Intel to split off foundary business

[HockeyJon]

https://www.cnbc.com/2024/09/16/int...s-into-subsidiary-weighs-outside-funding.html

Intel shares jumped almost 10% in extended trading on Monday after the company said it plans to turn its foundry business into an independent unit with its own board and the potential to raise outside capital.

As part of CEO Pat Gelsinger’s effort to turn around the struggling chipmaker, Intel said in a memo to employees that it will also sell off part of its stake in Altera.




Gelsinger said the restructuring would allow Foundry to “evaluate independent sources of funding,” and comes days after Intel’s board met to assess the direction and future of the company. The foundry business, which Intel plans to use to manufacture chips for other customers, has been a big drag on its bottom line, with the company spending roughly $25 billion on it for the last two years.

Beyond just considering outside funding, Intel is weighing whether to spin off the foundry business, possibly into a separate publicly traded company, according to a person with knowledge of the matter who declined to be named to discuss confidential information.

With a standalone “operating board” and a cleaner corporate structure, the mechanics of a separation become far easier than trying to turn a fully-integrated unit into a separate company.

Intel has lost almost 60% of its value this year, as the company has lost market share in its core PC and data center market and watched Nvidia run away with the market for chips that power artificial intelligence workloads. Last month, Intel reported disappointing quarterly results, sparking the sharpest selloff in 50 years, and said it would lay off over 15% of its workforce as part of a $10 billion cost-reduction plan.

Intel will also pause its fab efforts in Poland and Germany “by approximately two years based on anticipated market demand,” Gelsinger said. The chipmaker will also pull back on its plans for its Malaysian factory. U.S. manufacturing projects will remain unaffected, the company said.




The company is also roughly halfway towards the layoffs announced in August, Gelsinger said.

Intel also announced on Monday that it had entered into a deal with Amazon Web Services to produce custom chips for AI.

The move is a vote of confidence in Intel’s quest to manufacture custom chips for companies in its foundry business, in addition to designing its own products.

It extends a long-running partnership between the two companies. Amazon is a large end customer of Intel chips to power its AWS servers, and will buy a custom Xeon processor from Intel as well, Intel said.

It also will allow Intel a new foothold in the growing industry for AI server chips. While Intel has several products that can be used for AI, including one called Gaudi 3, Nvidia has largely taken control of the market.

Amazon has developed its own AI chips, including one called Trainium, for over five years. Companies such as Amazon, Microsoft, and Google have invested heavily in custom chips to run AI in hopes that theirs would be less expensive or offer other advantages over than Nvidia’s general-purpose GPUs.

The company said that the chips would be made on the company’s 18A process, which is expected to be used in production in 2025. Analysts expect the manufacturing technology to be similar to TSMC’s forthcoming 2nm process. (The first chips using TSMC’s 3nm process are shipping now in Apple’s iPhone.)

Intel said that it would perform its most advanced manufacturing, including the AI chip for AWS, at its plant in Ohio that’s currently under construction.

“All eyes will remain on us. We need to fight for every inch and execute better than ever before. Because that’s the only way to quiet our critics and deliver the results we know we’re capable of achieving,” Gelsinger said.

I had a feeling this was on the table. Hopefully this will be a good first step toward getting back into shape for Intel's now two divisions, but we'll see.

 

[cdabc123]

Very bad.. They are the single us based organization on the forefront of lithography. They cannot go in the way of amd and focus on design using tsmc exclusively.

Honestly one of the few companies the US should have continued to throw money at till they thrive again. The stock price is criminally low, in the US based market it should have been kept way higher if stocks were rational.

Absolutely disgraceful effort from intels management, board, and pat. There was no reason they couldn't have put intel in the position of a good silicon company facing some minor troubles. This is near catastrophic.

Altera on the chopping block too.. fpgas are a growing industry and the technology is exactly what intel should be investing in the their future. Fpgas are used heavily for military applications, the US should have ensured intel was able to manufacture, develop, and fab these devices.

Not good, hope Taiwan doesn't fall or shift too heavily to China.

 

[ochadd]

TSMC is going to rule the world if this trend continues.

 

[Gideon]

Very bad.. They are the single us based organization on the forefront of lithography. They cannot go in the way of amd and focus on design using tsmc exclusively.

Honestly one of the few companies the US should have continued to throw money at till they thrive again. The stock price is criminally low, in the US based market it should have been kept way higher if stocks were rational.

Absolutely disgraceful effort from intels management, board, and pat. There was no reason they couldn't have put intel in the position of a good silicon company facing some minor troubles. This is near catastrophic.

