Discussion Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel) - [2020 - 2025]

[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.

 

[Geddagod]

TSMC delays U.S. chip plant start to 2025 due to labor shortages

Taiwan chipmaker logs quarterly net profit drop on consumer electronics slump

asia.nikkei.com

TSMC is delaying the start date of the Arizona fab to 2025, claiming "labor shortages". I don't buy it.

It's a 5nm fab IIRC. An extremely mature node. I wouldn't be surprised if TSMC just didn't want to build up more 5nm, or there really were labor shortages. A couple months ago, TSMC were claiming they were shipping some engineers from Taiwan to the US to train them. It could also be 'labor shortages' of 'not enough labor at these salaries we want to pay them' considering that a couple months ago as well, TSMC was complaining about US engineers.

 

[jpiniero]

if TSMC just didn't want to build up more 5nm,

That's probably it.

 

[Doug S]

It's a 5nm fab IIRC. An extremely mature node. I wouldn't be surprised if TSMC just didn't want to build up more 5nm, or there really were labor shortages. A couple months ago, TSMC were claiming they were shipping some engineers from Taiwan to the US to train them. It could also be 'labor shortages' of 'not enough labor at these salaries we want to pay them' considering that a couple months ago as well, TSMC was complaining about US engineers.

I agree. I think they might delay it with the excuse, and then use the delay as an excuse to make it an N3E fab opening in late 2025 once they have N2 up and running in Taiwan (to maintain their policy of always reserving the best node for home)

 

[Ajay]

I agree. I think they might delay it with the excuse, and then use the delay as an excuse to make it an N3E fab opening in late 2025 once they have N2 up and running in Taiwan (to maintain their policy of always reserving the best node for home)

I hope you are right. We don't need another FOXCONN debacle. We also need to be able to source parts from TSMC for military purposes on home soil if need be. If Intel really is able to make a go of IFS over the next 5 years - then there will be more options.

 

[H433x0n]

TSMC Q2 Earnings

TSMC produced their earnings for Q2, notably there is no N3 revenue on the books yet.

 

[Doug S]

TSMC Q2 Earnings

TSMC produced their earnings for Q2, notably there is no N3 revenue on the books yet.

Where does that say there is no N3 revenue yet? All they said was that Q3 would be supported by a "strong ramp of [N3]", that doesn't mean Q2 was zero. Unless they specifically said in the call there was zero revenue from N3 (and if so why not link to a call transcript) I see no reason to believe this.

Even if they showed a breakdown of a few nodes like they sometimes do if N3 was a fraction of a percent (i.e. just bitcoin miners or whatever so far) they wouldn't break it out they'd do something like list 5, 7, 10, 16, 28, 40 and "other".

 

[jpiniero]

Where does that say there is no N3 revenue yet?

https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2023-07/aa7d1adbe692cf07f10aa1584c832facb43ee84b/2Q23Presentation%28E%29.pdf

You can see the chart in there. Must not be shipping for revenue yet. That pretty much confirms that the 2023 iPhone isn't using N3, but perhaps the next Macs could be using it since they don't need as much of a lead time.

 

[NostaSeronx]

Versus July 20th, 2023 earnings call:
"N3 is already involved in production with good yield. We are seeing robust demand for N3 and we expect a strong ramp of N3 in the second half of this year, supported by both HPC and smartphone applications.

N3 is expected to continue to contribute mid-single-digit percentage of our total wafer revenue in 2023."

 

[H433x0n]

Versus July 20th, 2023 earnings call:
"N3 is already involved in production with good yield. We are seeing robust demand for N3 and we expect a strong ramp of N3 in the second half of this year, supported by both HPC and smartphone applications.

N3 is expected to continue to contribute mid-single-digit percentage of our total wafer revenue in 2023."

iPhone was expected to release in September on N3B. I don’t see how that’s possible and have negligible revenue from N3 in Q2.

So it’s either 1 of 3 things:

Apple deferred payment on N3 wafers somehow, iPhone 15 isn’t coming out until later Q4 or iPhone 15 is not using N3.

 

[Doug S]

iPhone was expected to release in September on N3B. I don’t see how that’s possible and have negligible revenue from N3 in Q2.

So it’s either 1 of 3 things:

Apple deferred payment on N3 wafers somehow, iPhone 15 isn’t coming out until later Q4 or iPhone 15 is not using N3.

It has already been made public that Apple isn't paying per wafer for N3, but for known good dies. So TSMC can't recognize revenue when wafers roll off the line, but much later in the process.

