News Intel principal engineer bemoans potential TSMC takeover, touts company's 18A tech advantage

[Pierce2623]

He doesn't know what he is talking about. He only has a Ph.D. in Physics...

Which means very little vs a bachelors in computer engineering when we’re talking about process node development, especially considering we’re talking much more about logic chips at a high level view rather than things like the various chemical treatments involved in making chips.

 

[smartcom5]

I'm suspecting that this engineer's opinion is a lie he created to make Intel look desirable from outside, to make Intel worth more to any external "buyers". He just paints too rosy of a picture. Cloud 9 stuff.

With all due respect, I surely not know anyone at Intel, yet when he writes bits like this;

Handing over control of Intel's foundries solves this problem for TSMC, but hurts Intel, hurts US leadership, and gives credibility to the false idea that Intel Foundry is not as advanced as TSMC.

You clearly know, he's either unknowingly talking nonsense for whatever reason, intentionally spreads untruths or is straight-up in Lala-land, and people like him are the reason forwhy Intel is in sharp decline!

I mean, everyone knows for a fact, that we're well behind any schedule here, when Intel has to outsource to TSMC instead of manufacturing it ourselves. Intel is just not as advanced as TSMC. Period.

That said, I have no direct knowledge of Intel, but that's just my experience from other companies.

You're spot-on. These are the ones who fear the loss of their job the most, since they won't find that anymore.

Point is, a very speculative picture is being drawn out of thin air about this person's role and responsibilities, no less from a biased view.

The corporate clearance the recent times on marketing and other useless ranks, just scares such people, when actual engineers always find a company to work for easily. Knifing the bloat is dangerous for those above.

 

[Pierce2623]

With all due respect, I surely not know anyone at Intel, yet when he writes bits like this;


You clearly know, he's either unknowingly talking nonsense for whatever reason, intentionally spreads untruths or is straight-up in Lala-land, and people like him are the reason forwhy Intel is in sharp decline!

I mean, everyone knows for a fact, that we're well behind any schedule here, when Intel has to outsource to TSMC instead of manufacturing it ourselves. Intel is just not as advanced as TSMC. Period.


You're spot-on. These are the ones who fear the loss of their job the most, since they won't find that anymore.

The corporate clearance the recent times on marketing and other useless ranks, just scares such people, when actual engineers always find a company to work for easily. Knifing the bloat is dangerous for those above.

I’m sorry but all the semiconductor analysts that actually do that for a living disagree. While Intel has undoubtedly fumbled the ball lately

Did we missed something? In what alternate time-line did that happen? Or is he talking about a fever-dream of his SecondLife-commune Intel inside™ of former Intel-employees and now outsiders?

At no point in time (we're talking about actual reality here) did Intel ever move to anything EUVL prior to TSMC.

Heck, up until 2021 Intel didn't even had anything with regards to today's EUVL-standards, when being finally delivered their EUVL-packages in 2021, even if that was sneakingly accounted for by 2019 already.

TSMC already moved to 7nm-class in 2017 with their N7 node, and by the turn of 2018 was well underway towards EUVL with 10 machines like the AMSL Twinscan NXE:3600D, starting into their EUV-first N7+.

That guy got confused and meant High-NA EUV which Intel is using for 14a but TSMC have only recently gotten their first machines. It stands for “high numerical aperture” and supposedly it both increases precision on smaller discrete parts like transistors while simultaneously increasing the reticle limit. If 14a is actually on schedule, there’s a fairly good chance that Intel’s much longer experience with the high-NA machines should be sufficient enough to at least partially maintain the performance advantage 18a has over N2. But that’s just the info from semiconductor analysts. They have more info than us but 14a and A16 are both too far out for them to have reasonably accurate performance numbers like they do for 18a vs N2.

 

[dalek1234]

Fair.


I disagree, but there definitely can be a situation where "group think" occurs and the insiders all drink the kool aid.

I'm not sure that's happening, here. The only way we can be sure is to wait and see (assuming nobody does anything that kills 18A in the cradle).

"Group Think", interesting theory, probably more plausible than mine. Kind of like "The Traitors"; there is some crazy group-think there.

 

[phead128]

High-NA EUVL is merely an optical (lens) optimization, it's not a gamechanger as the shift from DUV was to Low-NA EUVL. Leave it up to the Intel engineer Product Manager to be absolutely clueless on the basics and exaggerate things. No wonder he deleted his LinkedIn post, it was so a glaring error.

18A tapeout is in less than 3 months, and engineering samples have been given to dozens of customers by now, with only AWS annoucing AI fabric chips on 18A. There hasn't been a major secured customer for 18A so far since then, and with rumors of Intel IFS being spinned out or acquired, I don't see any incentive for anyone to abandon TSMC to take a risk on 18A unless forced to by US gov't at gunpoint.