Altera on the chopping block too.. fpgas are a growing industry and the technology is exactly what intel should be investing in the their future. Fpgas are used heavily for military applications, the US should have ensured intel was able to manufacture, develop, and fab these devices.

Not good, hope Taiwan doesn't fall or shift too heavily to China.

Innovate or die, it's the way of business. However the fab side being separated from Intel should actually allow them to land contracts. They just need to learn to market themselves and show they can deliver. This also allows Intel to feel more free to pick and choose the process they want to use.

 

[LukeTbk]

The stock price is criminally low, in the US based market it should have been kept way higher if stocks were rational.

How much debt did you took to invest in way too criminally low stock ? With how much leverage in stock options ?

This a company that missed back to back the mobile and then the GPUs trend on the design side and the foundry is a bit trouble on the other side, with now around $25 billions of net debt.

the US should have ensured intel was able to manufacture, develop, and fab these devices.

They have spend a lot to do so (and splitting the 2 company could open the door for the government to only care about the foundry side, putting the other one at risk of not being saved by a next chips acts type), but the move to on-shore TSMC and Samsung advanced node foundry is an other way to do it:

https://semiconductor.samsung.com/u...ke ground on the,-performance computing (HPC).

But it is really hard for American workforce-supply chain to compete with Taiwan, end of the day are those American base foundry Intel-Samsung or TSMC, being physically in the US (or strong allied) can be more what matter, it is not like it is likely for all the needed supply chain to be on-shored.

A bit like the most American car made outside Tesla tend to be Honda-Toyota, not Ford-GM-Chrysler now.

 

[cdabc123]

Innovate or die, it's the way of business. However the fab side being separated from Intel should actually allow them to land contracts. They just need to learn to market themselves and show they can deliver. This also allows Intel to feel more free to pick and choose the process they want to use.

They needed to drag the fabs along, till they engeneer a node good enough to be competitive. Roi is brutal for fabs, intels choice to make abunch of mistakes and begin to remove themselves from the fab space is immensely short sighted.

There is no reason intel will land any more contracts, they have already been open for business and working with companies to fab silicon. Fpgas, custom asics, nics, gpus, hpc chips, and of course the fantastic x86 cpu.

I agree innovation is essential. Intel has been struggling when they really needed to focus on being a engineering company, not bloated big business, with a board behind their blunders.

Still a very very poor move in my opinion

 

[Legendary Gamer]

Very bad.. They are the single us based organization on the forefront of lithography. They cannot go in the way of amd and focus on design using tsmc exclusively.

Honestly one of the few companies the US should have continued to throw money at till they thrive again. The stock price is criminally low, in the US based market it should have been kept way higher if stocks were rational.

Absolutely disgraceful effort from intels management, board, and pat. There was no reason they couldn't have put intel in the position of a good silicon company facing some minor troubles. This is near catastrophic.

Altera on the chopping block too.. fpgas are a growing industry and the technology is exactly what intel should be investing in the their future. Fpgas are used heavily for military applications, the US should have ensured intel was able to manufacture, develop, and fab these devices.

Not good, hope Taiwan doesn't fall or shift too heavily to China.

China will eventually catch up. They just hit 65nm Lithography (they were doing 90nm in house).

Spinning the Fab off can mean only one thing. They bit off more than they could chew. They have fallen so far behind that they will never recover. The writing was on the wall with 10nm. It's yields took too long to become competitive and it never delivered on it's promises. By the time it was viable, it was obsolete compared to the competition. Yields on 7nm must be total poo.

I knew something was wrong when they rebranded their die shrinks to things like Intel 7 and 5 and 20A.... Cryptic B.S. to cover up how far behind their Fabs were.

It will be interesting to see Intel Fabless but this is a end of the road strategy. If they fail at all here, we may very well watch the company fold.

Not good, not good at all. Horribly mismanaged, sat on their asses for far too long and never innovated. ugh

 

[cdabc123]

How much debt did you took to invest in way too criminally low stock ? With how much leverage in stock options ?

This a company that missed back to back the mobile and then the GPUs trend on the design side and the foundry is a bit trouble on the other side, with now around $25 billions of net debt.


They have spend a lot to do so (and splitting the 2 company could open the door for the government to only care about the foundry side, putting the other one at risk of not being saved by a next chips acts type), but the move to on-shore TSMC and Samsung advanced node foundry is an other way to do it:

https://semiconductor.samsung.com/us/sas/company/taylor/#:~:text=Samsung broke ground on the,-performance computing (HPC).

But it is really hard for American workforce-supply chain to compete with Taiwan, end of the day are those American base foundry Intel-Samsung or TSMC, being physically in the US (or strong allied) can be more what matter, it is not like it is likely for all the needed supply chain to be on-shored.

