Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel)

[DisEnchantment]

TSMC's N7 EUV is now in its second year of production and N5 is contributing to revenue for TSMC this quarter. N3 is scheduled for 2022 and I believe they have a good chance to reach that target.


N7 performance is more or less understood.


This year and next year TSMC is mainly increasing capacity to meet demands.

TSMC Radically Boosts CapEx to Expand Production Capacities, To Reach $14B For 2019

www.anandtech.com

For Samsung the nodes are basically the same from 7LPP to 4 LPE, they just add incremental scaling boosters while the bulk of the tech is the same.

Samsung is already shipping 7LPP and will ship 6LPP in H2. Hopefully they fix any issues if at all.
They have two more intermediate nodes in between before going to 3GAE, most likely 5LPE will ship next year but for 4LPE it will probably be back to back with 3GAA since 3GAA is a parallel development with 7LPP enhancements.

Samsung’s Aggressive EUV Plans: 6nm Production in H2, 5nm & 4nm On Track

www.anandtech.com

Samsung's 3GAA will go for HVM in 2022 most likely, similar timeframe to TSMC's N3.
There are major differences in how the transistor will be fabricated due to the GAA but density for sure Samsung will be behind N3.
But there might be advantages for Samsung with regards to power and performance, so it may be better suited for some applications.
But for now we don't know how much of this is true and we can only rely on the marketing material.

This year there should be a lot more available wafers due to lack of demand from Smartphone vendors and increased capacity from TSMC and Samsung.
Lots of SoCs which dont need to be top end will be fabbed with N7 or 7LPP/6LPP instead of N5, so there will be lots of wafers around.

Most of the current 7nm designs are far from the advertized density from TSMC and Samsung. There is still potential for density increase compared to currently shipping products.
N5 is going to be the leading foundry node for the next couple of years.

For a lot of fabless companies out there, the processes and capacity available are quite good.

 

[SpudLobby]

It's completely apples to oranges. We have no particular reason to believe they're even the same core, then you get into silicon results vs simulation, power numbers, N3 vs N3B vs N3E, etc. There's simply no meaningful conclusion one can draw from those scattered points. I think Intel themselves choosing N3 over Intel 3 says far more about how those nodes compare.

Precisely. To be sure, directionally if they revealed it were (imagine) an A76 core and revealed the cache configuration and workload both - that would help a great deal in narrowing down the gap between Intel 4 & SS's nodes or TSMC's. It would be a tremendous boon.

But they didn't, and while I'm subtly optimistic about Intel this is getting ridiculous from some here.

Like you - and I pointed this out as well a few posts ago though H433x0n failed to address this, I also find it *incredibly* telling Intel are still going with N3 even for their dedicated GPU's or integrated GPU's by almost every rumor. For all the talk about Intel with HPC leadership, lol. And if Intel 3/4 is truly on par with N3 - why not use that? They don't even have to wait on 18A/20A, latter of which is allegedly "manufacturing ready" this next year.

 

[SpudLobby]

No one does. And Intel has been by far the worst at execution in the past decade, but you seem to be assuming that they will deliver everything on time and at mass production volume - you are full of criticism and doubt about TSMC but take everything Intel claims as gospel.

Intel will have the same issues with logic density scaling far better than cache, it isn't like Intel is going to double density from Intel 3 to 20A/18A either.

Personally I would love to see Intel catch up / take leadership and become a major player in the foundry business. Having all our eggs in Taiwan's basket is risky - not only from the "maybe China invades" but a really big earthquake could be just as devastating. But just because it is something I want to see doesn't mean I abandon all reason and decide Intel is going to do it just because they said they would. Intel has lied to us over and over again in the past decade, I need to see some proof before I start believing them again.

Yep. Same view. I think Intel will improve and probably (hopefully too) court some clients slowly, but right now I'm leaning towards a slower trek up the mountain than many realize.