 

[bit_user]

That guy got confused and meant High-NA EUV which Intel is using for 14a but TSMC have only recently gotten their first machines. It stands for “high numerical aperture” and supposedly it both increases precision on smaller discrete parts like transistors while simultaneously increasing the reticle limit.

TSMC says high-NA machines are too expensive and not necessary for their 1.6 nm node.

https://www.tomshardware.com/tech-i...ode-in-contrast-intel-has-championed-the-tech
​

Now, a cynical take would be that it's "sour grapes", but at least they have addressed the issue.

 

[Alvar "Miles" Udell]

Don't look at the board or Trump, look at institutions, they own over 2/3 of Intel, with Blackrock, Vanguard, and State Street owning combined 20% themselves (8/8/4%). If they, among other holders, demand a split up to recoup an investment that's still not even worth half of what it was a year ago it will happen.

As a shareholder myself still deep in the red, I'd vote yes to a sale if it brought me in the $65 a share range they're (analysts) give their tangible value as at the upper end.

 

[bit_user]

Don't look at the board or Trump, look at institutions, they own over 2/3 of Intel, with Blackrock, Vanguard, and State Street owning combined 20% themselves (8/8/4%). If they, among other holders, demand a split up to recoup an investment that's still not even worth half of what it was a year ago it will happen.

Only if the sale gets approved by regulators.

BTW, the institutional investors basically control most of the board. Along with their voting shares, that's how they usually get their way.

 

[Mr Majestyk]

So many stupid comments wanting Intel to die. That would be a disaster for consumers and allowing TSMC already a monopoly effectively to own their fabs would be an even bigger disaster. I haven't owned an Intel cpu for 9 years but that does not mean I don't want a strong viable alternative to AMD who would become exactly like Intel of old, if they were the only cpu show in town.

 

[Thunder64]

So many stupid comments wanting Intel to die. That would be a disaster for consumers and allowing TSMC already a monopoly effectively to own their fabs would be an even bigger disaster. I haven't owned an Intel cpu for 9 years but that does not mean I don't want a strong viable alternative to AMD who would become exactly like Intel of old, if they were the only cpu show in town.

Who is saying they want Intel to die?

 

[Mr Majestyk]

Who is saying they want Intel to die?

The tone says it all.

 

[Thunder64]

The tone says it all.

I think you are reading too much into it. At best I think the tone might be "karma's a bitch" after the shenanigans Intel has pulled over the years. Rebates, contra-revenue, etc. No sane person wants them to die though and create a monopoly in AMD and darn near one in TSMC.

 

[jkflipflop98]

I agree. Intel has pulled parity and in some respects advanced beyond TSMC. To force Intel into some type of joint venture with TSMC (who is "the enemy") would be a huge demoralizing slap in every engineer's face.

If the government wants to help, throw INTC some government contracts. Other than that, leave Intel the hell alone. Intel is going to be just fine.

 

[RUSerious]

Which means very little vs a bachelors in computer engineering when we’re talking about process node development, especially considering we’re talking much more about logic chips at a high level view rather than things like the various chemical treatments involved in making chips.

Process engineers are mainly PhD chemists. PhD physicists do work on device physics, particularly quantum effect mitigation, during early development. EEs do most of the work on logic design. FWIW.

 

[phead128]

So many stupid comments wanting Intel to die. That would be a disaster for consumers and allowing TSMC already a monopoly effectively to own their fabs would be an even bigger disaster. I haven't owned an Intel cpu for 9 years but that does not mean I don't want a strong viable alternative to AMD who would become exactly like Intel of old, if they were the only cpu show in town.

It's not the first time Intel Foundry failed. It failed in 2018 after a 6-year run, nobody cried about TSMC monopoly back then because Samsung exists. If Intel Foundry fails again in 2025, I don't see any monopoly issue, Samsung also exist too. Intel Products is highly lucrative without that deadweight foundry dragging it down. The fear mongering can't over come the deep institutionalized incompetence in Intel leadership, Team Blue cheerleading won't change that. They need structural reform starting by ending IFS foundry and spinning that deadweight off.

 

[bit_user]

It's not the first time Intel Foundry failed. It failed in 2018 after a 6-year run, nobody cried about TSMC monopoly back then because Samsung exists.

I think that was more to do with Intel's overall financial health, which was quite strong. Also, that was before Global Foundries cancelled their 7 nm node and TSMC hadn't really pulled ahead of everyone else, either. So, there wasn't the same level of concern about lack of competition or leading-edge capacity as there is now and people probably (rightly) assumed Intel would eventually pull through.

If Intel Foundry fails again in 2025, I don't see any monopoly issue, Samsung also exist too.

If you haven't been paying attention, they've been really struggling for the past few years. A quick search will turn up a string of bad news from them.

Intel Products is highly lucrative without that deadweight foundry dragging it down.