A bit like the most American car made outside Tesla tend to be Honda-Toyota, not Ford-GM-Chrysler now.

No debt, I don't gamble aggressively on stocks. Although my intc position has certainly not been as prosperous as nvda or amd.

Intel is worth more then 100b from a national defense standpoint alone.. between fpgas, hpc, networking, and the traditional x86 they are extremly valuable. Their stock is not a fair representation of this. And this move to spinoff is boneheaded and rooted in investors, boards and businessmen shortsightedness. The vertical integration and having cutting edge US based fabs is absolutely essential.

 

[LukeTbk]

I agree innovation is essential. Intel has been struggling when they really needed to focus on being a engineering company, not bloated big business, with a board behind their blunders.

Pat and his decision, that exactly Intel being focused as a engineering company lead by an electric Engineer/computer science guy as a CEO instead of a bean counter.

No debt, I don't gamble aggressively on stocks.

usually a sign that part of you agree on the price being possibly a good representation

 

[cdabc123]

Pat and his decision, that exactly Intel being focused as a engineering company lead by an electric Engineer/computer science guy as a CEO instead of a bean counter.

I disagree, he is a ceo and buisnessman first. If intel was engineering focused we would have seen leaps more development in the last 5 years. They are running like big business. Amd, Nvidia, apple, are all out engineering them.

 

[Zarathustra[H]]

Yeah, there were some news rumor posts a few weeks ago that they were considering this. Given their financial situation it does not surprise me.

TSMC is going to rule the world if this trend continues.

Very bad.. They are the single us based organization on the forefront of lithography. They cannot go in the way of amd and focus on design using tsmc exclusively.

"Splitting off" doesn't necessarily mean the fabs will turn to shit like they did with Global foundries, but they will likely need some serious investment to be competitive again.

There is something to be said for narrowing ones focus. Both latest node fab development and latest node chip design continue to become more and more complicated, and there is something to be said for being able to focus on one of them without getting distracted by the other.

The question is, can an independent ex Intel Fab business regain the focus and investment it needs to become competitive with the likes of TSMC and Samsung? Or will it b absorbed into some third party and become irrelevant sort of like Globalfoundries?

I have a feeling Uncle Sam might have their hands forced in stepping in and propping up the foundry business, what with all the money that has been provided to Intel under the chips act, and the desire to not be dependent on Taiwan for cutting edge defense tech.

This makes me concerned, but then again, maybe we will be lucky again, and repeat the GM bailout success with the Federal government regaining most of the money it spent to bail out GM in the end.

 

[LukeTbk]

I disagree, he is a ceo and buisnessman first. If intel was engineering focused we would have seen leaps more development in the last 5 years. They are running like big business. Amd, Nvidia, apple, are all out engineering them.

This seem using the conclusion for the attempt. How much more a deep tech guy to run Intel could they have picked.... They did try to move to a tiles system, make a entry in the gpu market and what not.

They did try to reach Samsung-TSMC level on the foundry side and spent a giant amount on it to do so and failed, I am not sure what running like a big business mean here when we talk about an over 100 billions company, Amd-Nvidia-Apple are all running like big business as well, as they are one.

 

[cdabc123]

China will eventually catch up. They just hit 65nm Lithography (they were doing 90nm in house).

Spinning the Fab off can mean only one thing. They bit off more than they could chew. They have fallen so far behind that they will never recover. The writing was on the wall with 10nm. It's yields took too long to become competitive and it never delivered on it's promises. By the time it was viable, it was obsolete compared to the competition. Yields on 7nm must be total poo.

I knew something was wrong when they rebranded their die shrinks to things like Intel 7 and 5 and 20A.... Cryptic B.S. to cover up how far behind their Fabs were.

It will be interesting to see Intel Fabless but this is a end of the road strategy. If they fail at all here, we may very well watch the company fold.

Not good, not good at all. Horribly mismanaged, sat on their asses for far too long and never innovated. ugh

China may catch up by assimilating the largest and most advanced semiconductor fab in the world. Or atleast shift its priorities enough to get some good chips from them. They can't compete much the same as Intel has been struggling substantially to compete.

I agree with most of the rest of your post. I was under the hopeful assumption that Intel 7nm would be competitive enough to revive their x86 desktop and server effort. Mobile is cutthroat. Hence Apple thriving on tsmc 3nm. If like you said 7nm is poo too, this move would make more sense but would still be a very grim option. Intel can't save face by spinning off their fabs like amd did. Intel 14th gen is way better then amd bulldozer.

 

[Zarathustra[H]]

China may catch up by assimilating the largest and most advanced semiconductor fap in the world. Or atleast shift its priorities enough to get some good chips from them. They can't compete much the same as Intel has been struggling substantially to compete.

