Monday, September 8th 2025

Intel 14A Node Expected to Cost More Than 18A, Driven by High-NA EUV

At Citibank's Global 2025 TMT conference, Intel CFO David Zinsner made an appearance and talked about the economics of Intel's next-generation 14A node. According to Zinsner, the "14A [node] is more expensive than 18A. It is not significantly [more expensive] in terms of investment. [But] it is a higher wafer cost, for sure and partly that is because we, we are expecting to use High-NA EUV tools in 14A, which was not the case in 18A." The use of High-NA EUV tools suggests that Intel will install the Twinscan EXE:5200B machine, which costs $380 million. Given the high cost of these tools, Intel will need to raise its wafer prices to break even, assuming the 14A node attracts external customers.

Intel claims its upcoming 14A node will deliver significant efficiency gains, achieving roughly 15-20% better performance per watt, or a 25-35% reduction in power consumption compared to the 18A node. Under the hood, 14A combines several process advances. RibbonFET 2 updates Intel's gate-all-around transistor architecture, while PowerDirect moves the power delivery network to the chip's backside so supply rails can feed transistor sources and drains more directly. The node also introduces Turbo Cells, taller high-drive cells placed within compact standard-cell libraries to shave critical timing paths and lift CPU and GPU frequencies without large area or power penalties. By contrast, scaling 18A will depend on more capable lithography tools with higher resolution, so the node can avoid relying on multi-patterning.
Source: via Tom's Hardware
Add your own comment

13 Comments on Intel 14A Node Expected to Cost More Than 18A, Driven by High-NA EUV

#1
Hakker
and water is wet. Offcorse it's more expensive than the previous node because it's more difficult to do. It always is.
Posted on Reply
#2
kondamin
would be happy with an intel 3 n cpu
Posted on Reply
#3
Squared
while PowerDirect moves the power delivery network to the chip's backside
Wasn't that supposed to be a 20A feature which was definitely ready because they tested it on Intel 4 and said it was working great? How is it new on 14A?
Posted on Reply
#4
overclockedamd
Guess we’ll be seeing these closer to 2029–2030 in real volume, but by then TSMC’s roadmap will be way ahead. N2 is set for 2025, N2P in 2026, and N2X already being talked up for 2027 mass production. If those ramps hold, Intel’s 14A will be landing right as TSMC’s high-performance N2X is mature. Hope Intel delivers, because that’s a brutal race to walk into late.
Posted on Reply
#5
john_
They better forget the "brake even" and just offer prices that others will think twice to ignore. And I believe behind close doors, that's what they will do. The "higher prices to brake even" are probably something intended for the smaller companies that will be interested using Intel nodes, in case those nodes attract big companies. Because right now Intel nodes are a huge bet for anyone thinking building new chips there. Intel is a recipe for disaster the last 10 years in manufacturing and they will need to offer huge discounts, even if this means loses to attract big players. And they need to create a customer base first. They can think profits latter.

Apple: What is the cost per wafer?
Intel: $10
Apple: TSMC's price is $12 but I know that my chips will be good.
Intel: $8
Apple: I know that they will be performing as expected.
Intel: $6
Apple: With no degradation effects.
Intel: ..............$5?
Apple: Deal.
Posted on Reply
#6
Wirko
Hakkerand water is wet. Offcorse it's more expensive than the previous node because it's more difficult to do. It always is.
What else is a CFO supposed to say?
Posted on Reply
#7
Deorc Mearh
I can't help thinking those at Intel that make these claims, first of all the new CEO, are totaly incapable or plan to make Intel fail. They are producing the best cpus for production softwares based on floating point calculus with a 265k costing a really fair price thou they are not as good in integer based application where amd works better: why not making a commercialncampaign to point that feature?? They keep on communicatimg they have failed and even 18A is a failure so much that they have to emphasize14A process. No one 18A Intel cpu has still been sold and they are letting intended they don't have any hoped improvement over ArrowLake. Even if it were true you must never tell it that way. When the government intervened I was indded astonished not because it did it but because it did it so late thou Intel ceo had released a ton of idiotic suicidal statements and was to be silenced as quick as possible.I think government must get rid of part of the board. Deleting globalists' inclusive marxist policies imposed for putting some useful stupid puppets in roles where they could harm the company but spread propaganda for globalists' agenda is not enough. Part of the board either works for selling off intel's foundries or it 's completely demented
Posted on Reply
#8
RandallFlagg
This is the problem with getting new management, in any company not just Intel.

The new management discards the accomplishments of the old management, because it wasn't theirs.

18A was supposed to be Intel's crown jewel and frankly I see no reason why it still can't be. Who here is running a 2nm CPU? 2nm GPU?

Nope, they don't exist yet. Intel will likely be the first to market with that by 6-9 months, and yet it is already a 'failure'? Why not just focus in on it , get the volume up, and make it better? Maybe if they did that then by this time in 2026 we could have laptop, desktop, and GPUs on Intel 18A while the competition just starts to release some token volume N2 parts for stupid high prices?

