Nvidia halts testing of 18A process

[prime007]

Happy holidays, everyone!

It's been a while but wanted to contribute something small to this forum I read practically everyday!

Intel Corporation stock dropped 2% Wednesday after Reuters reported that Nvidia stopped testing Intel’s 18A chip manufacturing process. The stock opened as much as 3.9% lower before erasing some losses.

Accordingly, Nvidia had been evaluating whether it could manufacture its chips using Intel’s advanced 18A production technology but decided not to move forward with the tests.

Related articles below:

https://finance.yahoo.com/news/intel-shares-slide-report-nvidia-120053021.html

https://finance.yahoo.com/news/nvidia-tested-making-chips-intel-113527089.html

https://finance.yahoo.com/news/uh-oh-did-nvidia-just-165713505.html

 

[siliconbruh999]

Happy holidays, everyone!

It's been a while but wanted to contribute something small to this forum I read practically everyday!

It appears Nvidia will not move forward with Intel's 18A process node.

Related articles below:

https://finance.yahoo.com/news/intel-shares-slide-report-nvidia-120053021.html

https://finance.yahoo.com/news/nvidia-tested-making-chips-intel-113527089.html

https://finance.yahoo.com/news/uh-oh-did-nvidia-just-165713505.html

Well Intel themselves said 18A will be internal only node

 

[MKWVentures]

Well Intel themselves said 18A will be internal only node

the revisionist history is quite stunning here.

2021: 18A was the external version of the Internal node. 50 products being testing. Everyone is looking at it. Whales and prepayments coming in. 5B in external foundry revenue in 2026. Fabs planned everywhere

2024: Internal version (20A) is cancelled and product moved to TSMC. This is because 18A was doing so well, there was no need for 20A.

1H2025: No one (except USG with 2030 volume) external has signed up. LBT says maybe the external foundry fanstasy needs to stop

2H2025: Apple, Google, Nvidia Looking at it. 18A will win.. ... Risk production started in April. Panther lake Launch in 2025 Yay!

18A is "internal only" because no one external wants to use it as of today. Its pretty simple.

Everyone looks at options, then they make decisions and have backup plans based on reality.

Perhaps we should track external tapeouts for products and external revenue instead of these "news articles"

We know there is no significant (defined as greater than half a fab) external foundry volume until at least 2028.

 

[Artificer60]

2024: Internal version (20A) is cancelled and product moved to TSMC. This is because 18A was doing so well, there was no need for 20A.

We know there is no significant (defined as greater than half a fab) external foundry volume until at least 2028.

I've said it before and I'll say it again. There was a need for 20A. It was a necessary step in the development of 18A. Even in Gelsinger's Pollyanna world there were only going to be a few SKU's run on 20A. In fact, what you are calling revisionist history is really only a replacement of Gelsinger's reality with the one the rest of us live in.

One of the biggest issues with 18A is that it was still designed as an Intel first node. It is my understanding that there are still many issues with the transistor library that make it less desirable for others to use it.

As to the 2028 claim, I'm assuming you are basing that on potential 14A adoption?

 

[MKWVentures]

I've said it before and I'll say it again. There was a need for 20A. It was a necessary step in the development of 18A. Even in Gelsinger's Pollyanna world there were only going to be a few SKU's run on 20A. In fact, what you are calling revisionist history is really only a replacement of Gelsinger's reality with the one the rest of us live in.

Great Comment. I agree on need for 20A. And reasons for not ramping it

One of the biggest issues with 18A is that it was still designed as an Intel first node. It is my understanding that there are still many issues with the transistor library that make it less desirable for others to use it.

I just respond to what the IFS people said to us. I think @Scotten Jones had some comments on the whole BSPD as well. If 18A was not great for external, then those "we were misled" lawsuits were correct.

As to the 2028 claim, I'm assuming you are basing that on potential 14A adoption?

2028 is not likely to be 14A volume. 2028 18A assuming that companies have started designing today on it and that Intel demonstrates 18A capability and stability. I am not sure how anything could be earlier than that. The finances on 18A are a serious challenge.

Again if we just focus on reality (tapeouts and revenue), I think we will be better off. Intel knows what is actually happening .... lets see what they tell us.