I agree with most of the rest of your post. I was under the hopeful assumption that Intel 7nm would be competitive enough to revive their x86 desktop and server effort. Mobile is cutthroat. Hence Apple thriving one tsmc 3nm. If like you said 7nm is poo too, this move would make more sense but would still be a very grim option. Intel can't save face by spinning off their fabs like amd did. Intel 14th gen is way better then amd bulldozer.

lol @ "semiconductor fap"

 

[cdabc123]

This seem using the conclusion for the attempt. How much more a deep tech guy to run Intel could they have picked.... They did try to move to a tiles system, make a entry in the gpu market and what not.

They did try to reach Samsung-TSMC level on the foundry side and spent a giant amount on it to do so and failed, I am not sure what running like a big business mean here when we talk about an over 100 billions company, Amd-Nvidia-Apple are all running like big business as well, as they are one.

Idk I think pat has a track record at this point for not being focused on engineering and development. He did almost nothing for VMware, where development in the software space is much easier then fabs and silicon.

Lisa su, and her endeavors seem to be much more engineering focused and amd appears to be a lean, thriving company. They appear to be doing far better in the fpga space with the xilinx accusition, Intel has altera for years longer and did nothing but sit back abit on the technology and let the companies development go on relatively unimpended.

VMware was sold, global founderies sold, and now Intel Fab and altera are on the chopping block? That is a very bad sign. Pat messed up by not laying off as many businessmen as possible and hiring flocks of engineers when he arrived.

 

[hititnquitit]

Without proper leadership and talent this spinoff is doomed. I have no idea who they plan on placing in charge of this new division but I have no faith in Gelsinger. I can see the government pulling it's investment in the foundry business tomorrow. Pat screwed this up in a bad way. With so much riding on the foundry business succeeding...this looks so bad.

 

[LukeTbk]

"famously", intel has more software engineer than AMD complete staff, not sure if it is still true in 2024, but in 2023 it was still really big

In a recent episode of the FLOSS Weekly podcast, Gupta highlights the numbers: “We have 19,000 plus software engineers at Intel. People think of us as a silicon company, but no: We have almost 20,000 software engineers.

Often at the very top of the most contributions on Linux and other open source project is from Intel, I am not sure a lack of engineers was ever an issue for them, arguably it became a bit of a crutch that made putting giant custom by application effort in managing P vs E core and so on possible and extremely costly, they have back in the giant margin days staff count it seems.

 

[cdabc123]

"famously", intel has more software engineer than AMD complete staff, not sure if it is still true in 2024, but in 2023 it was still really big

In a recent episode of the FLOSS Weekly podcast, Gupta highlights the numbers: “We have 19,000 plus software engineers at Intel. People think of us as a silicon company, but no: We have almost 20,000 software engineers.

Often at the very top of the most contributions on Linux and other open source project is from Intel, I am not sure a lack of engineers was ever an issue at Intel.

Don't they need more hardware engineers, too many software engineers talking to their rubber ducks? I haven't seen any software development focused on altera, they couldn't quickly resolve their software kernel issues with Microsoft which made them look bad. Is it miss management of engineers?

 

[LukeTbk]

Possibly that they are trying to do 2 of the hardest thing in the business at the same time and competing directly with some really good company at it like Apple-TSMC-Nvidia-AMD.

Samsung and everyone else was not able really able to keep up with TSMC either, arguably not many kept up with Nvidia at designing chips for money part and for a complete solution Apple is really something, like a team losing the Superbowl we can come up with a long list of reason, but we need to keep in mind that it is really hard and any strategy and management could have failed.

They did miss on mobile and on GPUs straight on putting them in a very hard position no matter what and do not seem particularly well-placed in the AI money race, they are very bloated as an old company that operate in sector in massive declines (In 1999 would have thought that the world PC market would have peaked by 2011.... I think the desktop peaked maybe around 2000, but we are talking laptop+desktop combined...) can easily be:

It seem really hard for anyone to not have issues with windows schedulers, they could very well have assigned a lot of good people, with a lot of time and money and yet still failed at it. It is possible to just fail, specially for very hard to do things vs great competition.

 

[KazeoHin]

I knew something was wrong when they rebranded their die shrinks to things like Intel 7 and 5 and 20A.... Cryptic B.S. to cover up how far behind their Fabs were.

This

So many people on here trying to justify it, where are they now?

 

[LukeTbk]

Is Intel 7 or 20A more cryptic than TSMC 4NX or Samsung 8 ? What would be a non-cryptic name for a node ?