But no, that was Geisinger's idea. Can't have that.
Posted on Reply
#9
kondamin
Deorc MearhI can't help thinking those at Intel that make these claims, first of all the new CEO, are totaly incapable or plan to make Intel fail. They are producing the best cpus for production softwares based on floating point calculus with a 265k costing a really fair price thou they are not as good in integer based application where amd works better: why not making a commercialncampaign to point that feature?? They keep on communicatimg they have failed and even 18A is a failure so much that they have to emphasize14A process. No one 18A Intel cpu has still been sold and they are letting intended they don't have any hoped improvement over ArrowLake. Even if it were true you must never tell it that way. When the government intervened I was indded astonished not because it did it but because it did it so late thou Intel ceo had released a ton of idiotic suicidal statements and was to be silenced as quick as possible.I think government must get rid of part of the board. Deleting globalists' inclusive marxist policies imposed for putting some useful stupid puppets in roles where they could harm the company but spread propaganda for globalists' agenda is not enough. Part of the board either works for selling off intel's foundries or it 's completely demented
Can you make sense of this wall of letters your self?

from what I can make out of it, you think intel is saying their latest node is bad.
They are not, they are saying the next one will be beter and more expensive.

Anonymous sources should be treated as such, keep it in the back of your head if you need to make a descission about an investment or something. But for all other things treat them like fan fiction you enjoy but don’t bother others with.
Posted on Reply
#10
Deorc Mearh
kondaminCan you make sense of this wall of letters your self?

from what I can make out of it, you think intel is saying their latest node is bad.
They are not, they are saying the next one will be beter and more expensive.

Anonymous sources should be treated as such, keep it in the back of your head if you need to make a descission about an investment or something. But for all other things treat them like fan fiction you enjoy but don’t bother others with.
I base what I said on all the public claims Intel 's new ceo did. When you are yet developing 18A process you don say publicly " oh well we spent a ton of money for it but it's not good enough to be sold to the other chip producers so we bet on 14A node". It's a clear ammission of having a lot of troubles and no significant improvement. Even if it were so you don't tell things like that if you are the CEO of a company with tens of billions dollar revenue. You can say " the 18A node is almost ready but at the moment we decided we won't immediately propose it to our possible clients keeping it for our own products for a while". This might imply it's good and you don't want to give it to competitors at the beginning for keeping an advantage: this way you don't make the stock fall. Everytime he opened his damn mouth he said something loser, something to depress the idea of Intel and its products the investor might have even when some products are good. He's clearly stupid or he works for some other companies. You don't save intel cancelling projects and foundries in construction
Posted on Reply
#11
RandallFlagg
kondaminCan you make sense of this wall of letters your self?

from what I can make out of it, you think intel is saying their latest node is bad.
They are not, they are saying the next one will be beter and more expensive.

Anonymous sources should be treated as such, keep it in the back of your head if you need to make a descission about an investment or something. But for all other things treat them like fan fiction you enjoy but don’t bother others with.
As someone said, this is like saying water is wet.

We don't see TSMC hyping up their 14A node. They are still working diligently on N2, improving yields, talking up its strengths. Why? Because the next few years is the era of N2, this is their flagship node and for now they'll be investing a lot in it.

What Lip-Bu is doing, is like a car company, say Toyota, coming out with a new Avalon for 2026. Then they say 'Yeah, this is our new 2026 one. It's so-so. But just wait until 2030 when the next gen comes out, that's when the real change happens and it'll be revolutionary!'

There's a reason no manufacturer ever does that.

Lip-Bu is either an idiot or something worse.
Posted on Reply
#12
TechBuyingHavoc
RandallFlaggThis is the problem with getting new management, in any company not just Intel.

The new management discards the accomplishments of the old management, because it wasn't theirs.

18A was supposed to be Intel's crown jewel and frankly I see no reason why it still can't be. Who here is running a 2nm CPU? 2nm GPU?

Nope, they don't exist yet. Intel will likely be the first to market with that by 6-9 months, and yet it is already a 'failure'? Why not just focus in on it , get the volume up, and make it better? Maybe if they did that then by this time in 2026 we could have laptop, desktop, and GPUs on Intel 18A while the competition just starts to release some token volume N2 parts for stupid high prices?

But no, that was Geisinger's idea. Can't have that.
And I don't know if cramming so many process advances into a single node is a great idea. RibbonFET 2 and PowerDirect technologies might be a step too far, especially because Intel still hasn't proved that it can reliably transition to new nodes without massive delays and cost overruns. This sounds like the 14nm node all over again.
RandallFlaggThere's a reason no manufacturer ever does that.

Lip-Bu is either an idiot or something worse.
I would say that he is not a idiot, he is trying to hide the fact that Intel fabs won't be competitive anytime soon. Even their new nodes coming out in 2027/2028 won't be competitive with TSMC's planned nodes for those years.

Of course, the Osborne effect still applies here and will only suppress current demand for Intel fabs.
Posted on Reply
#13
RandallFlagg
TechBuyingHavocI would say that he is not a idiot, he is trying to hide the fact that Intel fabs won't be competitive anytime soon. Even their new nodes coming out in 2027/2028 won't be competitive with TSMC's planned nodes for those years.

Of course, the Osborne effect still applies here and will only suppress current demand for Intel fabs.
I think they can focus on and iteratively improve their 18A node. i.e. by that I mean, improve yield, refine its efficiency and improve the libraries for better layout.

I also suspect the new nodes are going to turbocharge the price bellyaching that already goes on. Just wait and see how expensive N2 is. Intel may well have a huge advantage in that respect. I have no interest in seeing a return of the $600 midrange CPU a la the 1990s, but that is what N2 is likely to bring.
Posted on Reply
Add your own comment
Nov 20th, 2026 21:20 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts