The viability of fast following

[nghanayem]

While I am not sure it was intentional, UMC often would be a fast follower to TSMC, and before 28nm this worked out well for them as designs could be easily ported from TSMC nodes to UMC nodes. Once TSMC started accelerating away and UMC's nodes stopped being design compatible this model fell apart, and UMC needed a more differentiated offering. Samsung and Intel are doubling down in an attempt to accelerate ahead of TSMC. The thought then crossed my mind if fast following could work for these IDMs. TSMC used fast following to great effect to accelerate past the other semiconductor companies with HKMG last, finFETs on bulk Si, and EUV. In the case of intel they are now benefiting from the work that TSMC and Samsung put in at 7 and 5nm to accelerate their own EUV efforts. With this I wonder if it would potentially be desirable to remain a fast follower rather than attempting to zoom ahead of the current technological leader?

To clarify I don't think these firms have any plans of doing this, given that Samsung has the ambitious goal to be the number one foundry and intel has the much more modest goal of being No.2. But I wonder if these firms chose a different path, or if their current path leads to a financial situations that makes "being in first place" becomes too expensive to justify. In theory it should be both easier to catch up to whomever is the current technological lead than to beat them due to all of the whitepapers, teardowns, and vendor experience acting as a strong tailwind. While this stopped working for UMC because their nodes stopped being plug and play with TSMC - giving TSMC an unassailable first movers advantage - but in theory the IDMs are not bound in a similar manner because their primary business is selling end product not wafers. In theory these IDMs don't have to have the best cost structure because they are selling the wafer at ASP*number working dies rather than wafer cost*1.X. Their captive customers also either can't leave or face significant capacity or cost penalties to leave (reducing the risk of a UMC situation). One additional benefit of this scheme could be that when new nodes are rolled out they could be in an even healthier state than a normal leading edge node would be when it is new, given the extra time it had to incubate (think things like UMC 14nm or GF 22/12 FDX which had to already be mature when they entered HVM).

In short; could staying at N+1 with TSMC be something that is sustained for a long period of time (with maybe a few outsourcing jobs or parts of an MCM design being thrown to TSMC here and there)? Or even with things like MCM and the general slowing of transistor improvements would the IDMs start bleeding share to competitors if they are always a bit behind on the process front (potentially risking that these IDMs even lose the scale needed to even "run the race" at a slower pace)? While it would certainly not be a way to build up a strong healthy foundry, would this business model allow the IDMs to "stay in the race" without having to burn mountains of cash and developing a foundry to help amortize their costs?

 

[hist78]

TSMC is not just a fast follower. They built several critical capabilities (such as immersion technology and EUV) earlier than many peers.

Without TSMC's commitment and collaborations, ASML may have difficulties to achieve what they are today. On the other hand, I believe some technology solutions achieved by the ASML/TSMC partnership might be exclusive to TSMC. A TSMC's competitor might not get everything ASML knew even they bought EUV machines from ASML.

 

[nghanayem]

TSMC is not just a fast follower. They built several critical capabilities (such as immersion technology and EUV) earlier than many peers.

Without TSMC's commitment and collaborations, ASML may have difficulties to achieve what they are today. On the other hand, I believe some technology solutions achieved by the ASML/TSMC partnership might be exclusive to TSMC. A TSMC's competitor might not get everything ASML knew even they bought EUV machines from ASML.

A few notes. I never meant it as they never come up with new things. Moreso in the 2000s and 2010s they got to benefit from the work of intel to develop a better process. One example is going HKMG last when the rest of the industry wanted gate first. Another was that N16 originally was a SOI process to control fin height until 22nm came around and TSMC moved to a bulk process. The final example was with EUV. Intel and IBM were the two biggest investors in the technology for most of it’s looooooong development (not surprising given those two were always the ones who pushed new lithography techniques forward). Near the finish line IBM left HVM and intel seems to have lost some of their interest once they thought that EUV would miss 10nm. Samsung then ironed out the worst of the final HVM issues, allowing TSMC to adopt limited amounts of EUV for N7+ with minimal resistance. In my opinion a more fair statement about TSMC and EUV is that they were always a key second place player in its development, but once it was ready TSMC put their money where their mouth was and bought like mad (giving ASML extra scale).