 

[1_rick]

I knew something was wrong when they rebranded their die shrinks to things like Intel 7 and 5 and 20A.... Cryptic B.S. to cover up how far behind their Fabs were.

OTOH, compare (IIRC) Micron who didn't want to admit they couldn't shrink their processes, so started calling them 1X, 1Y, 1Z, 1α.

 

[kac77]

Is Intel 7 or 20A more cryptic than TSMC 4NX or Samsung 8 ? What would be a non-cryptic name for a node ?

Well Intel 7 wasn't actually 7nm it was 10 nm. Intel's 7nm was renamed Intel 4. So no completely different names.

 

[Lakados]

Well Intel 7 wasn't actually 7nm it was 10 nm. Intel's 7nm was renamed Intel 4. So no completely different names.

And??? TSMC 3nm is 13nm, and Intel 18A is 8nm, the names have all been made up since 90nm, naming for 90nm and everything that comes after is made up as there is a nearly infinite number of ways to measure some arbitrary aspect of the lithography process to justify the number placed on the marketing name for it.

 

[LukeTbk]

Well Intel 7 wasn't actually 7nm it was 10 nm. Intel's 7nm was renamed Intel 4. So no completely different names.

I feel it has been a very long time since the name of the node was linked to the actual size of the transistor.

Was TSMC 7 twice as small than intel 14 ? (apparently, the scan above show that Intel 14nm is like 24nm gate width on average and TSMC 7nm is about 22nm gate width.)
https://www.techpowerup.com/272489/...-7-nm-node-using-scanning-electron-microscope
The 14 nm transistor isn't 14 nm in width, and the 7 nm transistor isn't 7 nm wide. The naming of the node and actual size of the node have had a departure a long time ago, and the naming convention is really up to the manufacturer - it's become more of a marketing gimmick than anything else. This is the reason researchers have already proposed another density metric for semiconductor technology other than pure "nm" terms.

From my understanding node generation is more density base than actual size of transistor and density change for the type (memory-cache-sram vs logic vs something else) making it an average depending of their usage and using a short name that would not be just the marketing name for it would be hard.

it is not like the client are not very well-informed people here, that rely in any way on the naming.

The first generation of TSMC 4 and the latest Blackwell tsmc 4nx (or what they are naming it now) could have different sized transistor, the density is quite different at least, would it be better than the first months of the first tsmc 3 generation.... maybe)

 

[idiomatic]

Intel has survived this long while being a shite processor company by being a pretty transcendent *process* company. Without that edge...

 

[Lakados]

This

So many people on here trying to justify it, where are they now?

Is Intel 7 or 20A more cryptic than TSMC 4NX or Samsung 8 ? What would be a non-cryptic name for a node ?

OTOH, compare (IIRC) Micron who didn't want to admit they couldn't shrink their processes, so started calling them 1X, 1Y, 1Z, 1α.

In 2021 it was conventionally agreed by the IEEE that some numbers would denote generational improvements, 10, 7, 5, 3, 2, and everything falling between them was an improvement on the previous generation but not a new generation.
So the node names have less to do with the actual size as it was never about the size, but branding on a generation, because saying this was done on an 18'th generation process vs a 19'th generation process is more confusing for most.

I mean look at what the IEEE has agreed means 3nm:

The term "3 nanometer" has no direct relation to any actual physical feature (such as gate length, metal pitch or gate pitch) of the transistors. According to the projections contained in the 2021 update of the International Roadmap for Devices and Systems published by IEEE Standards Association Industry Connection, a 3nm node is expected to have a contacted gate pitch of 48 nanometers, and a tightest metal pitch of 24 nanometers.[12]

So Intel renaming their nodes simply brought them inline with where their hardware fell in the generational gap.

In regards to spinning off the fabs I was under the impression that it was always the goal of the IDM2.0 plan to spin them off.

 

[kac77]

And??? TSMC 3nm is 13nm, and Intel 18A is 8nm, the names have all been made up since 90nm, naming for 90nm and everything that comes after is made up as there is a nearly infinite number of ways to measure some arbitrary aspect of the lithography process to justify the number placed on the marketing name for it.

What do you mean? Who is ahead TSMC or Intel in manufacturing? I mean it's great to go with Intel's claims but performance wise has it matched up with reallity? If it (TSMC) was was so inferior how did AMD catch up? Why hasn't Intel locked in customers if its so much better? Broadcom surely isn't happy right now.

 

[kac77]

I feel it has been a very long time since the name of the node was linked to the actual size of the transistor.

