You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
We keep trying to find ways to make the Jets and Panthers competitive with the Chiefs.
Customers are free to bet on the Jets and Panthers to win the Superbowl. No one is stopping them. For some reason, they do not.
The Chiefs (Nvidia, Apple, TSMC) do not win because of government backing.
True but TSMC, Nvidia, and Apple have been at it for 30+ years. Intel and Samsung foundry are rising from the ashes so they will need to get creative.
Knowing Lip-Bu, I'm expecting other big potential customer investments in Intel like Softbank. Yes it is dilutive for share holders but hopefully there is a big upside around the corner. These investments bolster the image of Intel (trust) and are a big step forward in making Intel Foundry relevant again.
Again, I do not see any alternatives if we want a competitive foundry landscape.
True but TSMC, Nvidia, and Apple have been at it for 30+ years. Intel and Samsung foundry are rising from the ashes so they will need to get creative.
Knowing Lip-Bu, I'm expecting other big potential customer investments in Intel like Softbank. Yes it is dilutive for share holders but hopefully there is a big upside around the corner. These investments bolster the image of Intel (trust) and are a big step forward in making Intel Foundry relevant again.
Again, I do not see any alternatives if we want a competitive foundry landscape.
Alternatives ? Several of us have repeatedly suggested that IFS needs to be split from Intel products in order to produce a viable foundry company. In all the excitement of the ever increasing Intel newsflow we appear to have lost track of some basic business fundamentals. Why would a company - nVidia for example - willingly buy chips from another company that is actively seeking (albeit not very successfully so far) to sell similar chips ? Every order placed with IFS is a potential cross-subsidy for Intel products.
No doubt the market can be distorted in the short/medium term to try to force certain outcomes. But I doubt this is sustainable long term. And certainly not without unintended consequences we are yet to discover.
This sort of intellectual detachment from basic principles doesn't feel that far away from Trump's apparent shift from the longstanding US policy of strict export restrictions on sensitive technologies to potentially hostile regimes to opportunistically trying to take a cut on any potential deal going regardless of long term consequences.
Alternatives ? Several of us have repeatedly suggested that IFS needs to be split from Intel products in order to produce a viable foundry company. In all the excitement of the ever increasing Intel newsflow we appear to have lost track of some basic business fundamentals. Why would a company - nVidia for example - willingly buy chips from another company that is actively seeking (albeit not very successfully so far) to sell similar chips ? Every order placed with IFS is a potential cross-subsidy for Intel products.
No doubt the market can be distorted in the short/medium term to try to force certain outcomes. But I doubt this is sustainable long term. And certainly not without unintended consequences we are yet to discover.
This sort of intellectual detachment from basic principles doesn't feel that far away from Trump's apparent shift from the longstanding US policy of strict export restrictions on sensitive technologies to potentially hostile regimes to opportunistically trying to take a cut on any potential deal going regardless of long term consequences.
I guess I don't understand what splitting is. They are already functionally split, different financials, the same as Samsung Foundry is split from Samsung electronics. Do you mean Intel Products and Intel Foundry have no connection whatsoever? That would be like splitting Siamese twins. Even if they both survived there would still be a strong connection for years to come.
What I have said before is that in no world will Nvidia and AMD use Intel Foundry for HVM. Their tie to TSMC is very deep and that will not change unless Jensen Huang and Lisa Su step away from the companies. That also means Intel Products will never have an equal relationship with TSMC.
So what will a complete Intel split really accomplish?
Why would a company - nVidia for example - willingly buy chips from another company that is actively seeking (albeit not very successfully so far) to sell similar chips ? Every order placed with IFS is a potential cross-subsidy for Intel products.
Nvidia might not care so much because it has an over two decade software ecosystem moat, which Lip-bu Tan understands to the extent Intel is giving up on LLM training.
On the other hand he said Intel is going to compete in inference, and to the extent that's not on a CPU, maybe with a NPU it would try to compete with Nvidia.
Another issue is Nvidia is really pushing their designs, the big ones are EUV max aperture size, the latest GB200 uses liquid cooling. How long before Intel will show an 18A or 14A node can do the former with acceptable yield in HVM, doesn't the road map use chiplets for a long time? The latter requires especially close collaboration Intel has yet to master.
Our host probably has it right. Maybe the company could be bribed into doing this even with the opportunity cost, which includes a finite number of engineers in the world who can design highest end logic chips, but for now absent greater signs the mainland will destroy TSMC's operations on the island it doesn't make sense.
What Nvidia is doing is hard enough with the best foundry in the world, see for example the initial Blackwell problems that required a respin because TSMC's simulations didn't catch enough "hot spots" for yield.
I guess I don't understand what splitting is. They are already functionally split, different financials, the same as Samsung Foundry is split from Samsung electronics. Do you mean Intel Products and Intel Foundry have no connection whatsoever? That would be like splitting Siamese twins. Even if they both survived there would still be a strong connection for years to come.
