Potential CEO Candidate for Intel

[jabo]

I doubt that High NA will be featured in TSMC's A14, at least not in the first generation. Currently, top customers are primarily focusing on advanced packaging and HBM to enhance performance, making the High NA option relatively expensive and risky. Apple seems to be shifting its selling points to other areas, possibly AI. Qualcomm and Mediatek are aligned with Apple, while Nvidia and AMD require CoWoS. It would be quite amusing if Intel pressured TSMC to adopt High NA

These high NA machines are something like $350-$400M each. Intel has had at least one of these up and running now for about a year. A second one was installed about 6 months ago. Intel's Foundry Roadmap has some fuzzy dates on purpose, but suggests 14A on high NA being used in 2026, presumably late 2026, and then 14A-E in 2027, presumably late 2027. The far right year on the roadmap is 2027, under which shows 14A-E. Can this slip? Certainly. Is it likely to slip, odds are that it will. Intel also has a quote on their foundry site stating " High NA is foundational to cost-effective scaling", and that they are "previewing 14A" now. When do you guys think the first product with 14A will launch? Will this be NovaLake or whatever is after that? I can't imagine it being beyond whatever is after Nova Lake.

 

[Artificer60]

The Intel and TSMC foundry events are coming up in April. If TSMC does not mention HNA-EUV that means customers are not pressuring them for it. It really is all about the top 10 customers and what they want.

Maybe I'm just being thick here, but do customer's really care which tool you use? It is about getting good yields on an acceptable time frame with a price customers consider affordable, right? I don't believe that customers were telling TSMC "if you don't use EUV tools we won't buy your product". I suspect the only thing customers really care about is that you stick to the process roadmap you gave them to base their product roadmap on. TSMC felt they needed EUV to do that and Intel bet (foolishly as it turned out) on quad patterning. It seems to me it should be largely up to TSMC to decide which tools they need to hit those milestones.

 

[blueone]

Maybe I'm just being thick here, but do customer's really care which tool you use? It is about getting good yields on an acceptable time frame with a price customers consider affordable, right? I don't believe that customers were telling TSMC "if you don't use EUV tools we won't buy your product". I suspect the only thing customers really care about is that you stick to the process roadmap you gave them to base their product roadmap on. TSMC felt they needed EUV to do that and Intel bet (foolishly as it turned out) on quad patterning. It seems to me it should be largely up to TSMC to decide which tools they need to hit those milestones.

IMO, exactly correct!

 

[DanX]

These high NA machines are something like $350-$400M each. Intel has had at least one of these up and running now for about a year. A second one was installed about 6 months ago. Intel's Foundry Roadmap has some fuzzy dates on purpose, but suggests 14A on high NA being used in 2026, presumably late 2026, and then 14A-E in 2027, presumably late 2027. The far right year on the roadmap is 2027, under which shows 14A-E. Can this slip? Certainly. Is it likely to slip, odds are that it will. Intel also has a quote on their foundry site stating " High NA is foundational to cost-effective scaling", and that they are "previewing 14A" now. When do you guys think the first product with 14A will launch? Will this be NovaLake or whatever is after that? I can't imagine it being beyond whatever is after Nova Lake.

I think Intel should be more practical. Money talks, big money talks louder.
Why do you think Intel can surpass TSMC at 1nm when they couldn't do it at 5nm, even with more money.

 

[jabo]

I think Intel should be more practical. Money talks, big money talks louder.
Why do you think Intel can surpass TSMC at 1nm when they couldn't do it at 5nm, even with more money.

I am not necessarily saying Intel can surpass TSMC. But now that Intel has EUV up and running the playing field is level. When TSMC was doing N5, Intel still had no EUV up and running, so of course their Intel 7 node (on DUV) was not competitive. I'm saying Intel is back in the game and whatever nodes they are both producing later this year and going forward are likely to be similar in performance. I would guess some Intel nodes will be better, and some TSMC nodes will be better. Up until about 1.5 years ago, Intel was at a severe disadvantage by not having any EUV, now they do. And they also seem to have ~ 1 year headstart on high NA, which can only help their chances of surpassing. In addition to that, they have at least a 1 year headstart on backside power. It will be fun to watch them compete. It's good for the industry. Also, we have Samsung who could surprise us too.