Apple isn't starting mass production of iPhone 15 in June. Even if they did TSMC isn't going to add an entry into that pie graph for some tiny amount like 0.17% if they shipped Apple 100K chips in June.

 

[jpiniero]

Apple isn't starting mass production of iPhone 15 in June.

If they want to sell it in September, even starting in June would be cutting it close. Apple really does need 2-3 months, then a month to ship it around the world.

 

[Thunder 57]

It has already been made public that Apple isn't paying per wafer for N3, but for known good dies. So TSMC can't recognize revenue when wafers roll off the line, but much later in the process.

Apple isn't starting mass production of iPhone 15 in June. Even if they did TSMC isn't going to add an entry into that pie graph for some tiny amount like 0.17% if they shipped Apple 100K chips in June.

That's a sweetheart deal. What is Apple going to do, go to Samsung?

 

[H433x0n]

It has already been made public that Apple isn't paying per wafer for N3, but for known good dies. So TSMC can't recognize revenue when wafers roll off the line, but much later in the process.

Apple isn't starting mass production of iPhone 15 in June. Even if they did TSMC isn't going to add an entry into that pie graph for some tiny amount like 0.17% if they shipped Apple 100K chips in June.

They absolutely would’ve been at high throughput during June for a September launch. I’m guessing they weren’t paid much for this until recently though.

Apparently A17 is still good to go for September.

 

[NostaSeronx]

That's a sweetheart deal. What is Apple going to do, go to Samsung?

There is no signs of Apple going to Samsung. Pretty much every CV across the web is that it is TSMC 3nm.

However, 3GAE isn't a fantasy though....

 

[jpiniero]

They absolutely would’ve been at high throughput during June for a September launch. I’m guessing they weren’t paid much for this until recently though.

Usually they are paid when the chips are delivered.

 

[oak8292]

Usually they are paid when the chips are delivered.

I think the chips aren’t delivered until INFO POP packaging is complete. Details of the contract and partial payment aren’t public. How much yield risk is their in packaging and has Apple accepted all that risk from TSMC?

 

[Ajay]

That's a sweetheart deal. What is Apple going to do, go to Samsung?

TSMC borked the original N3 process. That’s well known. TSMC made a less favorable deal with Apple because Apple's been the first silicon design company to put huge orders in the books for the latest process node for quite a few years now. This sort of deal saves face for the TSMC C-suite because they delivered on N3 (even if it had to be revised) and allows their best customer to hit some target important to their bottom. It’s not ideal, but if it works out well enough, both company's largest shareholders are satisfied. Both companies remain happy to be doing business with each other. I’m sure TSMC is sufficiently chastened to not screw up in a big way for a while. They don’t want large customers to 'even' be considering the possibility of dual sourcing their chips from SS or IFS down the road. IMHO.

 

[DrMrLordX]

TSMC is delaying the start date of the Arizona fab to 2025, claiming "labor shortages". I don't buy it.

Given their attitudes towards pay/benefits, it may actually be a semi-honest response. @Geddagod is probably on to something.

 

[SpudLobby]

The A suffix stands for Automotive. It’s a specific node that’s hardened and ruggedized to last 20 years of -40 to 105c temperature range to meet AEC-100 standards. It also requires a far more stringent standard of defect reduction to meet IATF 16969 standard.

I noticed that. I wonder what exactly they do to the node in order to achieve that e.g. specific cell libraries and chemical treatments. Probably an advanced packaging bonus flavor they sell with it.

 

[Ajay]

Given their attitudes towards pay/benefits, it may actually be a semi-honest response. @Geddagod is probably on to something.

Yes, it's not really a labor shortage. It's a shortage of, what TSMC considers reasonably cost effective labor. Of course, pitching yourself as offering good paying American jobs to get CHIPS ACT money makes this look pretty disingenuous. TSMC has complained about not being able to find enough well qualified workers with sufficient skills in semiconductor manufacturing - yet they'd like to open more FABs in China, where there is even less talent available (but it sure is cheap talent). So, give us $15US and please shut up and let us do it our way isn't exactly working out for TSMC. I hope with the delay TSMC does indeed upgrade the fab to handle various N3 process nodes - it would be worth the wait then.

 

[Doug S]

That's a sweetheart deal. What is Apple going to do, go to Samsung?

Is it?

It is a sweetheart deal if the alternative was "pay per wafer at the same price we'd be charging if yields were 90% instead of 55%", but if they charge such that Apple pays the same price per working chip that they would have paid if they were paying per wafer and yields were good" then it isn't a sweetheart deal. It is just giving Apple - by far their largest and most important customer - what the same value for dollar they've previously received.