View attachment 680136

Was TSMC 7 twice as small than intel 14 ? (apparently, the scan above show that Intel 14nm is like 24nm gate width on average and TSMC 7nm is about 22nm gate width.)
https://www.techpowerup.com/272489/...-7-nm-node-using-scanning-electron-microscope
The 14 nm transistor isn't 14 nm in width, and the 7 nm transistor isn't 7 nm wide. The naming of the node and actual size of the node have had a departure a long time ago, and the naming convention is really up to the manufacturer - it's become more of a marketing gimmick than anything else. This is the reason researchers have already proposed another density metric for semiconductor technology other than pure "nm" terms.

From my understanding node generation is more density base than actual size of transistor and density change for the type (memory-cache-sram vs logic vs something else) making it an average depending of their usage and using a short name that would not be just the marketing name for it would be hard.

it is not like the client are not very well-informed people here, that rely in any way on the naming.

The first generation of TSMC 4 and the latest Blackwell tsmc 4nx (or what they are naming it now) could have different sized transistor, the density is quite different at least, would it be better than the first months of the first tsmc 3 generation.... maybe)

Didn't say it wasn't. But Intel's claims of performance vs reality isn't exactly rock solid. Again, if Intel was so great then it would have customers lining up around the block. It doesn't and there's a reason for that.

 

[KazeoHin]

In 2021 it was conventionally agreed by the IEEE that some numbers would denote generational improvements, 10, 7, 5, 3, 2, and everything falling between them was an improvement on the previous generation but not a new generation.
So the node names have less to do with the actual size as it was never about the size, but branding on a generation, because saying this was done on an 18'th generation process vs a 19'th generation process is more confusing for most.

I mean look at what the IEEE has agreed means 3nm:

The term "3 nanometer" has no direct relation to any actual physical feature (such as gate length, metal pitch or gate pitch) of the transistors. According to the projections contained in the 2021 update of the International Roadmap for Devices and Systems published by IEEE Standards Association Industry Connection, a 3nm node is expected to have a contacted gate pitch of 48 nanometers, and a tightest metal pitch of 24 nanometers.[12]

So Intel renaming their nodes simply brought them inline with where their hardware fell in the generational gap.

In regards to spinning off the fabs I was under the impression that it was always the goal of the IDM2.0 plan to spin them off.

In other words, they were trying as hard as they could to tell the world Everything is Fine™

Very good work on explaining exactly what Intel explained in their press releases! Such an amazing ability to repeat exactly what Intel wanted people to think! You're right: That is exactly what Intel said and hoped everyone would believe.



Now they're bleeding money and cannibalizing their fabs, using TSMC to manufacture 99.9% of their inventory, So really the whole 'renaming to be inline with the competition' was truly innocent and was in no way trying to cover up their incompetence, NO WARNING SIGN AT ALL NOPE Everything is Fine™

 

[LukeTbk]

What do you mean? Who is ahead TSMC or Intel in manufacturing? I mean it's great to go with Intel's claims but performance wise has it matched up with reallity? If it (TSMC) was was so inferior how did AMD catch up? Why hasn't Intel locked in customers if its so much better? Broadcom surely isn't happy right now.

is there currently used in the field Intel node has a name that suggest it is better than TSMC ? 14900k is on Intel10, the core ultra latest intel laptop are on Intel 7 (are those significantly worse than the old TSMC 10 and 7 chips... ?)

are you talking about from all we know node that did not work to scale manufacturing wise and no idea how good they are ?

Didn't say it wasn't. But Intel's claims of performance vs reality isn't exactly rock solid. Again, if Intel was so great then it would have customers lining up around the block. It doesn't and there's a reason for that.

Was performance the issue or yield-price, etc... ? Which node are we talking about, what was the claim and what was the actual performance ?

In regards to spinning off the fabs I was under the impression that it was always the goal of the IDM2.0 plan to spin them off.

Maybe, but for a while there a talk about being different division with foundry needing to beat the competition to get contract from the design team instead of 2 different company, would it had work I can imagine keeping it like that

 

[Lakados]

In other words, they were trying as hard as they could to tell the world Everything is Fine™

Very good work on explaining exactly what Intel explained in their press releases! Such an amazing ability to repeat exactly what Intel wanted people to think! You're right: That is exactly what Intel said and hoped everyone would believe.



Now they're bleeding money and cannibalizing their fabs, using TSMC to manufacture 99.9% of their inventory, So really the whole 'renaming to be inline with the competition' was truly innocent and was in no way trying to cover up their incompetence, NO WARNING SIGN AT ALL NOPE Everything is Fine™

A turd by any other name would smell as rank.