As for exclusive collaboration. The public moves and motivations would not support this. It is in ASML’s best interest to provide all of their learnings to all of their customers to create further demand for their systems. They are aided in this endeavor by doing most of the work on the EUV tools themselves. Additionally if there were secret collaborations then I don’t think that TSMC would need to be less transparent about their inner workings than they used to be. As a side note it kind of reminds of a slightly less closed off (kind of a necessity given their line of work) old intel. Not that I’m upset about it though, when you are in the lead you want to minimize the information available to your competitors.

 

[Daniel Nenni]

The relationship between TSMC and Applied Materials did result in some unique implementations. ASML, LCRX, KLAC, etc... I do not know first hand but it definitely does happen. I think it is highly unlikely that semiconductor equipment is plug and play between Intel, TSMC, and Samsung logic fabs. Let me ask around for more input. I know some people.......

 

[nghanayem]

The relationship between TSMC and Applied Materials did result in some unique implementations. ASML, LCRX, KLAC, etc... I do not know first hand but it definitely does happen. I think it is highly unlikely that semiconductor equipment is plug and play between Intel, TSMC, and Samsung logic fabs. Let me ask around for more input. I know some people.......

My understanding is that while it is not plug and play: especially if the fab is doing something unusual with the tool. However there are obviously issues that everyone has to deal with (PMOS mobility, EUV throughput, maintenance and process best practices to maximize PM lifetime, control over fin heights and profiles, copper running out of room to shrink, HNS uniformity, ect ect). For these situations the tool vendors can and do leverage their expertise on how they solved these issues in the past to help the fab in question figure out their particular solution for their process. Tool vendors also make upgrade packages to improve their customer's tools. While I have no doubt that many are only for X customer who is working on Y feature for their Z node; I would not be shocked if you told me that there are upgrade packages that are distributed to all of their customers whenever it would help.

TSMC and Applied Materials did result in some unique implementations

Interesting. I doubt you can speak about it though... Shame I would love to have insight into what sort of exclusive solutions are made, and how long (if ever) it takes for these to eventually be distributed to their other customers. Obviously I only have insight into my fab and the few vendor engineers I occasionally talk with. So I am far from the absolute authority on everything the tool vendors would know.

 

[Mooredaddy]

My understanding is that while it is not plug and play: especially if the fab is doing something unusual with the tool. However there are obviously issues that everyone has to deal with (PMOS mobility, EUV throughput, maintenance and process best practices to maximize PM lifetime, control over fin heights and profiles, copper running out of room to shrink, HNS uniformity, ect ect). For these situations the tool vendors can and do leverage their expertise on how they solved these issues in the past to help the fab in question figure out their particular solution for their process. Tool vendors also make upgrade packages to improve their customer's tools. While I have no doubt that many are only for X customer who is working on Y feature for their Z node; I would not be shocked if you told me that there are upgrade packages that are distributed to all of their customers whenever it would help.


Interesting. I doubt you can speak about it though... Shame I would love to have insight into what sort of exclusive solutions are made, and how long (if ever) it takes for these to eventually be distributed to their other customers. Obviously I only have insight into my fab and the few vendor engineers I occasionally talk with. So I am far from the absolute authority on everything the tool vendors would know.