And to be frank: even in the optimistic scenario, Intel is not going to court every client and their brother overnight for flagship orders in 2026. It won't happen even in the most optimistic case wherein TSMC ships N2 (and backside power delivery would start in 2026 with N2P) in 2026 sans BSPD, it won't turn Intel's fortunes around in the Titanic way people imagine. People need to be realistic. Intel's best case is becoming a significant 3rd player at all for one - one that can actually be trusted with leading designs and orders - and the optimistic case is getting close to or taking Samsung's place as number 2. Volume, yields *at volume production*, IP and experience with clients count and this will take time.

 

[desrever]

This isn't just binning. The same chip with a higher power limit will generally be less efficient at full load. Or the same power limit and less cores. I picked out these two specs just to illustrate where you'd start if you wanted to try to do such a comparison. Throw the architectural and design differences on top, and it's just useless for comparing nodes

It's not any more useless than the other BS in the thread. So far you haven't even given any real evidence to the contrary, just said things everyone already knows about comparing things.

What other metrics are you even going to use? Some fact sheet published by Intel and TSMC but isn't verifiable in a working product? There is million of variables in every design and changing 1 changes a lot of of characteristics. If you go by that, why even bother trying to compare? Most published metrics by both companies never even line up by design. They don't want to make it easy to compare.

If you want to isolate every variable, then you will end up with no comparison because there is always differences.

It's like comparing apple and oranges but its still a valid comparison.

 

[H433x0n]

Precisely. To be sure, directionally if they revealed it were (imagine) an A76 core and revealed the cache configuration and workload both - that would help a great deal in narrowing down the gap between Intel 4 & SS's nodes or TSMC's. It would be a tremendous boon.

But they didn't, and while I'm subtly optimistic about Intel this is getting ridiculous from some here.

Like you - and I pointed this out as well a few posts ago though H433x0n failed to address this, I also find it *incredibly* telling Intel are still going with N3 even for their dedicated GPU's or integrated GPU's by almost every rumor. For all the talk about Intel with HPC leadership, lol. And if Intel 3/4 is truly on par with N3 - why not use that? They don't even have to wait on 18A/20A, latter of which is allegedly "manufacturing ready" this next year.

I don’t have a definitive answer why they’re using TSMC silicon for dGPU and iGPU. I’m assuming for both costs and contractual reasons. They probably don’t have much margin on dGPU and given that TSMC can produce wafers that are solid on both performance and costs it would make sense to go that route. The last thing the GPU program needs is any unexpected delays or added expenses. That’s just a guess though.

These aren’t my claims that Intel 3/4 is competitive with TSMC N3. I’m going off of what others have wrote that have more expertise than I do. I don’t see why they’d have motive to lie about it.

 

[Exist50]

So far you haven't even given any real evidence to the contrary, just said things everyone already knows about comparing things.

Are you asking why I'm not drawing conclusions from a comparison I just called out as being woefully inadequate to draw conclusions?

 

[H433x0n]

It's not any more useless than the other BS in the thread. So far you haven't even given any real evidence to the contrary, just said things everyone already knows about comparing things.

What other metrics are you even going to use? Some fact sheet published by Intel and TSMC but isn't verifiable in a working product? There is million of variables in every design and changing 1 changes a lot of of characteristics. If you go by that, why even bother trying to compare? Most published metrics by both companies never even line up by design. They don't want to make it easy to compare.

If you want to isolate every variable, then you will end up with no comparison because there is always differences.

It's like comparing apple and oranges but its still a valid comparison.

The technical documents TSMC / Intel have released have been independently verified over the years. For example - TechInsights has analyzed Intel 7 on 4 different occasions over the years and found that the pitches and other measurements match what was published.

 

[Geddagod]

The technical documents TSMC / Intel have released have been independently verified over the years. For example - TechInsights has analyzed Intel 7 on 4 different occasions over the years and found that the pitches and other measurements match what was published.

