lack of new products, supply limited, price cuts ..etc and you are still limiting your comparison to quarters, a true year over year scale means having to measure the whole year not just a single quarter to try to account for all possible variable factors.

And there is nothing remarkable about this. That's how AMD operated in every Q2 since 2011, sometimes they remain flat, sometimes they fall in shipments, this is one of those times where they gained nothing. They virtually remained flat compared to previous quarter.

Which brings me to the crux of the matter, people will disagree on which is important, you value YoY more, I do not. I consider QoQ much more important since we are measuring market share in a given select period, If you want to look more broadly, then why not look at the whole year since a given quarter? for example Q2 2015: Q 2016, market share is still 80% 20% split.

The point of these numbers is to paint a picture of how a company is doing (well/bad). AMD didn't gain marketshare in Q2 because they operated exceptionally well, historically they remain flat in Q2 or dipped a notch. Which what they did this time, they remained flat. Only variable here is NV dipping sharply due to clearing stocks.

Really? Lets take a look at some data again. Since you claim this happens to AMD every Q2 since 2011.

And we'll even look at your preferred metric, Quarter to Quarter, which is mostly meaningless.

In terms of volume AIB shipments of discrete desktop GPUs.

• Q2 2011, AMD shipments decreased by 1.28 million units. A 16.6% drop.
• Q2 2012, AMD shipments decreased by 0.01 million units. A 0.2% drop.
• Q2 2013, AMD shipments increased by 0.04 million units. A 0.7% increase.
• Q2 2014, AMD shipments decreased by 0.54 million units. An 11.0% drop.
• Q2 2015, AMD shipments decreased by 0.85 million units. A 33.5% drop.
• Q2 2016, AMD shipments increased by 0.10 million units. A 3.7% increase.

Yup, that shows a clear trend that since Q2 2011, AMD is relatively unaffected by Q1 to Q2 variation. Erm, wait, no it doesn't. The trend for AMD is that volume typically drops from Q1 to Q2. If you go back before 2011, that trend continues as well.

Did you even bother to look at the data?

Now lets look at the YoY trend since 2011.

• Q2 2011, AMD shipments decreased by 1.5 million units. An 18.8% drop.
• Q2 2012, AMD shipments decreased by 0.54 million units. An 8.3% drop.
• Q2 2013, AMD shipments decreased by 0.61 million units. A 10.2% drop.
• Q2 2014, AMD shipments decreased by 0.96 million units. An 18.0% drop.
• Q2 2015, AMD shipments decreased by 2.67 million units. A 61.2% drop.
• Q2 2016, AMD shipments increased by 1.13 million units. A 66.9% increase.

Both NVidia and AMD were dropping shipments YoY, although NVidia did have some YoY numbers that increased. This mirrors the general YoY decrease in overall desktop discrete GPU shipments. However, NVidia had to this point generally lost less shipments YoY than AMD, which is reflected in the fact that from 2010-2014 they gradually increased their share of the market. While from 2014-2015 their share of the market exploded.

Looking over that data, it's interesting to see that NVidia's market share exploded due to the effect of Maxwell's release into the market.

However, even more interestingly, Maxwell 2's release into the market saw Nvidia gradually start to lose market share to AMD.

Anyway, as I mentioned previously it's certainly notable that AMD was able to not only gain that much market share, but that AMD AIB partners were able to increase shipments by that much in a quarter that typically sees flat or reduced shipments compared to the previous quarter.

Regards,
SB

 

[Razor1]

It wouldn't be surprising if the 470 was dominant on OEM sales. Considering we're just getting through back to school, I'd have to imagine that's where a lot of volume has been going lately. OEMs would likely be purchasing in Q2 (more towards the end) and probably reflected in the current share numbers we've seen.

Back to school already happening haven't seen one rx 470 in Dell, HP or Lenovo. They better get them in there soon.

Edit, yes have seen m470 but that isn't a 470 desktop chip so that doesn't count

 

[Razor1]

Really? Lets take a look at some data again. Since you claim this happens to AMD every Q2 since 2011.