 

[Daniel Nenni]

Happy holidays, everyone!

It's been a while but wanted to contribute something small to this forum I read practically everyday!

Intel Corporation stock dropped 2% Wednesday after Reuters reported that Nvidia stopped testing Intel’s 18A chip manufacturing process. The stock opened as much as 3.9% lower before erasing some losses.

Happy Holidays! It really was a great one this year. Food, family, friends, boating, definitely one of the better holidays!

Nvidia was never on the intel 18A customer short list as far as I know. Nvidia and AMD are so tight with TSMC I don't see it ever happening for CPU/GPUs and other big chips. Same with AMD. Jensen Huang, Lisa Su, and CC Wei are very close friends and compatriots. TSMC N2 has a record amount of tape-outs coming, even more than N3 if you can believe that. This all points to a stronger NOT TSMC market in the years to come.

Intel 18A will have external foundry customers in 2026. I know of 4, hopefully there will be more. These will be in HVM in the 2028-2030 time frame. This update is from IEDM in early December. Hopefully none will drop out but anything can happen before taping-out. We should hear more on the next investor call. I really am expecting good things from Lip-Bu and Intel in 2026.

Samsung 2nm will have some customers as well but nothing earth shattering. Elon Musk is using both TSMC N2 and Samsung 2nm but my guess is that the bulk of the volume will go to TSMC. I also believe Elon will do a deal with Intel Foundry of some sort. We shall see. QCOM will make some Samsung 2nm chips for Samsung but most of the wafer volume will be TSMC N3 and N2 moving forward.

Bottom line: The NOT TSMC business is alive and well. Intel Foundry will lead this market segment with 18A and 14A. Just my opinion of course.

 

[MKWVentures]

Happy Holidays! It really was a great one this year. Food, family, friends, boating, definitely one of the better holidays!

Nvidia was never on the intel 18A customer short list as far as I know. Nvidia and AMD are so tight with TSMC I don't see it ever happening for CPU/GPUs and other big chips. Same with AMD. Jensen Huang, Lisa Su, and CC Wei are very close friends and compatriots. TSMC N2 has a record amount of tape-outs coming, even more than N3 if you can believe that. This all points to a stronger NOT TSMC market in the years to come.

Intel 18A will have external foundry customers in 2026. I know of 4, hopefully there will be more. These will be in HVM in the 2028-2030 time frame. This update is from IEDM in early December. Hopefully none will drop out but anything can happen before taping-out. We should hear more on the next investor call. I really am expecting good things from Lip-Bu and Intel in 2026.

Samsung 2nm will have some customers as well but nothing earth shattering. Elon Musk is using both TSMC N2 and Samsung 2nm but my guess is that the bulk of the volume will go to TSMC. I also believe Elon will do a deal with Intel Foundry of some sort. We shall see. QCOM will make some Samsung 2nm chips for Samsung but most of the wafer volume will be TSMC N3 and N2 moving forward.

Bottom line: The NOT TSMC business is alive and well. Intel Foundry will lead this market segment with 18A and 14A. Just my opinion of course.

Great Update. Thanks.

 

[Dangerman1337]

Nvidia was never on the intel 18A customer short list as far as I know. Nvidia and AMD are so tight with TSMC I don't see it ever happening for CPU/GPUs and other big chips. Same with AMD. Jensen Huang, Lisa Su, and CC Wei are very close friends and compatriots. TSMC N2 has a record amount of tape-outs coming, even more than N3 if you can believe that. This all points to a stronger NOT TSMC market in the years to come.

I definitely don't think AMD will have anything on IFS but Nvidia did like RTX 30 on Samsung 8nm so I wonder why 18A-P or 14A derivative wouldn't be good for RTX 60 and 70?

 

[Artificer60]

I just respond to what the IFS people said to us. I think @Scotten Jones had some comments on the whole BSPD as well. If 18A was not great for external, then those "we were misled" lawsuits were correct.

I don't think it was a case of being misleading. I honestly believe that Gelsinger didn't fully understand the market and didn't fully grasp the importance of having an offering with broader applicability. As I understand it 18A-P is intended to address those shortcomings, but it is still largely a band-aid.