What I have said before is that in no world will Nvidia and AMD use Intel Foundry for HVM. Their tie to TSMC is very deep and that will not change unless Jensen Huang and Lisa Su step away from the companies. That also means Intel Products will never have an equal relationship with TSMC.
So what will a complete Intel split really accomplish?
It is certainly possible. Pat G seemed to be on an all-or-nothing mission which would have killed both sides of Intel.
My concern is Intel Products remaining competitive without manufacturing. Where do you think Nvidia and AMD would be without their close relationship with TSMC (manufacturing)? Do you think Intel Products will remain competitive using TSMC processes and packaging that AMD and Nvidia co-develop?
The best presentations at Hot Chips this week, in my opinion, were from Nvidia. Way above my pay grade but they were impressive. Google also did great presentations. How will AMD and Intel compete against companies with so much young talent and so much money being spent on design? Google is all-in on TSMC N3 and N2 by the way and I have seen Google write some VERY big design and IP checks. How do you compete with that?
I understand that keeping Intel together and investing in manufacturing has more risk but it also has more reward, absolutely.
It is certainly possible. Pat G seemed to be on an all-or-nothing mission which would have killed both sides of Intel.
My concern is Intel Products remaining competitive without manufacturing. Where do you think Nvidia and AMD would be without their close relationship with TSMC (manufacturing)? Do you think Intel Products will remain competitive using TSMC processes and packaging that AMD and Nvidia co-develop?
The best presentations at Hot Chips this week, in my opinion, were from Nvidia. Way above my pay grade but they were impressive. Google also did great presentations. How will AMD and Intel compete against companies with so much young talent and so much money being spent on design? Google is all-in on TSMC N3 and N2 by the way and I have seen Google write some VERY big design and IP checks. How do you compete with that?
I understand that keeping Intel together and investing in manufacturing has more risk but it also has more reward, absolutely.
If Intel Products becomes uncompetitive (not something I'm well qualified to comment on), that will be because something new and more competitive has come along and supplanted it in the marketplace. And the customers would be benefting from the improved products and/or pricing. This is surely just normal business evolution at work. Intel Products has no more right to exist than any other semiconductor company.
But I'm not seeing it that way.
Intel is sitting on a gold plated cash cow x86 business that while arguably in long term decline has years of highly profitable business ahead of it with only AMD for competition. If they can't make money at something like this (and this is regardless of who fabs the silicon), frankly they don't deserve to survive.
The problem you hint at is with future businesses and growth - how Intel competes for talent when it's unlikely it can offer the sort of stock option growth and compensation that's expected in the US. And that will be very difficult to fix while it remains what is essentially a conglomerate (to use an archaic 1970s term) of mature and emerging businesses. Note that the competitors like nVidia aren't carrying the lower growth legacy businesses that Intel has. I'd never thought of the word "conglomerate" in conjunction with Intel before - but that is what it had become - buying up endless unrelated businesses for supposed synergies, only to dispose of them at a loss later (Altera being only the latest). This lack of focus and concentration has been as much of a problem as the IDM model.
I'll return to a view I've given before - that people are hoping that Intel can become something it just cannot.
AMD will likely torpedo any x64 related IP sharing between Intel with Nvidia, AMD is the creator and the owner of the x64 IP, Intel has only a license.
AMD will likely torpedo any x64 related IP sharing between Intel with Nvidia, AMD is the creator and the owner of the x64 IP, Intel has only a license.
Intel has AVX/AV2/AVX-512/AMX and other stuff Intel also has Intel 64 AMD will be left with ISA of 2005 so it's a loss for both mainly for AMD though cause they don't have alternate to these ISA Features
Intel has AVX/AV2/AVX-512/AMX and other stuff Intel also has Intel 64 AMD will be left with ISA of 2005 so it's a loss for both mainly for AMD though cause they don't have alternate to these ISA Features
On the AMD side, AMD has already access to AVX/AV2/AVX-512, and with the size of their footprint in the server market, they do not have to adopt AMX. Regarding PTX ( the ISA under CUDA), it's a bit too late for AMD to bake it into its designs with the 2026 GPUs, like the Mi4xx, already taped out.
On the NVIDIA side, the AVX/AV2/AVX-512/AMX extensions do not have any real value without the foundational x64 ISA used by most of the SW in the entreprise/cloud world. And I do not see NVIDIA mixing the basic ARM64 ISA with non ARM ISA extensions.
On the AMD side, AMD has already access to AVX/AV2/AVX-512, and with the size of their footprint in the server market, they do not have to adopt AMX. Regarding PTX ( the ISA under CUDA), it's a bit too late for AMD to bake it into its designs with the 2026 GPUs, like the Mi4xx, already taped out.