And we'll even look at your preferred metric, Quarter to Quarter, which is mostly meaningless.

In terms of volume AIB shipments of discrete desktop GPUs.

• Q2 2011, AMD shipments decreased by 1.28 million units. A 16.6% drop.
• Q2 2012, AMD shipments decreased by 0.01 million units. A 0.2% drop.
• Q2 2013, AMD shipments increased by 0.04 million units. A 0.7% increase.
• Q2 2014, AMD shipments decreased by 0.54 million units. An 11.0% drop.
• Q2 2015, AMD shipments decreased by 0.85 million units. A 33.5% drop.
• Q2 2016, AMD shipments increased by 0.10 million units. A 3.7% increase.

Yup, that shows a clear trend that since Q2 2011, AMD is relatively unaffected by Q1 to Q2 variation. Erm, wait, no it doesn't. The trend for AMD is that volume typically drops from Q1 to Q2. If you go back before 2011, that trend continues as well.

Did you even bother to look at the data?

Now lets look at the YoY trend since 2011.

• Q2 2011, AMD shipments decreased by 1.5 million units. An 18.8% drop.
• Q2 2012, AMD shipments decreased by 0.54 million units. An 8.3% drop.
• Q2 2013, AMD shipments decreased by 0.61 million units. A 10.2% drop.
• Q2 2014, AMD shipments decreased by 0.96 million units. An 18.0% drop.
• Q2 2015, AMD shipments decreased by 2.67 million units. A 61.2% drop.
• Q2 2016, AMD shipments increased by 1.13 million units. A 66.9% increase.

Both NVidia and AMD were dropping shipments YoY, although NVidia did have some YoY numbers that increased. This mirrors the general YoY decrease in overall desktop discrete GPU shipments. However, NVidia had to this point generally lost less shipments YoY than AMD, which is reflected in the fact that from 2010-2014 they gradually increased their share of the market. While from 2014-2015 their share of the market exploded.

Looking over that data, it's interesting to see that NVidia's market share exploded due to the effect of Maxwell's release into the market.

However, even more interestingly, Maxwell 2's release into the market saw Nvidia gradually start to lose market share to AMD.

Anyway, as I mentioned previously it's certainly notable that AMD was able to not only gain that much market share, but that AMD AIB partners were able to increase shipments by that much in a quarter that typically sees flat or reduced shipments compared to the previous quarter.

Regards,
SB

SB, what is being said is generally true, when a company with a significant market share in any given market if demand drops their shares will drop more then a subdued competitor. This is because the subdued competitor, shops are already allocating their products tightly. They are more worried about left over stock for them than the ones they know will be selling.

Its a generalization but just ask any inventory manager they will tell you the same thing. And its logical.

So pretty much what happened was, retailers had more left over stock from Maxwell 2 from the previous quarter or how ever they do their inventory checks are, and told their AIB reps, sorry can't order any more this time. Still have left over stock. If you can get us more Pascal we will take those though, but being in tight supply they couldn't get those. Rock... Hard place....

 

[Silent_Buddha]

So pretty much what happened was, retailers had more left over stock from Maxwell 2 from the previous quarter or how ever they do their inventory checks are, and told their AIB reps, sorry can't order any more this time. Still have left over stock. If you can get us more Pascal we will take those though, but being in tight supply they couldn't get those. Rock... Hard place....

I kind of thought people would pick that up when I mentioned that the NVidia and their AIBs had to correct channel inventory. I guess I should have gone into more detail on that.

In comparison to previous years, channel inventory was overstuffed with Maxwell 2 cards which led lower than normal shipments of Maxwell 2 cards during that quarter. Which Pascal couldn't make up for. It isn't rare for channel corrections to happen in Q2, but this is the largest correction for NVidia in the past decade. That's magnified even more by the fact that the GPU market continues to shrink, making it that much more significant.

