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

[JRStern]

Cutting edge nodes are all well and good, but in the Angstrom Era the benefits aren't as drastic, and a mature older process is worth more than a cutting edge less refined one.

That's what I say.
I want to go back to single-exposure DUV and see if there are new ways to optimize it!

 

[JRStern]

So 18A isn't ready after all? Thank you Pat. Thank you affirmative action kid. Thank you the board for choosing some cr*ppy religious fanatic as CEO. Thank you for choosing somebody without even a PHD as CEO. At least Pat had style - he'd rather lose $100B (of somebody else's money) than a mere $1B.

I don't really want to agree with that, but ...

 

[bit_user]

IP are people.

At this level, it goes well beyond people. SMIC has lured away significant talent from TSMC and they're still way behind. The IP is a lot of infrastructure, processes, design work, formulas, methodologies, etc. There are probably tens of thousands of details that have to be right for successful wafer production at these scales. What we're talking about transcends people and rises to the level of an institution.

A system like TSMC could take 20 years to recreate.

Yes, quite possibly.

 

[bit_user]

That's what I say.
I want to go back to single-exposure DUV and see if there are new ways to optimize it!

Isn't that exactly what Intel spent more than half a decade doing, with their nodes from 10 nm to "Intel 7"?

 

[DS426]

They didn't get any loan to build the FABs. They used the profits they made during human malware plus the profits they make every year, plus some of the FABs are co financed by a company.
Also this:

I have to doubt that. Most businesses take out seasonal loans, term revolvers, and other types of loans and financing options to make sure they have sufficient working capital (free cashflow) to help make the actual payment on huge CapEx items and even OpEx. Then there's the Line-of-Credit to also help with floating cash. Businesses tend to have very close relationships with banks and financial institutions in general.

You really mean revenues, not profits as they've had operating losses ever since IFS was established in 2021.

I think the problem here is that Intel's fabs essentially always lost money in themselves but yielded very very profitable client and server chips for ages, so much so that there wasn't any concern or even relevance of their "profitability." Spinning out IFS from the rest of Intel has exposed this issue -- one that has proven to be more stubborn than Pat, Tan, and probably anyone at all could have imagined.

Intel spent all that R&D to nix 20A and only have three fabs (really two as Oregon isn't high-volume like Arizona) for 18A? Even if 18A is fully booked, and I actually don't doubt that this is true as Intel themselves said they are fully booked on 18A and that's one of their big reasons to focus on 14A now, it's their misjudgment on using economies-of-scale to swing all those huge fab costs into having a positive return on investment. 14A is only going to be even more expensive, and with the Ohio megafab project pushed way back, it's obvious that they aren't going to scale 14A massively either... maybe not even "aren't" but "can't." I suppose they thought that they would have been able to secure more of the CHIPS Act funding.

I remember vividly when Pat said himself that he bet the entire company on 18A:
https://www.tomshardware.com/pc-com...18a-chip-slated-for-high-volume-manufacturing

Fool me once, shame on you. Fool me twice, shame on me...

 

[phead128]

It makes sense to cancel 18A if 14A nodes are promising and designed for external customers. Why continue investing resources in 18A when 14A nodes are the main nodes to compete against TSMC and designed for external customers.

 

[DHB]

What do you mean by "writing off" ? They are going to use 18A the same way they used all of their nodes from since they became intel, they will make a full line up and sell them.
With only two foundries that make 18A there is probably no space for anybody else's products anyway.
Intel will have to get a few more fabs running for them to have enough fabs to produce for others as well.

Intel is considering removing 18a from external foundry market and focus on marketing 14a to foundry customers resulting in probable write offs for the investment in 18a (not getting hoped for foundry customers). So, maybe making the node pricing more attractive instead would increase interest in the node particularly if the 18a can be an introduction and becomes a jumpstart into using Intel’s 14a nodes. So it is actually a move forward for customers when they design for 18a?