If anything I believe Intels financial performance will hold them back from competing at the leading edge. Intels issues are primarily structural and financial rather then technical at this point imo. If they come in under already very low guidance for this quarter and then give continuing weak guidance for fiscal 2023. I’d start to wonder if they close the gap even if they regain relative process leadership. Thoughts from others welcome

 

[Daniel Nenni]

If anything I believe Intels financial performance will hold them back from competing at the leading edge. Intels issues are primarily structural and financial rather then technical at this point imo. If they come in under already very low guidance for this quarter and then give continuing weak guidance for fiscal 2023. I’d start to wonder if they close the gap even if they regain relative process leadership. Thoughts from others welcome

Personally I think expectations were set too high after Pat Gelsinger joined Intel His "AMD is in the rear view mirror" comment said it all:

"All of a sudden, boom, we are back in the game. AMD in the rear view mirror in clients, and never again will they be in the windshield," Gelsinger says with a smile. “We are just leading the market." If you’re an Intel fan, you’ll love to hear the confidence. But if you’re team AMD, you’re likely rolling your eyes.

The same can be said about IFS. The foundry business is a marathon not a sprint. It really is a difficult business and not to be taken lightly. Under estimating TSMC is also a huge mistake as they are champion counter punchers. If you hit them once they will hit you back ten times.

 

[Mooredaddy]

Personally I think expectations were set too high after Pat Gelsinger joined Intel His "AMD is in the rear view mirror" comment said it all:

"All of a sudden, boom, we are back in the game. AMD in the rear view mirror in clients, and never again will they be in the windshield," Gelsinger says with a smile. “We are just leading the market." If you’re an Intel fan, you’ll love to hear the confidence. But if you’re team AMD, you’re likely rolling your eyes.

The same can be said about IFS. The foundry business is a marathon not a sprint. It really is a difficult business and not to be taken lightly. Under estimating TSMC is also a huge mistake as they are champion counter punchers. If you hit them once they will hit you back ten times.

The prudent think to do would be to cut the dividend, but I reckon that would trigger a shareholder revolt. This is the exact scenario where a company SHOULD cut its dividend to ensure its ability to peruse strategic goals. I think Pat and the Intel c suite have done a very poor job of communicating to investors just how serious a task standing up to TSMC in its own backyard will be. A lot of the Intel investors I talk to are under the that impression once Intel gets even a marginal process lead then it will be off to races in acquiring major business from the likes of nvidia, Qualcomm etc. Some are even floating the idea of APPLE coming back for gods sake. Expectations are completely out of whack.

 

[Daniel Nenni]

The prudent think to do would be to cut the dividend, but I reckon that trigger a shareholder revolt. This is the exact scenario where a company SHOULD cut its dividend to ensure its ability to peruse strategic goals. I think Pat and the Intel c suite have done a very poor job or communicating to investors just how serious a task standing up to TSMC in its own backyard will be. A lot of the Intel investors I talk to are under the impression once Intel gets even a marginal process lead then it will be off to races in acquiring major business from the likes of nvidia, Qualcomm etc. Some are even floating the idea of APPLE coming back for gods sake. Expectations are completely out of whack.

True, I do Q&As with Wall Street quite a bit. It's not just what they ask, it's what they don't ask. Yes they do think 18A will be a landslide of foundry business. Pat even said they are in conversation with 8 of the top ten semiconductor companies. From conversation to revenue is such a long time I wouldn't even mention it. One thing you have to remember is that the last time Intel threw into the foundry business they did have a process advantage (22nm/14nm).

I'm hopeful that IFS will succeed because we need a multi source leading edge foundry business. Unfortunately, with expectations so high, I don't think that will be possible. And it is not just investors, it's employees, partners, and customers that will be judging if expectations have been met or not and that will ultimately be Pat's legacy at Intel.

 

[nghanayem]

True, I do Q&As with Wall Street quite a bit. It's not just what they ask, it's what they don't ask. Yes they do think 18A will be a landslide of foundry business. Pat even said they are in conversation with 8 of the top ten semiconductor companies. From conversation to revenue is such a long time I wouldn't even mention it. One thing you have to remember is that the last time Intel threw into the foundry business they did have a process advantage (22nm/14nm).

I'm hopeful that IFS will succeed because we need a multi source leading edge foundry business. Unfortunately, with expectations so high, I don't think that will be possible. And it is not just investors, it's employees, partners, and customers that will be judging if expectations have been met or not and that will ultimately be Pat's legacy at Intel.