But what you said WRT AMD doesn't hold for this quarter. If that was all that happened then AMD would have had a minor gain at best or a major decrease (more likely) in YoY shipments for Q2. Combined with the impending release of Polaris which would have curbed demand for midrange (the bulk of the GPU market) parts from AMD we should have seen a reduction in the number of GPU shipments from AIB partners.

I'm not even talking market share at this point, but purely shipments from AIBs. And a decrease in shipments from NVidia AIB partners isn't going to affect that as there was at no point a shortage of NVidia cards in the retail channel, and hence retail stores. IE - a shortage of NVidia GPUs didn't lead to more demand for AMD GPUs because there was never a shortage of NVidia GPUs.

But the strange thing is that despite AMD being at a worse performance advantage versus Maxwell 2 than they were versus Maxwell, AMD has been steadily increasing GPU shipments ever since Maxwell 2 launched.

And for Q2 2016, despite the launch of Pascal (impacting the performance and enthusiast markets) and the impending release of Polaris (impacting the midrange market), shipments actually went up from Q1 to Q2 which is very rare for either company.

Regards,
SB

 

[Razor1]

I kind of thought people would pick that up when I mentioned that the NVidia and their AIBs had to correct channel inventory. I guess I should have gone into more detail on that.

In comparison to previous years, channel inventory was overstuffed with Maxwell 2 cards which led lower than normal shipments of Maxwell 2 cards during that quarter. Which Pascal couldn't make up for. It isn't rare for channel corrections to happen in Q2, but this is the largest correction for NVidia in the past decade. That's magnified even more by the fact that the GPU market continues to shrink, making it that much more significant.

But what you said WRT AMD doesn't hold for this quarter. If that was all that happened then AMD would have had a minor gain at best or a major decrease (more likely) in YoY shipments for Q2. Combined with the impending release of Polaris which would have curbed demand for midrange (the bulk of the GPU market) parts from AMD we should have seen a reduction in the number of GPU shipments from AIB partners.

I'm not even talking market share at this point, but purely shipments from AIBs. And a decrease in shipments from NVidia AIB partners isn't going to affect that as there was at no point a shortage of NVidia cards in the retail channel, and hence retail stores. IE - a shortage of NVidia GPUs didn't lead to more demand for AMD GPUs because there was never a shortage of NVidia GPUs.

But the strange thing is that despite AMD being at a worse performance advantage versus Maxwell 2 than they were versus Maxwell, AMD has been steadily increasing GPU shipments ever since Maxwell 2 launched.

And for Q2 2016, despite the launch of Pascal (impacting the performance and enthusiast markets) and the impending release of Polaris (impacting the midrange market), shipments actually went up from Q1 to Q2 which is very rare for either company.

Regards,
SB

Ah ok, yeah nV and AMD do have to correct channel inventory, but when you have supply constraints which I don't care how nV spins it, they had/have supply constraints from their end and it isn't because of higher than expected demand lol sorry but 1% of the market is not going to become 10% over night, that is what they are making it sound like and that is just not realistic. AIB's aren't going to be fully aware of supply issues until they actually order. nV will not tell those types of things early on because they can't afford to do something like that. I have seen this done in business many times where a company will say sure we can do it. Because the moment they say they can't those "partners" will go else where. Of course nV's larger partners can't as they are exclusive and there really is only one of those EVGA and the smaller ones, but the others will. Its all about money at the end doesn't matter where the product comes from.

Sorry miss understood about the marketshare bit....

The uptake in AMD sales is not rare when you look at their bottom lines, they are getting rid of stock as much as possible by dropping prices. And this is why margins went down by 1% too.

Did we expect Dell and HP to pick up the Fury X now, its a dead in the water card, err bad pun lol, specially when its 150 bucks more than a founders edition 1080?

 

[CarstenS]

What I found interesting in this regard was that AMDs R9 3xx/3xxX generation seem to have had their clearance sale in spring already with prices climbing again when Polaris was released. While for Nvidia, prices stayed more or less constant until after Pascal was in the market. What I had originally attributed to already low channel inventory seems in retrospect rather like a backfiring attempt to cover for potential shortages of Pascal while not having to pump out massive amounts of cards for low prices which would in turn also reduce Pascal demand in the aftermath.