I would love to get a subscription but I've heard those things cost hundreds of thousands of dollars- they sell to businesses not individuals AFAIK. It's a shame.

 

[jpiniero]

I don’t have a definitive answer why they’re using TSMC silicon for dGPU and iGPU.

The 10 nm disaster. And then 7 nm as well, esp given whatever fine they had to pay because they couldn't make 7 nm work for Aurora and basically had to redo it at TSMC.

 

[H433x0n]

The 10 nm disaster. And then 7 nm as well, esp given whatever fine they had to pay because they couldn't make 7 nm work for Aurora and basically had to redo it at TSMC.

You’re right. They messed up 10nm so they’re doomed to fail forever now. Intel and Samsung might as well close up shop. Clearly only TSMC is capable of doing this now and forever; there must be something special in the water in Taiwan.

 

[A///]

I would like to clarify why Headaxon's sentiment and sarcasm is very much like mine we're not the same individual, though I do share his sentiment on the constant negativity, pessimism and nagging of all brands in this market and not specifically intel. i sometimes need to double check that I'm in fact on the ananad tech website and not a menopause forum.

 

[Geddagod]

It's not any more useless than the other BS in the thread. So far you haven't even given any real evidence to the contrary, just said things everyone already knows about comparing things.

What other metrics are you even going to use? Some fact sheet published by Intel and TSMC but isn't verifiable in a working product? There is million of variables in every design and changing 1 changes a lot of of characteristics. If you go by that, why even bother trying to compare? Most published metrics by both companies never even line up by design. They don't want to make it easy to compare.

If you want to isolate every variable, then you will end up with no comparison because there is always differences.

It's like comparing apple and oranges but its still a valid comparison.

You could do a much better job isolating the variables. I agree, you can never really have a perfect comparison, but the comparison you are using has so many different variables I agree with Exist50 when he says it's next to useless.

The comparison I personally have looked at is SNC vs WLC vs CML vs CYPC at package powers of 12-100 watts (equalized for 4c). And the reason I haven't really spoken at length about what I found between the power curves is because the data is so murky and lacking - and there are so many variables - that it's just not really a good comparison IMO.

What I think is the most fair test would be to look at an equal core TGL chip vs Zen 3 mobile chip at the same power level- and then look at frequency from there. Data on this is hard to find, so perfect comparisons again are limited. I believe this would be the best comparison we would have because cores are extremely similar IPC, both are mobile skus, both are monolithic. Both have similar amounts of L2+L3. Limitations are obviously architecture, cache setup, AMD using HD+HP while WLC I believe is HP+UHP, IO and non-Core (IO) power differences, boosting algorithms, binning, etc etc.

Looking at the 5600h vs 11400h at 45 watts, the targeted TDP, Jarrods Tech finds the 5600h to clock ~9% faster in Blender.
Cutting down the power per core, computer base testing via Blender BMW shows the 8 core 5980hs to clock ~15% higher at 35 watts than the 11800h.

I just want to mention that these are the 'worst case' scenarios I could find- as in the lowest TDP iso core count for testing I could find. Again, data here is very limited. Intel 10SF, or at least TGL, scales better at higher power than Zen 3 mobile variants, so the frequency gap would shrink or even reverse when you get to ~70 watts for the 8 core variants.

The data here seems to show that Intel 10SF is 10-15% lower perf/watt than TSMC 7nm, at lower power levels. 10-15% is the exact margin Intel claimed Intel 7 improved perf/watt over Intel 10SF. And we also have Intel 7 Ultra which increased perf/watt a bit over again.

But what about those self claimed perf/watt improvements Intel touts for their nodes? Again, direct comparisons are hard, but a decent one here is once again, TGL vs ICL. HWCooling shows us that at ~15 watts, TGL clocks ~20% faster than ICL (1165g7 vs 1065g7). Intel claimed a 18% gain with Intel 10SF. Lines up. Though I do want to stress, from what I see, at higher power levels this gap does shrink, but that's true for frequency/watt across the board.