Does hype matter to engineers? If the IFS customers are well taken care of, have passible PDKs/IP support, competitive nodes, reasonable wafer allocation, IFS followed through on it's promises, and they see a path for further improvement why would they not stick with it? I don't say this to be mean, but frankly why should an engineer care if Wall Street or armchair CEOs are disappointed that IFS didn't overtake TSMC in market share with their first node. I'd imagine all they care about is if IFS was a boon or an ankle weight, and if they think that through continued collaboration they can enjoy more profitable ventures.

 

[nghanayem]

The prudent think to do would be to cut the dividend, but I reckon that trigger a shareholder revolt. This is the exact scenario where a company SHOULD cut its dividend to ensure its ability to peruse strategic goals. I think Pat and the Intel c suite have done a very poor job or communicating to investors just how serious a task standing up to TSMC in its own backyard will be. A lot of the Intel investors I talk to are under the that impression once Intel gets even a marginal process lead then it will be off to races in acquiring major business from the likes of nvidia, Qualcomm etc. Some are even floating the idea of APPLE coming back for gods sake. Expectations are completely out of whack.

I totally agree with you on this. Hence my little thought experiment. Obviously not starting IFS and contenting yourself with being a fast follower who integrates the learning of the front runners might offer a cheap enough/smooth enough process development pathway for the economics of new node developments to be platiable without a foundry to amortize costs, and or a memory business/giant conglomerate to bankroll you. 5 "nodes" in 4 years or the desire to be the number one foundry are expensive ventures that will require risks (financial and technical) to be taken in an effort to slingshot TSMC. 5 "nodes" in say 6 or 8 years would allow one to profit from increased technical maturity, better tools, and a lower yearly development cost. The question is that if these savings/more thoroughly cooked nodes would be enough to stave off a bleeding of market share? Or would enough outsourcing to not lose market share be required that any savings an IDM made on node development would be wiped out by said outsourcing?

 

[Mooredaddy]

True, I do Q&As with Wall Street quite a bit. It's not just what they ask, it's what they don't ask. Yes they do think 18A will be a landslide of foundry business. Pat even said they are in conversation with 8 of the top ten semiconductor companies. From conversation to revenue is such a long time I wouldn't even mention it. One thing you have to remember is that the last time Intel threw into the foundry business they did have a process advantage (22nm/14nm).

I'm hopeful that IFS will succeed because we need a multi source leading edge foundry business. Unfortunately, with expectations so high, I don't think that will be possible. And it is not just investors, it's employees, partners, and customers that will be judging if expectations have been met or not and that will ultimately be Pat's legacy at Intel.

Why is everyone expecting 18A to be the saviour of Intel? Are people really expecting TSMC not to have a process that’s either better or close enough that dealing with all of Intels BS isn’t worth it. It just seems to me that Intels strategy here in contingent on everything going right for them while simultaneously TSMC stumbles multiple times. Not impossible but hardly anything I’d bet my reputation as an analyst on. Then again 2022 showed us how valuable analysts are aren’t they? Did anyone here carvana price target of $100? LOL

 

[Mooredaddy]

I totally agree with you on this. Hence my little thought experiment. Obviously not starting IFS and contenting yourself with being a fast follower who integrates the learning of the front runners might offer a cheap enough/smooth enough process development pathway for the economics of new node developments to be platiable without a foundry to amortize costs, and or a memory business/giant conglomerate to bankroll you. 5 "nodes" in 4 years or the desire to be the number one foundry are expensive ventures that will require risks (financial and technical) to be taken in an effort to slingshot TSMC. 5 "nodes" in say 6 or 8 years would allow one to profit from increased technical maturity, better tools, and a lower yearly development cost. The question is that if these savings/more thoroughly cooked nodes would be enough to stave off a bleeding of market share? Or would enough outsourcing to not lose market share be required that any savings an IDM made on node development would be wiped out by said outsourcing?

