I don't like how some people simplify things, it makes it seem like simply buying EUV machines first did all the magic, and Samsung and Intel could do the same.
no they couldn't because it takes forever to build these machines, so TSMC was able to control the largest amount of EUV machines by far. samsung and intel attempted to rush to catch up in orders but ASML can only make so many machines per year and its first come first serve.
intel is actually attempting to do the same thing with highNA by being the first to buy them in bulk.
Which is all the more reason I think something unspoken has already transpired. My speculation is the US government has required ASML to ship 100% of high-NA machines to USA for a limited time period.
Which I guess is another reason to question why Intel would be buying every commercial unit shipped for the whole of 2024. They haven't even got the first one going yet.
Not that much money, just on expectations! And definitely not while you still don't have a working line. Rather than expectations, I think the money has been allocated despite Intel's position.
It will be interesting to see how this plays out. As TSMC’s largest customer, I’m sure Apple has a lot of sway over TSMC’s choices. I can’t imagine that TSMC would refrain from investing in high NA tools if Apple requested that they make that investment. So I think we can infer that Apple has made no such request. The question then becomes: why has Apple not made that request?
One possible answer is that Apple does not believe it is necessary to move to the new tools yet. But I am skeptical that is the case. Apple has shown that they see great value in being on the most advanced available node. And what we know of apples future product plans suggest they will continue to want the very best (for example, the Vision product line needs the best performance/watt available if it is to go mainstream).
So, I wonder if the answer is that Apple wants a competitor for TSMC. Intel is making these investments using taxpayer money, not Apple’s money. That has to be appealing to Apple. And yet Apple could, by virtue of size, still secure most of Intel’s foundry output and be first in line.
Perhaps we will see that intel’s secret “whale” customer with the big prepayment is Apple, and that apple will then turn around and squeeze TSMC, forcing TSMC to scramble to buy the high NA tools on their own dime, without prepayment support. Or in some other way secure more favorable terms than currently.
It really doesn't matter what Apple wants, there is only 1 High NA machine in the world, and it is just being set up and tested right now- Intel owns it. It is going to be a while-2025? before anything hits the market, and I bet if Apple sees an advantage- they will hire intel to make their chips. Intel wants to be a foundry for hire again, that would be a giant leap...just a thought.
Not in 2025, earliest 2026, and TSMC doesn't plan to use HIgh-NA until 2027 or 2028, I'm sure they have the tools when they want them. That everyone is making waves about who takes first delivery of a new machine being put in a lab to develop future tools is pointless, it's the production processes and their characteristics that matter to us consumers in the end. What's interesting is if one of them are unable to follow their plans. Btw TSMCs best low-NA scanner allows for better characteristics than the first High-NA machine.
The whole point is to reduce costs of producing wafers, not to have the biggest EUV machine from ASML. So, if TSMC can produce wafers overall cheaper with Low NA EUV machines using customised pelliceles and resists, then they will beat Intel.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
20 Comments
Back to Article
Terry_Craig - Friday, May 17, 2024 - link
I don't like how some people simplify things, it makes it seem like simply buying EUV machines first did all the magic, and Samsung and Intel could do the same.whatthe123 - Friday, May 17, 2024 - link
no they couldn't because it takes forever to build these machines, so TSMC was able to control the largest amount of EUV machines by far. samsung and intel attempted to rush to catch up in orders but ASML can only make so many machines per year and its first come first serve.intel is actually attempting to do the same thing with highNA by being the first to buy them in bulk.