They brought their naming inline with what everybody else was already doing, doesn't change the fact that their design process isn't as good.
TSMC uses an Nvidia powered AI to design chips for their fab and it's good, really good.
Intel uses what it uses but it really isn't that good, TSMC has done a far better job at improving its gates, power delivery, voltage regulation, etc.
Intel has tunnel vision from making Intel processors, TSMC has learned a few tricks as it's circled the block a few times making everything for everybody, and they have incorporated those improvements into their processes for all their customers to use.

Give Intel and TSMC the exact same hardware to make their chips and TSMC will make the better chip because they have a better design process.

Node names are completely irrelevant to the process and they never have been.

 

[kac77]

is there currently used in the field Intel node has a name that suggest it is better than TSMC ?

Intel's renaming implies that inherently.

14900k is on Intel10, the core ultra latest intel laptop are on Intel 7 (are those significantly worse than the old TSMC 10 and 7 chips... ?)

Um, yeah. Intel isn't leading in mobile, or server at this point, and barely in desktop. You don't see AMD recalling two generations of chips do you? If their manufacturing process was that much better they would be ahead given their size.

Was performance the issue or yield-price, etc... ? Which node are we talking about, what was the claim and what was the actual performance ?

Manufacturing at scale is about yield, performance, and time to market. Intel hasn't been hitting those targets for quite some time. If it were people would be using it over TSMC. It really is that simple.

 

[Lakados]

What do you mean? Who is ahead TSMC or Intel in manufacturing? I mean it's great to go with Intel's claims but performance wise has it matched up with reallity? If it (TSMC) was was so inferior how did AMD catch up? Why hasn't Intel locked in customers if its so much better? Broadcom surely isn't happy right now.

In theory Intel should be ahead as they spent tens of Billions securing the latest and greatest ASML fabrication equipment. So TSMC is using the old NXE 3600 D series hardware for their 2 and 3 nm processes, but Intel is using the brand new hot off-the-line ASML EXE 5000 series which is better faster stronger with a tighter resolution, and blah blah blah.

So Intel should be ahead with the 18A series, but TSMC has nicer design software, a proven track record, and all those things that big companies who are betting 10's of Billions of dollars on want.

Intel only got their 18A design software out the door to partners earlier this year, and nobody is getting anything worth putting on that node designed in a few months, Blackwell's design work started 4-6 years ago depending on the leaks, and that is only coming out now. TSMC has had its designs decoupled from the nodes for almost a decade now, Intel only recently made that change. Intel has been fishing for a whale to make their IFS profitable when really they needed a lot of little fish testing the waters.

Intel should have been trying to get Broadcom or Qualcomm to release their new lineup of microcontrollers and ARM chips on one of those nodes, most of those are still on Samsung or TSMC 12nm or larger, putting them on Intel 10,7,4 could have been an easy win, it would probably struggle to break even, but a chance of breaking even is a hell of a lot better than sitting idle and eating losses. It would at least show the world that it can be done, and they would then have 2 large and happy clients who could put larger more profitable projects in later after they were comfortable.

 

[KazeoHin]

A turd by any other name would smell as rank.

They brought their naming inline with what everybody else was already doing, doesn't change the fact that their design process isn't as good.
TSMC uses an Nvidia powered AI to design chips for their fab and it's good, really good.
Intel uses what it uses but it really isn't that good, TSMC has done a far better job at improving its gates, power delivery, voltage regulation, etc.
Intel has tunnel vision from making Intel processors, TSMC has learned a few tricks as it's circled the block a few times making everything for everybody, and they have incorporated those improvements into their processes for all their customers to use.

Give Intel and TSMC the exact same hardware to make their chips and TSMC will make the better chip because they have a better design process.

Node names are completely irrelevant to the process and they never have been.

The issue I see here is that if you're lifting weights, and someone else is lifting a bunch of weights that say "200" but are actually fake "10" plates, the appropriate response as a market LEADER is to show how powerful and strong you are by ignoring the fakes. If they are lifting fake weights, then their results wouldn't be as good as yours, and time will show that. Confidence and competence will shine through.

As an insecure market FOLLOWER though, you quickly get some of those fake ego-plates and follow the same pattern as the other guy, because you don't have the competence to show results or the confidence to get better.

 

[Lakados]

The issue I see here is that if you're lifting weights, and someone else is lifting a bunch of weights that say "200" but are actually fake "10" plates, the appropriate response as a market LEADER is to show how powerful and strong you are by ignoring the fakes. If they are lifting fake weights, then their results wouldn't be as good as yours, and time will show that. Confidence and competence will shine through.

As an insecure market FOLLOWER though, you quickly get some of those fake ego-plates and follow the same pattern as the other guy, because you don't have the competence to show results or the confidence to get better.

Closer to each is benching 200, but TSMC is happily doing sets of 20 while Intel struggles to do 10.