 

[DavidGraham]

Really? Lets take a look at some data again. Since you claim this happens to AMD every Q2 since 2011.

Apparently you didn't read my post. The numbers you posted just foster my argument, I said they either remain flat or dip a notch. In 2 quarters they were virtually flat (-0.2%/ +0.7%), 2 other quarters they dipped a notch (-11% / -16%)! Only in 2015 they dropped massively, In 2016 now they are back to their normal rhythm! Did you even read what I posted?

 

[CSI PC]

It wouldn't be surprising if the 470 was dominant on OEM sales. Considering we're just getting through back to school, I'd have to imagine that's where a lot of volume has been going lately. OEMs would likely be purchasing in Q2 (more towards the end) and probably reflected in the current share numbers we've seen.

Yep I agree with that and was thinking the same, although the 470 did not launch until the very end of July so even that would had limited shipments in June for end of Q2.
Which is why I am wondering if some figures also include booked orders, which would be more applicable to OEMs and IHVs rather than consumer retail.

Cheers

 

[Krumme]

Looking back at the deal when AMD sold GF and the corresponding wsa agreements over the years.
How would you guys structure the deal back then /put together the incentive structure to get the most for both gf and amd? What would you do differently than what was done?

 

[3dilettante]

Per the then-CEO, the deal's initial price was already somewhat more favorable to AMD because they had convinced ATIC that there were other bidders for the fabs.
It's possible this lead to some blowback once the buyers realized this. If AMD had been more successful, then everyone would have made more money and AMD could have shrugged it off as long as it kept to the optimistic side of its WSA and didn't need more financing.
Instead, AMD wound up in having to go back for more money and concessions. However, there's no guarantee that a lack of ill will would necessarily make an investor forgiving if money is on the line, although perhaps they'd consider AMD's projections more credible.

The deal was structured to work around a thicket of legal issues with Intel, give AMD a chance to not implode back then, and protect a sovereign investment fund from buying a multi-billion dollar investment that became instantly worthless if AMD quickly transitioned to another foundry.
I am not sure I could credibly come up with a "better" arrangement when so many of the things in the deal are not public, and "better" for one party would most likely make it a non-starter for everyone else.

 

[Krumme]

How do you cope when yield or freq is better or lower than in your scope of the deal? Whose fault or benefit is it. I mean its very hard to specify.
Especially when process problems meet arch problems. And gf process havnt attracted any customers besides amd in real numbers and bd was a failure. It qualify for problems.

I simply think the deal was structured wrong when looking how the new wsa is different. The guilty was Hector as he had the insight to know. But he probably got a fat bonus i recon but thats another challenge in these cases.

What about this idea:

Had they mutually shared lots of stock options from day one and until 2024 where the agreement stops. Shared interest thats goes far beyond what a deal can specify.

Instead of agreeing on a specified wafer amount - Using market thinking like the former soviet - why not solely use the fines for using tsmc/samsung capacity. Even if the fines for using those were huge?

Perhaps it had eg been better for gf to go for the arm market. The money would have been there for doing that from Mubadala because if amd could use tsmc better they wouldnt constantly need that cash or consessions.

Besides i guess the wsa more fixed wafer agreement makes communication difficult. But thats for another discussion.

 

[Silent_Buddha]

Apparently you didn't read my post. The numbers you posted just foster my argument, I said they either remain flat or dip a notch. In 2 quarters they were virtually flat (-0.2%/ +0.7%), 2 other quarters they dipped a notch (-11% / -16%)! Only in 2015 they dropped massively, In 2016 now they are back to their normal rhythm! Did you even read what I posted?

Sure I'll give you that. But a 3.7% increase is anything but flat. And you are still ignoring the forest for the tree.