evanh - Friday, May 17, 2024 - link
Which is all the more reason I think something unspoken has already transpired. My speculation is the US government has required ASML to ship 100% of high-NA machines to USA for a limited time period.Dante Verizon - Saturday, May 18, 2024 - link
Both have had EUV machines for years and are still embarrassed by TSMC's superior processes.evanh - Saturday, May 18, 2024 - link
Which I guess is another reason to question why Intel would be buying every commercial unit shipped for the whole of 2024. They haven't even got the first one going yet.Dante Verizon - Saturday, May 18, 2024 - link
Desperation heightened by the expectations of the government and investors. :)evanh - Saturday, May 18, 2024 - link
Not that much money, just on expectations! And definitely not while you still don't have a working line. Rather than expectations, I think the money has been allocated despite Intel's position.Exotica - Saturday, May 18, 2024 - link
Now that Intel is buying up the high na capacity you do have to wonder if within 2-3 years if intel will have a lead over TSMC…. At sub 2nm.Blastdoor - Saturday, May 18, 2024 - link
I think that’s the hope.It will be interesting to see how this plays out. As TSMC’s largest customer, I’m sure Apple has a lot of sway over TSMC’s choices. I can’t imagine that TSMC would refrain from investing in high NA tools if Apple requested that they make that investment. So I think we can infer that Apple has made no such request. The question then becomes: why has Apple not made that request?
One possible answer is that Apple does not believe it is necessary to move to the new tools yet. But I am skeptical that is the case. Apple has shown that they see great value in being on the most advanced available node. And what we know of apples future product plans suggest they will continue to want the very best (for example, the Vision product line needs the best performance/watt available if it is to go mainstream).
So, I wonder if the answer is that Apple wants a competitor for TSMC. Intel is making these investments using taxpayer money, not Apple’s money. That has to be appealing to Apple. And yet Apple could, by virtue of size, still secure most of Intel’s foundry output and be first in line.
Perhaps we will see that intel’s secret “whale” customer with the big prepayment is Apple, and that apple will then turn around and squeeze TSMC, forcing TSMC to scramble to buy the high NA tools on their own dime, without prepayment support. Or in some other way secure more favorable terms than currently.
Threska - Saturday, May 18, 2024 - link
Maybe, but I'm sure Apple remembers why they left Intel in the first place.Papaspud - Sunday, May 19, 2024 - link
It really doesn't matter what Apple wants, there is only 1 High NA machine in the world, and it is just being set up and tested right now- Intel owns it. It is going to be a while-2025? before anything hits the market, and I bet if Apple sees an advantage- they will hire intel to make their chips. Intel wants to be a foundry for hire again, that would be a giant leap...just a thought.Zoolook13 - Sunday, May 19, 2024 - link
Not in 2025, earliest 2026, and TSMC doesn't plan to use HIgh-NA until 2027 or 2028, I'm sure they have the tools when they want them.That everyone is making waves about who takes first delivery of a new machine being put in a lab to develop future tools is pointless, it's the production processes and their characteristics that matter to us consumers in the end.
What's interesting is if one of them are unable to follow their plans.
Btw TSMCs best low-NA scanner allows for better characteristics than the first High-NA machine.
evanh - Monday, May 20, 2024 - link
Intel's outlay in such a small space of time is extraordinary. Some of the machines will end up collecting dust for a while me thinks.Dante Verizon - Sunday, May 19, 2024 - link
Apple is not a foundry, guy. It's up to TSMC, which dominates the sector, to see whether it has an advantage or not.Blastdoor - Monday, May 20, 2024 - link
https://www.extremetech.com/computing/report-intel...Five or six machines to be delivered in 2024, all to Intel.
evanh - Monday, May 20, 2024 - link
Machine #2 shipped a month back. It went to an unnamed customer in the US. Speculation has it at a research facility that IBM is partnered with. https://www.anandtech.com/show/21356/asml-patterns...lefty2 - Sunday, May 19, 2024 - link
The whole point is to reduce costs of producing wafers, not to have the biggest EUV machine from ASML. So, if TSMC can produce wafers overall cheaper with Low NA EUV machines using customised pelliceles and resists, then they will beat Intel.Blastdoor - Monday, May 20, 2024 - link
I think the whole point for leading edge customers is to increase performance/watt.Anymoore - Monday, May 20, 2024 - link
The wafers per day was increased by getting higher EUV source power (400W vs. 250W). That's the key advantage over competitors.Anymoore - Monday, May 20, 2024 - link
This article doesn't tell the whole story.From 2022 to 2023 the wafer output was flat even though EUV tools increased ~30%. N3 appears to have reduced the productivity.
https://semiwiki.com/forum/index.php?threads/intel...