This the crux of my argument exactly. TSMC and Samsung are already at such a critical mass of resources that I think it’s a mistake to chase leading edge processes for Intel at this point. Make the best processors you can and if TSMC is the process that works best. Use it!! I know Pats pride won’t allow it but when IFS fails (my opinion) this will be where Intel inevitably end up. Both Samsung and Intel by nature having to compete with their customers will always make their relationships not as good as TSMC’s. No matter how good your internal controls are you can’t replicate the trust that comes from a total lack of financial incentive to f over your customer. Trust is Intels biggest problem to solve.

 

[nghanayem]

Why is everyone expecting 18A to be the saviour of Intel? Are people really expecting TSMC not to have a process that’s either better or close enough that dealing with all of Intels BS isn’t worth it. It just seems to me that Intels strategy here in contingent on everything going right for them while simultaneously TSMC stumbles multiple times. Not impossible but hardly anything I’d bet my reputation as an analyst on. Then again 2022 showed us how valuable analysts are aren’t they? Did anyone here carvana price target of $100? LOL

Not really. Looking at the roadmaps and current public statements i4 is comparable on density and performance to N3, when the full node rolls out it will be competing with N3E (less density and like 5% more performance). Compare this to i3 being 18% and a density increase (it also likely has a cheaper cost structure then N3E due to the smaller number of EUV layers). Even if i3 doesn't live up to expectations it should still have better PPW and maybe tied on density. 20A should start ramping in 2024 allowing for smooth transition to 18A in late 2024 for products in 2025. N2 is trending at like 10% improved density and like 10% more performance. TSMC's own words have risk starts in late 2024 with HVM in late 2025 (3-4 quarters later than 20/18A). N2 is a smaller PPW jump than the 15% for 20A and the like 5-10% for 18A (don't remember what the roadmap says fo 18A so don't quote me on this). N2 might also be a smaller density jump than 20A. At least on the performance side TSMC should have N3X (given N4X's jump, maybe a 10-15% bump over N3E) to stand by 18A. But even then that still might lose on performance and it probably will on density. Will 18A save intel no. Design side and to a lesser degree IFS need to save intel, 18A is just a very good asset to have if you were trying to save intel.

In my opinion if intel wants to have a PPAC advantage over TSMC they can do it, and TSMC doesn't even need to slip up anymore for it to happen. Of course this is all with the caveat that their roadmap is met. Even if they do "fail" (as long as we aren't talking about 10nm and 7nm level performance or cadence slipups) the roadmap could almost be considered a "success". If 20/18A comes out a quarter or two late and are both 5-10% weaker than where the roadmaps would place them, intel would have come back to rough parity with TSMC after being so far behind that many (including myself) thought that they would never leave TSMC's and maybe even Samsung's rearview mirror. In my opinion this in of itself would be a Herculean achievement. If intel can manage to do that or even pull ahead of TSMC, then that is an AMD with Bulldozer to Zen level of comeback story.

 

[nghanayem]

This the crux of my argument exactly. TSMC and Samsung are already at such a critical mass of resources that I think it’s a mistake to chase leading edge processes for Intel at this point. Make the best processors you can and if TSMC is the process that works best. Use it!! I know Pats pride won’t allow it but when IFS fails (my opinion) this will be where Intel inevitably end up. Both Samsung and Intel by nature having to compete with their customers will always make their relationships not as good as TSMC’s. No matter how good your internal controls are you can’t replicate the trust that comes from a total lack of financial incentive to f over your customer. Trust is Intels biggest problem to solve.

Intel is already outsourcing the best node for the job at the time and capacity that intel needs. i4 is a better choice for MTL than N3 given the limited capacity (who do you think TSMC is going to prioritize Apple or intel?), high structural cost, and roughly equivalent performance to i4. Besides MTL SOC, look at GPUs, ARL, ect.