That's why the short-term money people want to separate them so badly. I'm for separating them, too - just not before IFS has its legs under it.

 

[JayNor]

Intel funded much of the research by ASML on EUV. They also had their own internal research on EUV from as early as 2004, according to this article.

https://phys.org/news/2004-08-significant-intel-euv-lithography.html

I recall Intel was evaluating ASML EUV tools at the same time as others, but determined that it wasn't ready yet to go into their hvm due to not meeting some up-time target and not having pellicle technology until 2021. It is not hard to find that info.

I believe TSM went ahead, initially without the pellicle solution and improved their own pellicle solution. That was where they gained the advantage.

We could be seeing Intel taking the advantage now with BSPD and, perhaps, with their many years of R&D on EMIB. TSM is reportedly struggling with their CoWoS-L.

 

[phead128]

We could be seeing Intel taking the advantage now with BSPD and, perhaps, with their many years of R&D on EMIB. TSM is reportedly struggling with their CoWoS-L.

So 18A engineering samples have been distributed to dozens of potential customers for well over 6 months now. The supposed advantage of BSPD should result in secured customers announcements by now right?

 

[TerryLaze]

Intel Products is highly lucrative without that deadweight foundry dragging it down. The fear mongering can't over come the deep institutionalized incompetence in Intel leadership, Team Blue cheerleading won't change that. They need structural reform starting by ending IFS foundry and spinning that deadweight off.

The foundry is the only reason intel products are highly lucrative, they don't have the crazies that nvidia or apple has that will pay anything for a new product.
Also intel having to pay to build fabs isn't making them deadweight or financially bad decisions, you have to pay to build things.

Second also, building the fabs does not hurt intel, before they started building the fabs they would spend that amount of money on stock buy backs,
2018 10..7 bil
2019 13.6bil
2020 14.2 bil
now they can spend the same amounts of money on building fabs and declare it on their tax returns to get support and tax refunds.

https://www.intc.com/stock-info/dividends-and-buybacks

 

[bit_user]

The foundry is the only reason intel products are highly lucrative, they don't have the crazies that nvidia or apple has that will pay anything for a new product.

We can look at gross margins. AMD just closed out their FY 2024 with a gross margin of 49%.

https://www.tomshardware.com/pc-com...-the-datacenter-for-the-first-time-in-q4-2024
​

A month earlier, Intel closed their FY 2024 with a gross margin of 32.7%.

https://www.intc.com/news-events/pr...s-fourth-quarter-and-full-year-2024-financial
​

Also intel having to pay to build fabs isn't making them deadweight or financially bad decisions, you have to pay to build things.

It's the cost that determines whether it's good or bad. Revenue minus expenses = profit.

Second also, building the fabs does not hurt intel, before they started building the fabs they would spend that amount of money on stock buy backs,

That's how they got into this mess - by underinvesting in their future.

Now, even after they cut all shareholder dividends, they're still in bad shape. Prior to that, they maintained those dividends for like 2 or 3 decades. So, that was a very unprecedented move and it still wasn't enough.

now they can spend the same amounts of money on building fabs

No, both because their revenues dropped and making stuff on newer nodes is less profitable. That leaves them in a pickle, which is why the board kicked out Gelsinger and now everybody is worried the company is going to be torn apart.

 

[RUSerious]

2018 10..7 bil
2019 13.6bil
2020 14.2 bil

Are those the amounts of the stock buybacks?? If so, very short sighted. They were already having problems with both designs and node advancement.

 

[phead128]

The foundry is the only reason intel products are highly lucrative, they don't have the crazies that nvidia or apple has that will pay anything for a new product.

AMD leapfrogged Intel after GloFlo spinoff and outsourcing to TSMC. That just proves you don't need your own fabs to be wildly profitable.

 

[Pierce2623]

So 18A engineering samples have been distributed to dozens of potential customers for well over 6 months now. The supposed advantage of BSPD should result in secured customers announcements by now right?

You do realize that most customers don’t want announcements that tell the competition what they’re doing right?

 

[phead128]

You do realize that most customers don’t want announcements that tell the competition what they’re doing right?

While it's true that some customers may prefer to keep their plans confidential, it's also common in the semiconductor industry for foundries to announce secured customers, especially for strategic nodes like 18A. TSMC, Samsung, and even Intel in past generations have all publicized major customer wins to build confidence in their process technology. If 18A truly has a competitive advantage with BSPD, at least some key customers would likely have been willing to make announcements by now. The complete lack of public confirmation raises reasonable questions about traction.

 

[bit_user]

You do realize that most customers don’t want announcements that tell the competition what they’re doing right?

Well, Intel has to disclose revenue projections to their investors for IFS, so I think the rationale is that they might as well come out and announce when they get a big win. In case you haven't noticed, Intel has sometimes made these announcements without specifying exactly who the customer is, for the sake of the customer's own competitive reasons.