 

[Doug S]

And to be frank: even in the optimistic scenario, Intel is not going to court every client and their brother overnight for flagship orders in 2026. It won't happen even in the most optimistic case wherein TSMC ships N2 (and backside power delivery would start in 2026 with N2P) in 2026 sans BSPD, it won't turn Intel's fortunes around in the Titanic way people imagine. People need to be realistic. Intel's best case is becoming a significant 3rd player at all for one - one that can actually be trusted with leading designs and orders - and the optimistic case is getting close to or taking Samsung's place as number 2. Volume, yields *at volume production*, IP and experience with clients count and this will take time.

Yep, in the near term the only winner if Intel executes perfectly on their plans is Intel, and the only loser is AMD. They aren't going to get any major foundry customers until they prove they can execute AND will have the capacity ready (including fab equipment) to handle them. They have a lot of stuff on the drawing board with the CHIPS funding in Ohio, plus the other countries doing similar funding on a smaller scale but it takes a minimum of four years from breaking ground to mass production so we're talking something post 18A if not post post 18A before that capacity comes online even if Intel delivers 18A when they say they will.

So best case from now until the end of 2026 is Intel wins back some x86 market share from AMD. TSMC is still going to be the #1 foundry by a long shot even if they totally screw up and Intel gets a clear lead, because Intel doesn't have customers or the capacity to serve them. And frankly, they'll need to spin off their fabs to be taken seriously by TSMC's biggest customers, no one is going to believe "assurances" that their internal x86 customer will be treated the same as external customers until it is a separate company not owned or controlled by Intel. They have an in built advantage vs TSMC for seeking business from Apple, Qualcomm, AMD and Nvidia from "made in the USA" but none of them will accept even the slightest possibility that their wafers will ever take a back seat to Intel x86 wafers.

 

[jpiniero]

You’re right. They messed up 10nm so they’re doomed to fail forever now. Intel and Samsung might as well close up shop. Clearly only TSMC is capable of doing this now and forever; there must be something special in the water in Taiwan.

I mean that's the reason. It certinately looks like Raja/etc wanted to use TSMC, and used the fab disasters to convince upper management to let them do it.

 

[SpudLobby]

Yep, in the near term the only winner if Intel executes perfectly on their plans is Intel, and the only loser is AMD. They aren't going to get any major foundry customers until they prove they can execute AND will have the capacity ready (including fab equipment) to handle them. They have a lot of stuff on the drawing board with the CHIPS funding in Ohio, plus the other countries doing similar funding on a smaller scale but it takes a minimum of four years from breaking ground to mass production so we're talking something post 18A if not post post 18A before that capacity comes online even if Intel delivers 18A when they say they will.

So best case from now until the end of 2026 is Intel wins back some x86 market share from AMD. TSMC is still going to be the #1 foundry by a long shot even if they totally screw up and Intel gets a clear lead, because Intel doesn't have customers or the capacity to serve them. And frankly, they'll need to spin off their fabs to be taken seriously by TSMC's biggest customers, no one is going to believe "assurances" that their internal x86 customer will be treated the same as external customers until it is a separate company not owned or controlled by Intel. They have an in built advantage vs TSMC for seeking business from Apple, Qualcomm, AMD and Nvidia from "made in the USA" but none of them will accept even the slightest possibility that their wafers will ever take a back seat to Intel x86 wafers.

If you mean that Intel with its internal nodes would be more inclined to preserve it's best processes for X86 designs internally - I see their current strategy with Intel 4 or 20A (non-Foundry nodes kind of) as more borne of practical concerns as they work kinks out, especially as they are just now starting up. Going forward I expect similar cadence for the same reason but as IFS is structured separately from the IDM division, it's untenable to withhold their best - and they do not and likely will not ever have the fabrication advantages (certainly not at scale) that they enjoyed at times during the 00's or early 10's (beating the others to the punch). This ensures clients have a stronger hand. The current cadence is a one year ish thing, and not a big deal, I expect clients if anything are happy to see that so they know Intel isn't selling them damaged goods.