I also wouldn't exactly exclude Samsung from this argument since their logic and design business are currently on fire in a bad way (and SFS wasn't exactly a cash cow when they were doing well either).

 

[blueone]

This the crux of my argument exactly. TSMC and Samsung are already at such a critical mass of resources that I think it’s a mistake to chase leading edge processes for Intel at this point. Make the best processors you can and if TSMC is the process that works best. Use it!! I know Pats pride won’t allow it but when IFS fails (my opinion) this will be where Intel inevitably end up. Both Samsung and Intel by nature having to compete with their customers will always make their relationships not as good as TSMC’s. No matter how good your internal controls are you can’t replicate the trust that comes from a total lack of financial incentive to f over your customer. Trust is Intels biggest problem to solve.

Intel has to chase leading edge nodes for its own business. If they don't, and AMD gets a better process from TSMC (or Samsung, for that matter), then Intel is going to have a screwed-up cost structure from the fabs without industry-leading margins on their processors to absorb the non-diversified costs. You know, like now. From everything I can see, IFS really is a bet-the-company proposition. Gelsinger has done this before, when he pushed a CISC processor design over Intel's own RISC designs. (i860 and i960)

I also think the trust factor is being exaggerated. Does anyone really believe that Intel would risk the IFS investment on letting the design team examine a competitor's mask set or whatever? Or their 3D chiplet strategy? Seriously? For one thing, if IFS is a success then Intel's future is that the Intel design groups may no longer be the manufacturing group's biggest customer segment. And if it did happen even once, that would be the end of Gelsinger as CEO. And he's wanted that job for decades.

IMO, Intel's highest priority problems to solve are execution, ecosystem development, and developing better relationships with suppliers.

 

[Mooredaddy]

Intel is already outsourcing the best node for the job at the time and capacity that intel needs. i4 is a better choice for MTL than N3 given the limited capacity (who do you think TSMC is going to prioritize Apple or intel?), high structural cost, and roughly equivalent performance to i4. Besides MTL look at GPUs, ARL, ect.

I also wouldn't exactly exclude Samsung from this argument since their logic and design business are currently on fire (and SFS wasn't exactly a cash cow when they were doing well either).

Intel is already outsourcing the best node for the job at the time and capacity that intel needs. i4 is a better choice for MTL than N3 given the limited capacity (who do you think TSMC is going to prioritize Apple or intel?), high structural cost, and roughly equivalent performance to i4. Besides MTL look at GPUs, ARL, ect.

I also wouldn't exactly exclude Samsung from this argument since their logic and design business are currently on fire (and SFS wasn't exactly a cash cow when they were doing well either).

On fire as in bad right? Could read that one of two ways. Everything I’ve heard out of Samsung foundry recently is that it’s a complete sh*t show.

 

[nghanayem]

if IFS is a success then Intel's future is that the Intel design groups may no longer the manufacturing group's biggest customer segment. And if it did happen even once, that would be the end of Gelsinger as CEO. And he's wanted that job for decades.

Why would Pat lose his job if his brainchild became the most successful part of the business? If anything wouldn't that put him on a Grove/Moore level of bailing on memory for CPUs? I assume I am misunderstanding your statement.

 

[benb]

My opinion, the leading edge is the only edge you can be on with $17-20B/fab investments, but let me explain.

AMZN is, or was, the leader in inventory turns, which created their famous “flywheel”—sell more stuff more quickly, move more volume, lower the price, rinse and repeat. Until someone sells even more stuff more quickly at an ever lower price, then you become Kmart or Sears. Languish for a few decades then disappear.

This is even more true for a fab than it is for AMZN! Fabs are a cost/chip business, and whoever has the lowest cost/chip, wins all the volume. TSMC has been doing this convincingly for a decade, building more and more fabs, lowering their costs, acquiring more customers, lowering their costs more. Rinse and repeat.