Here's a hint. Look at the quarterly trends (trend per 4 quarters) leading up to this one for AMD. That also backs up the YoY trend that you see. You are so focused on the one metric that you favor, which has almost no meaning on its own, that you ignore all the other data available.

It's the same as focusing only on market share and ignoring shipment data. Which some people have a tendency to do if it favors their particular company of choice. Or ignoring it if it doesn't.

Looking at the same quarterly trends, what we see is that Nvidia is likely to rebound sharply in shipments for Q3 (assuming shipments of Pascal can ramp up). With AMD, it's anyone's guess as their current 4 quarter trend doesn't match anything in the past decade. Hence, why I expect Nvidia to regain ground.

Regards,
SB

 

[Krumme]

Amd are in an econimic position where they need revenue above anything else. Speeding up oem sales is the way to get there fast because it is price sensitive market.

Nv is in a far better economic position and can prioritize profit and brand development. Retail and high cost high margin pascal is the way to do it.

To me the development we see makes sense and what both nv and amd does seems like the optimal in the situation.

1040 and 1050 will land in a higher price than 460. Titan is here and 1080ti is comming soon.

I expect amd share to rise a bit further or stay flat and nv profit to rise.

 

[3dilettante]

How do you cope when yield or freq is better or lower than in your scope of the deal? Whose fault or benefit is it. I mean its very hard to specify.
Especially when process problems meet arch problems. And gf process havnt attracted any customers besides amd in real numbers and bd was a failure. It qualify for problems.

AMD would live with those problems whether or not they owned the fabs. Up until the last update to the cross-licensing deal with Intel, beyond a minority share of production had to do so.
Bulldozer's problems were effectively baked in. GF didn't materialize from nowhere, it was AMD's fabs and staff, so it wouldn't be until post-32nm that most decisions weren't actually made by AMD prior.

Had they mutually shared lots of stock options from day one and until 2024 where the agreement stops. Shared interest thats goes far beyond what a deal can specify.

GF could not own anything of AMD's. It was a subsidiary that AMD had theoretically half control over. What would be possible now was not allowed then, and even then there were legal challenges.

Instead of agreeing on a specified wafer amount - Using market thinking like the former soviet - why not solely use the fines for using tsmc/samsung capacity. Even if the fines for using those were huge?

Fabrication is a massively expensive process with high capital costs and long lead times. Planning is necessary because underutilization goes immediately to the detriment of the manufacturer.

Perhaps it had eg been better for gf to go for the arm market. The money would have been there for doing that from Mubadala because if amd could use tsmc better they wouldnt constantly need that cash or consessions.

Mubadala owns GF, and were the ones tricked into paying more than they needed to, for an investment that in hindsight seems unlikely to realize the goals it was sold for.
Why would it pay AMD to make GF lose more money? Mubadala's stake in AMD is smaller.

 

[Krumme]

AMD would live with those problems whether or not they owned the fabs. Up until the last update to the cross-licensing deal with Intel, beyond a minority share of production had to do so.
Bulldozer's problems were effectively baked in. GF didn't materialize from nowhere, it was AMD's fabs and staff, so it wouldn't be until post-32nm that most decisions weren't actually made by AMD prior.


GF could not own anything of AMD's. It was a subsidiary that AMD had theoretically half control over. What would be possible now was not allowed then, and even then there were legal challenges.


Fabrication is a massively expensive process with high capital costs and long lead times. Planning is necessary because underutilization goes immediately to the detriment of the manufacturer.


Mubadala owns GF, and were the ones tricked into paying more than they needed to, for an investment that in hindsight seems unlikely to realize the goals it was sold for.
Why would it pay AMD to make GF lose more money? Mubadala's stake in AMD is smaller.

I am not inplying mubadale should pay more or less. That way of looking at it is uninteresting in my world and you have lawyers and accountants for that. I am just interested in how those agreements should be made to make most business for all.

There is a reason we see a change to the wsa today. Longer period, fines for using tsmc/ss and inderectly mubadala ownership in amd.

Another way to frame the question is why then those changes? And why is it possible now and not back then if it was even considered?