News Intel might axe the 18A process node for foundry customers, essentially leaving TSMC with no rival — Intel reportedly to focus on 14A

[phead128]

So when Pat said: "I've bet the whole company on 18A", he really meant he bet the company on 14A and 10A.

 

[JRStern]

if you look at the high density EUV nodes they're significantly ahead of DUV

If something has half the yield is it ever really ahead?
OK the last DUV nodes had poor yield too, that's why I say have to go further back, to single mask.
Maybe 99% yield beats 15% yield in price/performance.
Only has to be half as fast.
Scale out instead of down.

 

[DHB]

So IFS loses billions of dollars by having all of their advanced node capacity full to the brim? I get that CapEx for new nodes is CRAZY expensive, but loans aren't going to get paid off when profits are in the negative billions. TSMC invests tens of billions per year on new nodes, but they're profitable.

I don't think so. Intel is its own customer is down on chip volume, so that's one hit right there.

I don't believe in this company at all at this point. Sell the future's concepts while literally selling off current developments that have reached fruition is not a business strategy that can continue on forever. Unless someone can show me hard concrete evidence otherwise, it's literally living on hope.

I know banks don't run this way, so how is Intel even able to continue reneging on their promises? I wouldn't loan them one penny if I was a business, and I sure as heck aren't about to invest in them.

How difficult/costly is it for designers to move to the next node? Is it less costly to move from Intel 18a to Intel 14a than to move from TSMC (N2) to Intel 14a for designers? If 14a is attractive enough will 18a be a “stepping stone“ to 14a and the next nodes? Will adopting earlier nodes (18a) make next nodes (14a) more “useful”/ better utilized by designers having experience with IFS?

 

[DHB]

If I said scrap it, that's not what I meant. They have what they have and it's booked, including the fab that hasn't come online yet. That's not all that he said though:
"According to the new report, by June, he began sharing with colleagues that the 18A manufacturing process — a technology designed to showcase Intel's manufacturing prowess — was losing appeal to outside customers, which is why he believed it made sense for the company to shift away from offering 18A and its performance-enhanced 18A-P version to foundry customers." - this article

If 18A is already losing appeal, why wouldn't 14A do the same at the same point in time (when it starts ramping)? Intel is finally going to magically hit this one out of the ballpark? It just feels like more kicking the can down the road. I mean, use 18A to springboard to 14A? After using 20A to springboard to 18A? Yeah, I guess it probably is an effective catch-up strategy, but it also comes with the additional costs and risk. IMO, IFS needs to show confidence in a product that they really put a lot of effort into and promoted. Obviously things do need to start shifting to a 14A focus, but that would be internal more than external.

This article mentioned a write-off and Reuters did as well:
"Potential write-off for "18A" process could cost hundreds of millions of dollars"

https://www.reuters.com/business/retail-consumer/intels-new-ceo-explores-big-shift-chip-manufacturing-business-2025-07-02/

It's not my words, lol.

Delivering Panther Lake on time and in quantity may make 18a more interesting to customers and prove capacity is available as well.

 

[bit_user]

If something has half the yield is it ever really ahead?

What has half the yield? What's your source on that?

Maybe 99% yield beats 15% yield in price/performance.

A node with 15% yield would be non-viable.

Only has to be half as fast.

It's a lot worse than that, because you're ignoring the density increases.

Scale out instead of down.

You can't. You're ignoring a fundamental dynamic that drove Moore's Law. Density is a key enabler of more performance, not only because you can fit more computing elements, but also because they can increase in complexity and use less power. Furthermore, communication becomes more efficient as density increases. This whole dynamic drove more than 5 decades of stellar advances in computing. If density didn't really matter, people wouldn't be so eager that they'd pay top dollar for it.

In case you haven't noticed, datacenters are being increasingly power and space-constrained. There's no way they'd go with a less dense and less efficient option. One of the reasons Datacenters do upgrades is to increase density and efficiency.

By limiting access to TSMC nodes, China is effectively being frozen on older, less-efficient, less-dense nodes and they're extremely unhappy about it. And if anyone out there is optimizing DUV nodes, it's definitely going to be SMIC.

 

[bit_user]

How difficult/costly is it for designers to move to the next node?

By this point, it should be more expensive to move between node generations than between foundries.

Is it less costly to move from Intel 18a to Intel 14a than to move from TSMC (N2) to Intel 14a for designers?

I'd guess it would be fairly similar, since Intel now uses industry standard toolchains. There's obviously going to be overhead, when you switch from one foundry to another, as each company will do various things differently and whole new sets of relationships need to be built.

Intel also has a huge trust deficit, which their financial problems and habit of canceling nodes & products doesn't help. What tech company wants to risk an entire generation of products on Intel following through, at this point? In order to make attractive offerings, Intel might also have to do some significant discounting, which is exactly the opposite of what they want to be doing, right now.

 

[DS426]

...

Intel also has a huge trust deficit, which their financial problems and habit of canceling nodes & products doesn't help. What tech company wants to risk an entire generation of products on Intel following through, at this point? ...In order to make attractive offerings, Intel might also have to do some significant discounting, which is exactly the opposite of what they want to be doing, right now.

That is kind of a good summarization of what I'm trying to say in this thread. Intel setting the goal post of 18A being sold out isn't the same as the rest of the world's goal post to prove (not claim) that 18A is successful -- final end-products still have to reach the market in volumes and have competitive cost, performance, and efficiency characteristics that we won't see until later this year, both Intel's and their clients'.

That said, I'm sure many clients are intentionally sitting this round out to let others be the guinea pigs. If 18A succeeds, it's a much easier road to cross on assuming that 14A will be competitive as well. Industry-standard tooling is definitely a big plus as well, yes.