Intel is, effectively, starved for volume, despite full fabs today. They can’t operate the flywheel of growing volume to lower costs to grow volume. The only way to increase volume is to sell at a lower price than TSMC, and that’s a game for all the marbles, only 1 market leader gets the majority of the volume. Shrinks reduce cost by increasing chips/wafer, so, that’s why you must stay on the leading edge. Intel needs IFS customers to grow volume, to lower costs.

What about the other force, differentiation? For all the commotion about nodes and technology, the differentiation is purchased from US, Japan and Netherlands equipment vendors, for a price. Most fabs are highly, highly standardized, so much so that engineers can move from Intel to Samsung to TSMC without missing a beat. Innovation was outsourced to vendors, leaving cost as the only factor of relevance, and that’s how we got to this state of affairs. A commoditized market.

 

[Mooredaddy]

Intel has to chase leading edge nodes for its own business. If they don't, and AMD gets a better process from TSMC (or Samsung, for that matter), then Intel is going to have a screwed-up cost structure from the fabs without industry-leading margins on their processors to absorb the non-diversified costs. You know, like now. From everything I can see, IFS really is a bet-the-company proposition. Gelsinger has done this before, when he pushed a CISC processor design over Intel's own RISC designs. (i860 and i960)

I also think the trust factor is being exaggerated. Does anyone really believe that Intel would risk the IFS investment on letting the design team examine a competitor's mask set or whatever? Or their 3D chiplet strategy? Seriously? For one thing, if IFS is a success then Intel's future is that the Intel design groups may no longer the manufacturing group's biggest customer segment. And if it did happen even once, that would be the end of Gelsinger as CEO. And he's wanted that job for decades.

IMO, Intel's highest priority problems to solve are execution, ecosystem development, and developing better relationships with suppliers.

Fair enough. Perhaps I overstated the trust factor.

Not really. Looking at the roadmaps and current public statements i4 is comparable on density and performance to N3, when the full node rolls out it will be competing with N3E (less density and like 5% more performance). Compare this to i3 being 18% and a density increase (it also likely has a cheaper cost structure then N3E due to the smaller number of metal layers). Even if i3 doesn't live up to expectations it should still have better PPW and maybe tied on density. 20A should start ramping in 2024 allowing for smooth transition to 18A in late 2024 for products in 2025. N2 is trending at like 10% improved density and like 10% more performance. TSMC's own words have risk starts in late 2024 with HVM in late 2025 (3-4 quarters later than 20/18A). N2 is a smaller PPW jump than the 15% for 20A and the like 5-10% for 18A (don't remember what the roadmap says so don't quote me on this). It might even be a smaller density jump than 20A. At least on the performance side TSMC should have N3X (given N4X's jump, maybe a 15% bump over N3) to stand by 18A. But even then that still might lose on performance and it probably will on density. Will 18A save intel no. Design side and to a lesser degree IFS need to save intel, 18A is just a very good asset to have if you were trying to save intel.

In my opinion if intel wants to have a PPAC advantage over TSMC they can do it, and TSMC doesn't even need to slip up anymore for it to happen. Of course this is all with the caveat that their roadmap is met. Even if they do "fail" (as long as we aren't talking about 10nm and 7nm level performance or cadence slipups) the roadmap could almost be considered a "success". If 20/18A comes out a quarter or two late and are both 5-10% weaker than where the roadmaps would place them, intel would have come back to rough parity with TSMC after being so far behind that many (including myself) thought that they would never leave TSMC's and maybe even Samsung's rearview mirror. In my opinion this in of itself would be a Herculean achievement. If intel can manage to do that or even pull ahead of TSMC, then that is an AMD with Bulldozer to Zen level of comeback story.

N2 GAA is a yield learning node from what I understand so I don’t expect it to be used very much and not be around for long. We can expect the derivations of N2 I.E N2P, N2X to be appreciably more performant. N2 being conservative in its density improvements is consistent with TSMC’s regimented approach to nodes. Once they get GAA down I imagine what ever follows the base N2 will be quite competitive to 18A. Have you guys heard anything to the contrary?