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.

 

[Doug S]

Well, stock buybacks should be illegal, IMHO.

So are you saying companies should only be able to return capital to shareholders via dividends, or should not be able to do so at all? What do you think Apple should be allowed to do with all their excess cash generated each year if you ban them from buying back stock? Bigger dividends? Massive increase in R&D with questionable benefits? Hookers and cocaine for the C-suite?

Stock buybacks are the cousin to dividends, a way to return capital to shareholders - but different in two important ways. One, dividends are taxable to the recipient while buybacks aren't taxed until the stock is sold (i.e. to the extent a share is worth more when eventually sold creating a capital gain) and two, dividends tend to come with a shareholder expectation that a dividend stream will be maintained or increased absent special circumstances (unless a "special dividend" is declared that's generally seen as a one time thing)

The first point could be addressed by making buybacks taxable at a similar rate to dividends. Either pass that tax through to the stockholder like dividends or via an additional tax paid by the company when a buyback occurs. The second point is a matter of stockholder psychology. Companies don't want to declare a dividend sufficient to return all their excess capital to shareholders because next year if there is less excess capital (either because of lower profits or because of increased capital spending) then Wall Street tends to get unhappy and punishes the stock, which makes stockholders unhappy. So they would probably continue to use buybacks even if they were taxed sufficiently that there was no advantage over dividends or even a slight disadvantage to dividends, unless they were outlawed completely, simply because they are more flexible without the stigma of YoY decreases of dividends.

Yes buybacks can be abused - if a CEO is compensated in a way that his chances of earning a bonus / stock award / whatever are increased by reducing share count (thus increasing stock price and EPS, two common performance benchmarks) then he has incentive to push for buybacks even when that isn't the best use of the company's capital. But that's the fault of the compensation committee who should (and almost certainly does) know better, not the fault of the concept of stock buybacks.

 

[Ajay]

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.

Hmm, thought they used to be illegal. Maybe the SEC just had some cumbersome rule that lead to it not being worth it.

So are you saying companies should only be able to return capital to shareholders via dividends, or should not be able to do so at all? What do you think Apple should be allowed to do with all their excess cash generated each year if you ban them from buying back stock? Bigger dividends? Massive increase in R&D with questionable benefits? Hookers and cocaine for the C-suite?

Hookers and blow for the shareholders ;-) I understand what you've said - I know Buffett is a big fan of buybacks (owning more of the company that way). Still, while the math may work out the same either way (wrt to taxation), I prefer dividends. As you mention, it is tougher to abuse dividends. As it is, the level of transparency for a fair number of companies is atrocious. Buffett and Munger have pointed this out - in somewhat harsh terms w/o naming names. But they and their team have the wherewithal to ferret out BS. The rest of us don't. I was burnt on a stock once over 'acceptable accounting rules' - fortunately I only lost a few thousand dollars.

 

[DrMrLordX]

Hmm, thought they used to be illegal. Maybe the SEC just had some cumbersome rule that lead to it not being worth it.

Sort of. There was an anti-market manipulation restriction that was mostly knocked down in 1982. In practice stock buybacks are the only way for a public company to take itself private without an outside investor swooping in and doing it for them. So there has to be a way to retract your shares if for some reason your company no longer wishes to be publicly-traded (and has amassed the cash to pull back shares).

As for Intel, there's no obvious attempt to take themselves private going on, so of course their buyback patterns should be met with some skepticism. They could really use that money right about now.

 

[Doug S]

Oh forgot to mention that stock buybacks are (very lightly) taxed now, at 1%. That was added into the Inflation Reduction Act passed last year. Not sure why it was passed at such a low rate since I don't think it will affect things.

I guess having it on the books will allow the IRS and corporations to build that into their processes and the rate can be adjusted later if congress decides they want to make the treatment of buybacks and dividends more equivalent from a tax perspective.

Having the company pay the tax isn't quite the same as having stockholders pay it, since it would not be owed in a retirement account and can be avoided by taxpayers who have enough losses to offset it.