 

[DHB]

They didn't get any loan to build the FABs. They used the profits they made during human malware plus the profits they make every year, plus some of the FABs are co financed by a company.
Also this:

They got funding from the Chips and Science act.

 

[DHB]

It makes sense to cancel 18A if 14A nodes are promising and designed for external customers. Why continue investing resources in 18A when 14A nodes are the main nodes to compete against TSMC and designed for external customers.

It has to do with capacity and it’s being used for internal (company /non-foundry customer) products.

 

[phead128]

It has to do with capacity and it’s being used for internal (company /non-foundry customer) products.

Outsource 18A to TSMC for internal use, then focus all engineering resources on 14A to leapfrog TSMC.

 

[TerryLaze]

I have to doubt that. Most businesses take out seasonal loans, term revolvers, and other types of loans and financing options to make sure they have sufficient working capital (free cashflow) to help make the actual payment on huge CapEx items and even OpEx. Then there's the Line-of-Credit to also help with floating cash. Businesses tend to have very close relationships with banks and financial institutions in general.

I'm not saying that they don't have any loans, but they didn't take any to build the FABs.

You really mean revenues, not profits as they've had operating losses ever since IFS was established in 2021.

No I really mean profits, from 2018 to 2021 they made double the profit per year than usual.
The operating losses they have since 2021 are because the money they would normally make goes to the FABs as well.
https://www.macrotrends.net/stocks/charts/INTC/intel/net-income

I think the problem here is that Intel's fabs essentially always lost money in themselves but yielded very very profitable client and server chips for ages, so much so that there wasn't any concern or even relevance of their "profitability." Spinning out IFS from the rest of Intel has exposed this issue -- one that has proven to be more stubborn than Pat, Tan, and probably anyone at all could have imagined.

They used to be profitable because they used to make products, ever since 2021 they are only building stuff without being able to produce stuff, it's like being a baker and having your kitchen remodeled...there is no way to bake in that kitchen until it's done.

Intel spent all that R&D to nix 20A and only have three fabs (really two as Oregon isn't high-volume like Arizona) for 18A? Even if 18A is fully booked, and I actually don't doubt that this is true as Intel themselves said they are fully booked on 18A and that's one of their big reasons to focus on 14A now, it's their misjudgment on using economies-of-scale to swing all those huge fab costs into having a positive return on investment. 14A is only going to be even more expensive, and with the Ohio megafab project pushed way back, it's obvious that they aren't going to scale 14A massively either... maybe not even "aren't" but "can't." I suppose they thought that they would have been able to secure more of the CHIPS Act funding.

I remember vividly when Pat said himself that he bet the entire company on 18A:
https://www.tomshardware.com/pc-com...18a-chip-slated-for-high-volume-manufacturing

Fool me once, shame on you. Fool me twice, shame on me...

They stopped promoting it, at least that's the only thing the article says....they did not scrap it, that's your phantasy making that stuff up.
And not even stopped, only considering it.

Lip-Bu Tan, the chief executive of Intel, is considering stopping the promotion of the company's 18A fabrication technology (1.8nm-class) to foundry customers,

 

[TerryLaze]

They got funding from the Chips and Science act.

Which is not a loan, and also laughably little in contrast to what they spend.

 

[PCWarrior]

This doesn't imply that Intel's 18A node is not ready or has issues. Neither does it imply it is inferior to TSMC’s N2. Nor does it mean that the node won’t be available to others to use. What it does mean is that Intel won’t actively pursue an advertising campaign to persuade major companies like Nvidia and Apple to use it. This could be because of two reasons:

1. Intel may not have enough production capacity to meet both its internal demands and those of major external clients.

2. Intel has determined that big potential external customers (like Nvidia and Apple) would be difficult to be persuaded to sway away from TSMC (their long-standing manufacturing partner), in favour of Intel when all Intel can offer is mere equivalency. Besides, by this point in time, these companies have likely already committed to TSMC’s N2 node for most of their upcoming products anyway.