Intel will do the same they won't allow AMD to sell AVX/AVX2/AVX-512 Chips so no modern chips for them Intel has self developed extension for AMD64 not to mention x86 license is with Intel it's a weird thing that no other company will have any use if one revokes the license
Intel will do the same they won't allow AMD to sell AVX/AVX2/AVX-512 Chips so no modern chips for them Intel has self developed extension for AMD64 not to mention x86 license is with Intel it's a weird thing that no other company will have any use if one revokes the license
There is a vicious circle involved in separating the parts of Intel, which is insoluable. The cash required to do it would be so great, that burdening Products that way would ensure Products dies, while failing to burden Products ensures Foundry is under capitalized and dies. There is a reason PE has absented themselves from the discussion--there is no deal to be made here.
TSMC is the cookie-cutter silicon Foundry, fully realized and evolved. But what if silicon is no longer the material of the future? Whatever that material may be, I can't picture the disruption coming from TSMC. It would have to come from an IDM; armed with foundational patents which would effectively prevent Foundry participation in that new industry for as long as 18 years.
My hope is the silicon technology remains a strength at Intel and Samsung, but my hope is silicon fades away soon. When that happens, TSMC will fade away too.
My hope is the silicon technology remains a strength at Intel and Samsung, but my hope is silicon fades away soon. When that happens, TSMC will fade away too.
There's also "silicon fades into a permanent commodity" from a foundry perspective. There is going to be a financial point where throwing $xxB at 5% performance gains over 2-3 years isn't worth it economically for chip designers. Once that happens, the last foundry standing will rake in a decent amount of cash, but will also be limited in how much they can raise prices lest they tempt competition or alternatives (similar to oil pricing).
I know this is somewhat 'betting against Moores Law', which has been a losing proposition for 50+ years, but there's only so many ways to cheat death.
There is a vicious circle involved in separating the parts of Intel, which is insoluable. The cash required to do it would be so great, that burdening Products that way would ensure Products dies, while failing to burden Products ensures Foundry is under capitalized and dies. There is a reason PE has absented themselves from the discussion--there is no deal to be made here.
TSMC is the cookie-cutter silicon Foundry, fully realized and evolved. But what if silicon is no longer the material of the future? Whatever that material may be, I can't picture the disruption coming from TSMC. It would have to come from an IDM; armed with foundational patents which would effectively prevent Foundry participation in that new industry for as long as 18 years.
My hope is the silicon technology remains a strength at Intel and Samsung, but my hope is silicon fades away soon. When that happens, TSMC will fade away too.
N2 GAA and then CFET are independent of the substrate base wafer used. So essentially it doesnt matter if it is silicon or anything else. TSMC is the leader in GAA and will be in CFET.
On separating Intel, this has been done before
1) create separate company. Separate finances completely
2) Foundry is found to have severe financial issues that people do not want to pick up. It cannot IPO as it is a financial disaster.... Losses, Massive cash need, Massive depreciation
3) Foundry is "sold" to a new company with requirement that the new company provide wafer capacity to the product company. The sale price is negative. The new company accepts the depreciation and losses in exchange for payment which the product company lists as "pre-payment" on future wafer purchases
4) Product/Parent company take one time write off on amount paid to eliminate the foundry.
5) Product company goes on to be very successful and profitable.
There is a vicious circle involved in separating the parts of Intel, which is insoluable. The cash required to do it would be so great, that burdening Products that way would ensure Products dies, while failing to burden Products ensures Foundry is under capitalized and dies. There is a reason PE has absented themselves from the discussion--there is no deal to be made here.
TSMC is the cookie-cutter silicon Foundry, fully realized and evolved. But what if silicon is no longer the material of the future? Whatever that material may be, I can't picture the disruption coming from TSMC. It would have to come from an IDM; armed with foundational patents which would effectively prevent Foundry participation in that new industry for as long as 18 years.
My hope is the silicon technology remains a strength at Intel and Samsung, but my hope is silicon fades away soon. When that happens, TSMC will fade away too.
There is no more reason that the disruption would come from an existing IDM than it would from a foundry. And every chance if might come from somewhere unexpected - even China.
Uneccessarily disrepectful to TSMC who clearly must be a little more than a mere "cookie cutter foundry" to have achieved what they have. When Intel is choosing TSMC in preference to their own fabs, I think that deserves a little more respect.
IDM is noisy
Before splitting up, Intel has no value and will have no choice but to go bankrupt.
Semiconductors should never have been manufactured in the United States in the first place.
And people here are forgetting that the semiconductor industry is incredibly dynamic.
The changes are drastic
Can you imagine what TSMC will be like in 10 years?
Will this huge demand for semiconductors continue? Even if demand increases, if this is just a temporary silicon cycle, it will hurt companies that have made large investments.
This is the scary thing about the semiconductor industry.