It could be an issue, but IMO you're overemphasizing this in the Gelsinger era-Intel what with the restructuring. Intel's biggest issue (that far too many, though not you, tend to overlooks when we talk about process technology and manufacturing for the mass market) is yielding competitive nodes across the two primarily important leading markets - HPC-class chips and mobile-oriented (broad term) chips. They have to prove they can do that.

At any rate just about everything else I think we see eye to eye on, and yes, Intel getting process technology back on track will have the largest impact on I think AMD provided their design division improves significantly too. The knock-on effects for firms like Nvidia, Qualcomm, MediaTek, Amazon, etc in terms of having a reliable third client are not to be understated but still would put Intel IDM as the single largest impact of a turnaround - I don't think it will end AMD or something though.

TSMC is still going to be the #1 foundry by a long shot even if they totally screw up and Intel gets a clear lead, because Intel doesn't have customers or the capacity to serve them.

Exactly.


Big tell from a design courting perspective is really this: By EOY 2027, can IFS court a Nvidia auto, GPU, or laptop chip? Qualcomm mobile (or laptop I think is arguably more likely given lower risk and Intel focus) chip? A MediaTek laptop or TPU? A Graviton or Amazon AI chip? Something of that class, arguably two products at the leading edge with significant (nothing crazy like with Apple at TSMC though) volume. If they can grab two of that sort, and things go well, then I think they're on the right track.

 

[DrMrLordX]

You’re right. They messed up 10nm so they’re doomed to fail forever now. Intel and Samsung might as well close up shop. Clearly only TSMC is capable of doing this now and forever; there must be something special in the water in Taiwan.

Nah, not Samsung. Just Intel. Intel is already positioning themselves to spin off their fabs.

 

[H433x0n]

Yep. Same view. I think Intel will improve and probably (hopefully too) court some clients slowly, but right now I'm leaning towards a slower trek up the mountain than many realize.

And to be frank: even in the optimistic scenario, Intel is not going to court every client and their brother overnight for flagship orders in 2026. It won't happen even in the most optimistic case wherein TSMC ships N2 (and backside power delivery would start in 2026 with N2P) in 2026 sans BSPD, it won't turn Intel's fortunes around in the Titanic way people imagine. People need to be realistic. Intel's best case is becoming a significant 3rd player at all for one - one that can actually be trusted with leading designs and orders - and the optimistic case is getting close to or taking Samsung's place as number 2. Volume, yields *at volume production*, IP and experience with clients count and this will take time.

I don't disagree. I hope I didn't give the impression that IFS is a threat to TSMC market share. Intel could have a clear process leadership and they're still not going to take significant market share away from TSMC. I view it similarly to how I view AMD's slow climb in market share with Epyc. They've had a performance advantage for the past 3-5 years and it didn't mean an immediate turn around in their fortunes since Intel was so entrenched. It doesn't matter the performance of Intel's nodes, nobody is going to put all their eggs in the IFS basket for a decade. TSMC will still retain the price advantage, yield advantage and most likely have a denser library when comparing N3E HD cells against Intel 3 HD cells.

Having a lead in HPC process tech and being the #1 fab is not mutually exclusive. The only company that will feel any effects of Intel's fabs turning things around in the short term will be AMD, not TSMC.

 

[Curious_Inquirer]

https://open.substack.com/pub/explo...-who?r=reiiy&utm_medium=ios&utm_campaign=post If you guys would like to know about how much everyone is spending on TSMC Sravan Kundojjala sub stack has some great information on who’s the top 10 customers. He also has a Twitter account that has great insights.