To win over external customers, Intel really needs to deliver a next-generation process node that launches a full year ahead of TSMC’s equivalent AND also maintains some technical superiority even after TSMC catches up. Otherwise, they would only be fighting for scraps, that is produce chips when TSMC is at full capacity and unable to take on new orders. The alternative would be to offer their nodes at a lower price compared to TSMC or even operate at a loss. But most likely Intel figured that it would probably be more profitable to them to just use the whatever excess capacity they have for their own products. After all they would earn more by selling heavily discounted cpus or gpus to consumers and data centres than selling chips to Nvidia or Apple for nothing.

 

[bit_user]

Outsource 18A to TSMC for internal use, then focus all engineering resources on 14A to leapfrog TSMC.

That's like asking 9 women to collaborate on making a baby in 1 month. There's only so much you can do to speed up projects by throwing resources at them, especially late in the cycle.

This is a mistake often made by management - they think they can make up for poor planning or underinvestment by throwing resources at a project when it becomes clear that it will be late. But, by that point, there's not much you can do except mess it up even worse.

New process nodes take a long time to develop, deploy, and ramp. That's why they're pipelined, with usually at least the next two being developed concurrently.

 

[bit_user]

They used to be profitable because they used to make products, ever since 2021 they are only building stuff without being able to produce stuff, it's like being a baker and having your kitchen remodeled...there is no way to bake in that kitchen until it's done.

That's not even true. Their server CPUs all still use in-house nodes and Meteor Lake (now rebranded as the lower-end Arrow Lake laptop CPUs) use Intel nodes for their compute tiles.

 

[phead128]

That's like asking 9 women to collaborate on making a baby in 1 month.

I mean, let's say there is 9 females in the kitchen. Right now, you are dedicating 7 females to cook a dish (18A) using a older utensil (low-NA EUV), and 2 females to cook another dish (14A) using another new utensil (High-NA EUV). The new utensil (High-NA EUV) is another beast in it's own that has it's own learning curve, idiosyncrasies, but ASML claims it's more cost-efficient per wafer throughput, which could lead 14A to have a price-per-unit advantage over TSMC. Why not abandon Low-NA EUV for 18A and go 'all-in' on the High-NA EUV for 14A to secure the next process node leadership for Intel? TSMC only has 1 High-NA EUV as a reference. 14A will restore Intel to process leadership AND have a price advantage.

 

[magbarn]

You're missing the point, which is that process R&D costs have ballooned to the point where they were becoming unsustainable on the old Intel's production volume. The new Intel's production volume is even less. So, unless the business somehow magically starts selling an order of magnitude new chips, they can't continue to fund development of new nodes. Becoming a foundry is the most obvious way of increasing production volume, at this point.


First, TSMC is the goose that lays golden eggs. It's the biggest prize of capturing Taiwan. There's no way China would damage it. Second, they have fabs outside Taiwan, so even if an earthquake causes an extended outage to their Taiwanese fabs, it wouldn't completely shut them down, nor erase their IP advantage.


I agree. Personally, I don't understand why Intel's fabs aren't seen as being more strategically important.

You betcha if/when China takes over Taiwan, it'll be Tomahawks hitting the foundries and not Chinese cruise missiles.

 

[thestryker]

I think the problem here is that Intel's fabs essentially always lost money in themselves but yielded very very profitable client and server chips for ages, so much so that there wasn't any concern or even relevance of their "profitability."

This wasn't accurate until 14nm because of how they reused equipment, facilities and nodes rather efficiently. 14nm had some pretty nasty delays and they like to pretend the first iteration (BDW) wasn't real. Eventually over the lifetime it did pay for itself but that's because it was used in 6 generations worth of products unlike the usual 2-3.

Spinning out IFS from the rest of Intel has exposed this issue -- one that has proven to be more stubborn than Pat, Tan, and probably anyone at all could have imagined.

The issue was never really hidden assuming someone actually paid attention to the company. I don't think any of them were surprised at anything except for the wallstreet whiplash when the numbers were made public.