Apple can't just drop everything and go to someone else, and Apple is unlikely to ever return to Samsung's foundry unless they had absolutely no choice. What Apple COULD eventually do though is go to Intel. They can't today because Intel doesn't have a node as advanced as N3B in mass production, and wouldn't have the capacity to handle Apple even if they did. But those things can change, and TSMC needs to keep Apple happy as the pull of going with Intel will be strong if Intel can simply match TSMC performance/power wise.

All else being equal Apple would prefer buying from a US company and definitely would prefer buying from a company that doesn't have a policy of making all its leading edge chips (i.e. what Apple needs) in a country that could possibly be involved in a war or blockade at some point in the future. Or could have a major earthquake knock out production for months at ANY point in the future.

Can't happen overnight, but if Intel can deliver on their promises I would expect Apple to be their lead customer by the end of the decade.

 

[moinmoin]

It is just giving Apple - by far their largest and most important customer - what the same value for dollar they've previously received.

And that's what makes it a sweetheart deal. As you pointed out Apple couldn't switch foundry now. TSMC could have moved the loss due to lower yield to Apple but didn't, taking the loss itself. That makes the cost predictable for Apple whereas TSMC has a huge incentive to improve the yield, whereas previously Apple profited of improving yield but wasn't really able to do much about it.

All else being equal Apple would prefer (...) [not] a country that could possibly be involved in a war or blockade at some point in the future.

This is getting OT and into P&N territory so I'll stop after this one time: Apple is a big lobby pro China due to it being a major market for them. Even though it's a country that is getting dangerously close to Russia and may well end up facing similar far reaching economic sanctions if they proceed as is.

 

[Geddagod]

And that's what makes it a sweetheart deal.

I think what constitutes a 'sweetheart deal' is very different for you, compared to the rest of us haha. Apple getting the same deal they usually get isn't TSMC heroically taking on the loss for its own sake, this was definitely to ensure good working business relationships with Apple. Ik we all joke about 'what will Apple do, move to Samsung?' but like I genuinely believe they could just move foundries (as long as it's not a regression gen on gen) and still maintain a large percentage of their consumer base. The M1 is the special case where it's not just Apple fanatics buying, bcuz battery life and all that, but lemme tell you what a bunch of kids in school appreciate just as much is the iphone/apple pairing that macs have.

Apple is a big lobby pro China due to it being a major market for them. Even though it's a country that is getting dangerously close to Russia and may well end up facing similar far reaching economic sanctions if they proceed as is.

Not nearly as much publicly. There's a reason Apple boxes have 'designed in California' and not 'assembled in china' on their boxes. All things being equal being the key, I think Apple would rather go with Intel rather than TSMC just for the positive PR.

 

[moinmoin]

I think what constitutes a 'sweetheart deal' is very different for you, compared to the rest of us haha.

What's a sweetheart deal to 'the rest of you'? In an economic reality where companies left and right use inflation to push hidden price rises for even higher margins?

 

[Ajay]

It is a sweetheart deal if the alternative was "pay per wafer at the same price we'd be charging if yields were 90% instead of 55%", but if they charge such that Apple pays the same price per working chip that they would have paid if they were paying per wafer and yields were good" then it isn't a sweetheart deal. It is just giving Apple - by far their largest and most important customer - what the same value for dollar they've previously received.

The math here doesn't work out in your favor here. If Apple is paying the same amount per working chip that they would if yields were >= 90%, but actually getting fewer chips, TSMC is losing money on that wafer (since about 1/2 off it is being thrown in the dumpster). Eh, what @moinmoin said.

Apple can't just drop everything and go to someone else, and Apple is unlikely to ever return to Samsung's foundry unless they had absolutely no choice. What Apple COULD eventually do though is go to Intel. They can't today because Intel doesn't have a node as advanced as N3B in mass production, and wouldn't have the capacity to handle Apple even if they did. But those things can change, and TSMC needs to keep Apple happy as the pull of going with Intel will be strong if Intel can simply match TSMC performance/power wise.

Agree. Probably THE best reason to give Intel CHIPS $$s - being able to reshore foreign chip production. If we can do a large amount of that, and reshore the a bunch of other processes/chemicals needed, then the benefits to the US High Tech manufacturing assets reduce our concerns over the dangers of China taking Taiwan by force. Semiconductor supply lines from Europe, So. Korea and Japan are much, much less risky.