 

[SpudLobby]

I don't disagree. I hope I didn't give the impression that IFS is a threat to TSMC market share. Intel could have a clear process leadership and they're still not going to take significant market share away from TSMC. I view it similarly to how I view AMD's slow climb in market share with Epyc. They've had a performance advantage for the past 3-5 years and it didn't mean an immediate turn around in their fortunes since Intel was so entrenched. It doesn't matter the performance of Intel's nodes, nobody is going to put all their eggs in the IFS basket for a decade. TSMC will still retain the price advantage, yield advantage and most likely have a denser library when comparing N3E HD cells against Intel 3 HD cells.

Having a lead in HPC process tech and being the #1 fab is not mutually exclusive. The only company that will feel any effects of Intel's fabs turning things around in the short term will be AMD, not TSMC.

Something we can agree on. I could actually see Intel having a cost advantage though, I don't think they'll charge what TSMC gets away with. I mean, initially they will lack the economies of scale that TSMC have, but TSMC are also at the point of borderline monopoly rents seeing as Samsung is somewhat intolerable for the likes of Qualcomm and Nvidia's premium lines. I think that will turn around, though, and I think Samsung will prove to have a decent 4NM process (which is important instead of brute forcing a 3NM GAAFET that just matches like N4P FinFET from TSMC).

At any rate, there is good reason to be optimistic, I think probably we often talk past one another given the perceptions vary so much (often for the worst, and silliest) on these topics at risk of drawing out in meta stuff. I'm very much rooting for Intel more than anyone else (Samsung second, TSMC last for selfish geopolitical reasons + monopolies or quasi-monopolies just suck).

What I'd really just like to see is Intel pull off an 18A chip for Qualcomm, MediaTek, Broadcom, Amazon. One of those feels like the most likely candidates to me. Frankly, Intel 3 is what I'd rather see, but it doesn't seem like there's much interest from those parties there.

 

[H433x0n]

Something we can agree on. I could actually see Intel having a cost advantage though, I don't think they'll charge what TSMC gets away with. I mean, initially they will lack the economies of scale that TSMC have, but TSMC are also at the point of borderline monopoly rents seeing as Samsung is somewhat intolerable for the likes of Qualcomm and Nvidia's premium lines. I think that will turn around, though, and I think Samsung will prove to have a decent 4NM process (which is important instead of brute forcing a 3NM GAAFET that just matches like N4P FinFET from TSMC).

At any rate, there is good reason to be optimistic, I think probably we often talk past one another given the perceptions vary so much (often for the worst, and silliest) on these topics at risk of drawing out in meta stuff. I'm very much rooting for Intel more than anyone else (Samsung second, TSMC last for selfish geopolitical reasons + monopolies or quasi-monopolies just suck).

What I'd really just like to see is Intel pull off an 18A chip for Qualcomm, MediaTek, Broadcom, Amazon. One of those feels like the most likely candidates to me.

From multiple reports it seems TSMC is signaling that their wafers will cost significantly more coming out of the Arizona fab. This is why I assumed they’d have a coat advantage since it’s cheaper to run a fab in Asia.

Frankly, Intel 3 is what I'd rather see, but it doesn't seem like there's much interest from those parties there.

Per the IFS investor webinar and a few people on Semiwiki there appears to be an Intel 3 customer they plan to announce in H2 2023. I can’t think of a good reason why they’d wait to announce it if it was already a done deal like they’ve been hinting at. I imagine it’s somebody dual sourcing a product on both N3 and Intel 3 and not some huge blockbuster deal.

 

[Doug S]

From multiple reports it seems TSMC is signaling that their wafers will cost significantly more coming out of the Arizona fab. This is why I assumed they’d have a coat advantage since it’s cheaper to run a fab in Asia.

I wonder about that. We can probably assume the land acquisition cost is a wash since it is such a tiny component of the overall cost to build a fab, so the main difference would be construction cost and labor cost - I'm assuming ASML scanners cost the same wherever they are delivered. I have no clue how construction cost in Taiwan compares to the US, but I could easily believe it is more expensive here. People here are more expensive, but the labor cost component of fabs gets smaller the more expensive fab equipment gets.