Intel spent all that R&D to nix 20A and only have three fabs (really two as Oregon isn't high-volume like Arizona) for 18A?

Actually it's only one because AZ Fab 62 hasn't opened yet. Killing 20A for the reason they claimed they did would make all the sense in the world. Using Intel 4 for MTL turned out to be a mistake due to EUV capacity issues . If they truly are messing around with 18A and are de-emphasizing it then I think it's fair to say that reasoning was likely false.

 

[thestryker]

Outsource 18A to TSMC for internal use, then focus all engineering resources on 14A to leapfrog TSMC.

I'm assuming you mean outsource products which would be made on 18A to TSMC. If so then they already did that with LNL and ARL which has hurt the margins on both of those parts. Doing that again with a full lineup would be very bad.

I mean, let's say there is 9 females in the kitchen. Right now, you are dedicating 7 females to cook a dish (18A) using a older utensil (low-NA EUV), and 2 females to cook another dish (14A) using another new utensil (High-NA EUV). The new utensil (High-NA EUV) is another beast in it's own that has it's own learning curve, idiosyncrasies, but ASML claims it's more cost-efficient per wafer throughput, which could lead 14A to have a price-per-unit advantage over TSMC. Why not abandon Low-NA EUV for 18A and go 'all-in' on the High-NA EUV for 14A to secure the next process node leadership for Intel? TSMC only has 1 High-NA EUV as a reference. 14A will restore Intel to process leadership AND have a price advantage.

There are zero High-NA machines installed outside of OR right now. ASML hasn't shipped any of the production model at all yet either (I don't know how big of a problem this is in a practical sense, but it's not good). This is likely why Intel was looking at potentially not using it at all for 14A. Even if they do use it for 14A it will be targeted usage which means the vast majority of the equipment is going to be EUV. Having another one and done situation with 18A would likely be a financial disaster.

 

[DS426]

...

They stopped promoting it, at least that's the only thing the article says....they did not scrap it, that's your phantasy making that stuff up.
And not even stopped, only considering it.

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.

 

[acadia11]

I've seen no evidence they care about that. The main issue of concern seems to be where the factories are geographically located. They don't seem to show a lot of interest in IP, in general.

I’d be hard pressed to believe any of the players in this administration responsible for this domain could be that out of touch. As much railing and complaining about China they can’t just boil it down to factory jobs for a labor force that the Us doesn’t have the skills to supply in the first place? Then again I just read the announcement that they will ease the export of of technology on chips in exchange for rare earth metal deal. The art of the deal I guess … so Joss whedon had it right we’ll all be fluent in mandarin by 2050.

 

[bit_user]

I mean, let's say there is 9 females in the kitchen. Right now, you are dedicating 7 females to cook a dish (18A) using a older utensil (low-NA EUV), and 2 females to cook another dish (14A) using another new utensil (High-NA EUV). The new utensil (High-NA EUV) is another beast in it's own that has it's own learning curve, idiosyncrasies,

Haven't you every heard the phrase "too many cooks in the kitchen"? When you throw people at a project, what happens is that the experienced and most productive contributors get bogged down having to bring the new people up to speed. This slows down the project and not everything can be parallelized, so the opportunities for the new people to help out are limited. Also, being new, they're way down the learning curve, so even when they can find a place to contribute, they're slow and clumsy, which takes additional time from the existing contributors.

This is a well-studied phenomenon, famously described in The Mythical Man Month. And that's just talking about software projects. What you're talking about is lithography, where there are a limited number of high-NA machines and it sounds like they're already fully-subscribed.

https://www.tomshardware.com/tech-i...0-000-wafers-with-high-na-euv-chipmaking-tool
​

Once the project is fully-resourced, throwing anything more at it would be disruptive. I'm not saying that 14A is wanting for nothing, but the benefit to 14A from cancelling 18A is surely quite limited, at this stage.