14A is going to be the first process Intel is designing to service the broader market from the get go with the intent to fully meet industry standard design practices. They have learned a lot from 18A and I expect 14A to be a much more attractive offering.

 

[blueone]

I don't think it was a case of being misleading. I honestly believe that Gelsinger didn't fully understand the market and didn't fully grasp the importance of having an offering with broader applicability.

If you are correct, that level of hubris and lack of industry knowledge is unbelievable. I have attributed it to the IFS group not doing their homework, and not hiring the right people to drive the strategy (regardless of how they looked on paper). Come to think of it, I'm not sure which explanation is more damning.

 

[lispwright]

If you are correct, that level of hubris and lack of industry knowledge is unbelievable. I have attributed it to the IFS group not doing their homework, and not hiring the right people to drive the strategy (regardless of how they looked on paper).

According to unreliable sources (Reddit, Wikipedia ... and SemiWiki) Gary Patton, hired in 2019 from Global Foundries after three decades at IBM, was the Corporate Vice President and General Manager of Design Enablement (and later more), which included creation of the 18A PDK. And his retirement was announced August 1st.

Several comments about him are not favorable, here at the time of hiring:

"Patton is known in the industry for his ability to translate research and development into practice inside chip factories."

Like the 7nm program he shitcanned with McKinsey against the spirit of the deal with IBM?

I've never heard anything good about this guy from anyone on the ground, but maybe it makes sense if Intel is about to start emulating the layoff and outsourcing culture of 21st century IBM.

An "Exist50" on Reddit who quoted the above finished a comment with:

And that prediction seems to have proved prophetic. He grew Intel's PDK team mostly in India (where it only had a minority presence prior), and now they're in the layoff stage, which is naturally going to disproportionately impact the "high cost geo" employees, regardless of relative productivity.

And on the retirement announcement:

I know Gary, shocked he stayed as long as he did. He definitely had a window seat. He is past retirement age.

It's been obvious to me since the 1990s that Intel's highest level technical management was very uneven, we know it failed hard with 10 nm putting the company in existential danger, then with PG's conspicuous failure to fire a lot of these people we should not be surprised if that also was a favor in failing to get any whales for 18A.

 

[tomatoma]

According to unreliable sources (Reddit, Wikipedia ... and SemiWiki) Gary Patton, hired in 2019 from Global Foundries after three decades at IBM, was the Corporate Vice President and General Manager of Design Enablement (and later more), which included creation of the 18A PDK. And his retirement was announced August 1st.

Several comments about him are not favorable, here at the time of hiring:An "Exist50" on Reddit who quoted the above finished a comment with:And on the retirement announcement:It's been obvious to me since the 1990s that Intel's highest level technical management was very uneven, we know it failed hard with 10 nm putting the company in existential danger, then with PG's conspicuous failure to fire a lot of these people we should not be surprised if that also was a favor in failing to get any whales for 18A.

This is correct

 

[tomatoma]

the revisionist history is quite stunning here.

2021: 18A was the external version of the Internal node. 50 products being testing. Everyone is looking at it. Whales and prepayments coming in. 5B in external foundry revenue in 2026. Fabs planned everywhere

2024: Internal version (20A) is cancelled and product moved to TSMC. This is because 18A was doing so well, there was no need for 20A.

1H2025: No one (except USG with 2030 volume) external has signed up. LBT says maybe the external foundry fanstasy needs to stop

2H2025: Apple, Google, Nvidia Looking at it. 18A will win.. ... Risk production started in April. Panther lake Launch in 2025 Yay!

18A is "internal only" because no one external wants to use it as of today. Its pretty simple.

Everyone looks at options, then they make decisions and have backup plans based on reality.

Perhaps we should track external tapeouts for products and external revenue instead of these "news articles"

We know there is no significant (defined as greater than half a fab) external foundry volume until at least 2028.

Intel is truly a bunch of incompetent people, they should just abandon Fab, but even if they do, Intel itself will probably go bankrupt eventually lol

 

[tomatoma]

According to independent information, Intel will be 100% dependent on external foundries by 2030, which will spell the end of semiconductor manufacturing in the US lol.