 

[LukeTbk]

Intel's renaming implies that inherently.

Yes but which node that was renamed that did not match the competition are we talking about ? The original post talk about A20 and so on, do we have any idea of their performance of those ?

Um, yeah. Intel isn't leading in mobile, or server at this point, and barely in desktop. You don't see AMD recalling two generations of chips do you? If their manufacturing process was that much better they would be ahead given their size.

I am not sure to follow, Intel naming that they use is not imply that they are close to lead, mobile competition are on second generation TSMC 3 while Intel is on Intel 7, Desktop competition are on TSMC 4, Intel is on intel 10, that not signaling Intel claiming that they kept up with TSMC. Maybe I am missing the latest Intel product with a different node.

Manufacturing at scale is about yield, performance, and time to market. Intel hasn't been hitting those targets for quite some time. If it were people would be using it over TSMC. It really is that simple.

Yes, why the claim about the performance of those node being the issue and not as good as advertised... I mean maybe, do we know ?

 

[ChadD]

Very bad.. They are the single us based organization on the forefront of lithography. They cannot go in the way of amd and focus on design using tsmc exclusively.

Honestly one of the few companies the US should have continued to throw money at till they thrive again. The stock price is criminally low, in the US based market it should have been kept way higher if stocks were rational.

Absolutely disgraceful effort from intels management, board, and pat. There was no reason they couldn't have put intel in the position of a good silicon company facing some minor troubles. This is near catastrophic.

Altera on the chopping block too.. fpgas are a growing industry and the technology is exactly what intel should be investing in the their future. Fpgas are used heavily for military applications, the US should have ensured intel was able to manufacture, develop, and fab these devices.

Not good, hope Taiwan doesn't fall or shift too heavily to China.

Oh no. The time to turn Intel around was actually back when they hired Pat. He has 100% screwed up. If they fired him tomorrow the stock would go up another 20%.

This is just the beginning. Nothing is going to much change for Intel... and within the next 6 month. I suspect Intel will be the Fab business and that is it. The rest of Intel is probably going to get parceled out into divisions at other companies. I think its actually a very real possibility some combo of companies buys all but the fab business. My money is on Qualcomm and Microsoft... MS taking the AI/GPU design team, Qualcomm the x86 design. Companies like Nvidia may even be interested. I mean at this point Intels Enterprise value is 110 Billion... but that is as a whole not two pieces, a merging company could still complete a full spin off of the fabs. MS has around that in cash on hand. Don't be shocked if a hostile play for Intel happens before the end of the year.

 

[ChadD]

Innovate or die, it's the way of business. However the fab side being separated from Intel should actually allow them to land contracts. They just need to learn to market themselves and show they can deliver. This also allows Intel to feel more free to pick and choose the process they want to use.

Intel needs to learn how to fab things other then CPUs. The largest issue with their fabs IMO.. is Intel was always fine with fairly high defect rates. I mean a 30% defect rate isn't a big deal if you can recover 80% othe 30% by fusing cache/cores and selling the silicon anyway. Broadcomm as not happy as things they fab probably have to actually be fully functional... a 30% defect rate is a 30% write off rate.

Hopefully an official spin off of the fabs gives them a little more leaway to try and rise to the challenge of fabbing those other types of chips.

 

[ChadD]

They needed to drag the fabs along, till they engeneer a node good enough to be competitive. Roi is brutal for fabs, intels choice to make abunch of mistakes and begin to remove themselves from the fab space is immensely short sighted.

There is no reason intel will land any more contracts, they have already been open for business and working with companies to fab silicon. Fpgas, custom asics, nics, gpus, hpc chips, and of course the fantastic x86 cpu.

I agree innovation is essential. Intel has been struggling when they really needed to focus on being a engineering company, not bloated big business, with a board behind their blunders.

Still a very very poor move in my opinion

Intel last year purposelly started operating their Fabs as a seperate company internally. This probably hurts their own design teams to be fair. It still hasn't really attracted any big clients... Broadcomm ran some tests and noped out when they realized Intel wasn't going to much improve their defect rate. The issue is no one wants to contract trade secrets to their competition, or trust their companies future to the competition. Nvidia does not want to fab their latest greatest for '26 in Intels foundry in '24. Same goes for Intel Quallcomm Apple. I mean do you think Apple wants to employee a team of engineers to work WITH Intel engeineers to fabricate a M4? Intel frankly stupidly thought they could be both a design and fabrication business. Nvidia isn't going to trust the ENTIERITY of their company to a direct compeditor. Imagine... Intels fab never quite gets it right or fabs you an under performaing generation. Why would you put your fate as a company in the hand of the people that benifit if you fail?