You also seem to have a very short memory. It was just under 1 year ago that Intel cancelled 20A in order to shift resources onto 18A. So, now you're saying cancel 18A to focus on 14A, then you'll be saying they should cancel 14A to focus on 10A, and so on.

but ASML claims it's more cost-efficient per wafer throughput,

TSMC directly contradicted this assertion, saying they don't expect high-NA to be more cost effective for a couple more generations, at least.

 

[bit_user]

You betcha if/when China takes over Taiwan, it'll be Tomahawks hitting the foundries and not Chinese cruise missiles.

If the US attacks anything in Taiwan, it will find ballistic missiles raining down on all its naval vessels in the eastern Pacific. Starting a hot war with China would be the height of stupidity. It's also a war crime to attack civilian targets, like fabs.

 

[acadia11]

If the US attacks anything in Taiwan, it will find ballistic missiles raining down on all its naval vessels in the eastern Pacific. Starting a hot war with China would be the height of stupidity. It's also a war crime to attack civilian targets, like fabs.

lol a war crime… I’ve always found this to be hilarious war crime, war is the crime anf the whole point is to use violence and impose one’s will. You’ve already crossed the rubicon and unless someone can actually stop your violence what so “war crimes” matter?

 

[PCWarrior]

>This doesn't imply that Intel's 18A node is not ready or has issues. Neither does it imply it is inferior to TSMC’s N2...This could be because of two reasons:

People should just read the article. Here's the source article that the THW piece is derived from:

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

The reason is stated plainly enough, "By June, [Lip-Bu Tan] started voicing that a manufacturing process known as 18A, in which prior CEO Pat Gelsinger had invested heavily, was losing its appeal to new customers.
...
"Persuading outside clients to use Intel's factories remains key to its future. As its 18A fabrication process faced delays, rival TSMC's N2 technology has been on track for production.
...
"Tan's preliminary answer to this challenge: focus more resources on 14A, a next-generation chipmaking process where Intel expects to have advantages over Taiwan's TSMC.
...
"...according to some industry analysts, the 18A process is roughly equivalent to TSMC's so-called N3 manufacturing technology, which went into high-volume production in late 2022."

‘Losing its appeal’ and 'the 14A node will have advantages over TSMC' (implying the 18A doesn't) are the only things that were allegedly said by Intel’s CEO. Where exactly in my response did I contradict these? To me the phrase ‘losing its appeal’ is vague and implies none of the things about the supposed bad state of 18A that many people here and elsewhere speculate. For all we know one major appeal Intel’s 18A had over competing TSMC’s N2 was coming to market earlier. Intel’s 18A was supposed to launch in late 2024 with high-volume manufacturing in 2025 putting it almost a year ahead of TSMC’s N2. This hasn’t happened, so from that perspective it has lost its appeal. As I said and earlier equivalency is not enough to persuade big customers to leave their trusty and neutral manufacturer in favour of a direct market rival. That being said despite the delay and lost appeal to external customers, Intel is not late in launching 18A compared to TSMC N2. By the end of this year we will have the first real products in market that actually use these nodes (18A for Intel Panther lake and N2 for iPhone 17Pro/Pro Max).

I also see you citing the Reuters report. These so-called industry analysts they are citing are not people from the tech/semiconductor industry but people in the finance sector that read the news/articles and listen to rumours and speculations to ‘determine’ the newest developments and trends. The fact of the matter is that 18A is not comparable to TSMC’s N3 as per the some analysts’ suggestion but to N2. It is even written in their very article (that 18A compares to N2) yet at the end the Reuters article proceeds to contradict it with the analysts' misinformation. In contrast the Tom’s hardware article did not include any of this and just reaffirms the equivalency of TSMC’s N2 and 18A. It’s an irony that you accuse others of not reading the article. By the way I fully expect someone to come here and give me density numbers taken out of context. Below is my pre-emptive response.

View: https://www.youtube.com/watch?v=KpZAhQvQMGk