One big fact that's ignored in these "see look how much more expensive it is to make chips in the US" is the scale of the fabs. TSMC's Arizona fabs are nowhere near the size of the fabs they built in Taiwan. You get economies of scale when you go bigger, but when TSMC is building a trailing edge fab (N+1 at best since they said they will always keep their leading edge stuff in country) they 1) can't build it as big and 2) don't benefit from operating it as a leading edge fab with customers like Apple backing up dumptrucks full of cash to pay for a half million leading edge wafers a year. That really helps amortize your investment, so it makes sense you have to charge more in the US where you aren't getting all that juicy amortization.

There are other things that aren't accounted for, like the business risk of getting your wafers from Taiwan when a big earthquake could take it out for months with zero warning. Not a lot of earthquakes in Arizona. It wouldn't be with zero warning, but there's also that small chance that China decides to invade. Again, no worries that California or New Mexico is going to invade Arizona and cause TSMC's fabs to shut down. Is that worth paying 35% more? I can't answer that but it is clearly worth a lot more than paying 0% more.

 

[Ajay]

I wonder about that. We can probably assume the land acquisition cost is a wash since it is such a tiny component of the overall cost to build a fab, so the main difference would be construction cost and labor cost - I'm assuming ASML scanners cost the same wherever they are delivered. I have no clue how construction cost in Taiwan compares to the US, but I could easily believe it is more expensive here. People here are more expensive, but the labor cost component of fabs gets smaller the more expensive fab equipment gets.

One big fact that's ignored in these "see look how much more expensive it is to make chips in the US" is the scale of the fabs. TSMC's Arizona fabs are nowhere near the size of the fabs they built in Taiwan. You get economies of scale when you go bigger, but when TSMC is building a trailing edge fab (N+1 at best since they said they will always keep their leading edge stuff in country) they 1) can't build it as big and 2) don't benefit from operating it as a leading edge fab with customers like Apple backing up dumptrucks full of cash to pay for a half million leading edge wafers a year. That really helps amortize your investment, so it makes sense you have to charge more in the US where you aren't getting all that juicy amortization.

There are other things that aren't accounted for, like the business risk of getting your wafers from Taiwan when a big earthquake could take it out for months with zero warning. Not a lot of earthquakes in Arizona. It wouldn't be with zero warning, but there's also that small chance that China decides to invade. Again, no worries that California or New Mexico is going to invade Arizona and cause TSMC's fabs to shut down. Is that worth paying 35% more? I can't answer that but it is clearly worth a lot more than paying 0% more.

Excellent post!

I did chafe at M. Chang's comment the costs of US fabs, when, obviously, he made no mention of the economy of scale. US labor laws, union pay and environmental impact and remediation requirements almost certainly increase construction costs - but the amortization of those costs over time should reduce the impact substantially. The solution, should Intel succeed, is to build bigger fabs and continue to find ways to get central and local governments to co-invest in expansion plans. Maybe this isn’t the best use of public monies, that’s a wide ranging debate, but it is good business sense.

 

[DrMrLordX]

Maybe this isn’t the best use of public monies, that’s a wide ranging debate, but it is good business sense.

Yes, pawning failures off on the taxpayer makes excellent business sense. For the business.

 

[scannall]

Yes, pawning failures off on the taxpayer makes excellent business sense. For the business.

Just more welfare for the rich. So everyday people have to pay for Intel's stock buybacks enriching stockholders. Oh, and pay for 15 years+ of bad decisions..

 

[Ajay]

Just more welfare for the rich. So everyday people have to pay for Intel's stock buybacks enriching stockholders. Oh, and pay for 15 years+ of bad decisions..

Well, stock buybacks should be illegal, IMHO.

 

[DrMrLordX]

Well, stock buybacks should be illegal, IMHO.

You can't really make them illegal. Congress just has to be smart enough not to subsidize companies that are burning capital buying up their own stock prior to begging for cash.