 

[ashFTW]

The stock price is such a “weak function” of EPS that stock buyback has never made sense to me from the corporation’s perspective. A large stock buyback could boost the stock price, by creating more buy pull. But I would rather have that cash in the bank and maybe use it for diversifying R&D and for strategic acquisitions. For example Apple could have potentially bought Netflix and/or Disney …

 

[Doug S]

The stock price is such a “weak function” of EPS that stock buyback has never made sense to me from the corporation’s perspective. A large stock buyback could boost the stock price, by creating more buy pull. But I would rather have that cash in the bank and maybe use it for diversifying R&D and for strategic acquisitions. For example Apple could have potentially bought Netflix and/or Disney …

Sure companies can trade at overly high or overly depressed multiples for some time (like Buffett says, the market can stay irrational longer than you can stay solvent) but in the long run stock price is the sum of expected future earnings so it is all about EPS. If you buy back 1% of your float then each share should be worth about 1% more - you just can't see that effect easily because of all the noise in day to day trading, quarter to quarter changes in expectations, and overall market fluctuations. But only a fool would doubt that reducing the number of shares makes each share more valuable. Go read what Buffett says about owning Apple and how much Berkshire Hathaway's percentage ownership of Apple has increased since they bought it despite hardly changing the number of shares they own, thanks to Apple reducing the share count substantially since his initial investment.

Apple buying Netflix or Disney would have been a disaster. I would have sold my Apple stock immediately if they did something that stupid! One of the best things about them is that they only make small carefully targeted acquisitions, and have a narrowly focused business. They know what their business is, who their customers are, and think carefully before expanding into other areas, despite the years of rumors (always from Wall Street investment bankers who love the idea of the massive fees they would collect for brokering such a deal) that they would go on a wild spending spree buying companies left and right.

Paying for Netflix's overpriced stock (and at a premium to its already too high price like any buyout requires) would be a terrible use of their capital. Return it to the shareholders via dividends or buybacks and let US choose if we want to own that stock! Disney would be an even worse purchase, because then Apple would be in all kinds of crazy businesses like toys, theme parks and cruises, and in the middle of political battles in Florida. Say goodbye to Apple's narrow focus, they might as well buy a chain of grocery stores and an airline if all that narrow focus is going out the window!

What would be the point of buying Netflix? Buying super fickle streaming customers? A content library made up of mostly crap? If Apple wanted to spend all those billions to improve their fortunes in the streaming business, think how much content they could develop for 1/10th of that money. Buying your way into a market with minimal barriers to entry makes zero economic sense.

 

[ashFTW]

Of course reducing the number of shares, increases the % of company owned. That’s pure math. But how much does it influence the actual share price? A much bigger influencer of stock price is the potential of increased earnings, which is demonstrated by capturing more market share and/or creating new markets.

Anyway, this discussion is off topic and doesn’t belong here.

 

[RnR_au]

https://twitter.com/x/status/1673503824865300480

Not sure if N1.4 has had a tentative production date published before. A 2 year window is not much of a 'date', but its a beginning for a long engineering effort.

 

[moinmoin]

Good story of South Korean engineers moving to China trying to build fabs there (the tweet's summary is jumping to conclusion a little much):

https://twitter.com/x/status/1673489327731675136

 

[A///]

China's paying big bucks to foreign engineers. the offers are lush. gonna be interesting to see how everyone sits at the dining table by the end of this decade. who will be at the head of household seat who knows.

 

[Ajay]

I think I'm in the wrong thread, but here goes anyway:

Update on the Chips act from Techpowerup.com

U.S. Administration Outlines Plan to Strengthen Semiconductor Supply Chains

Today, the U.S. Department of Commerce shared the Biden-Harris Administration's strategic vision to strengthen the semiconductor supply chain through CHIPS for America investments. To advance this vision, the Department announced a funding opportunity and application process for large...

www.techpowerup.com

Current CHIPS manufacturing incentives application processes and timelines:
For leading-edge commercial facilities: Both pre-applications (optional) and full applications are being accepted on a rolling basis.
For current-generation and mature-node commercial facilities: Pre-applications (optional but recommended) are currently being accepted on a rolling basis and full applications will be accepted on a rolling basis starting June 26, 2023.
For larger materials and manufacturing equipment supplier facility projects with capital investments equal to or exceeding $300 million: Pre-applications (optional but recommended) will be accepted on a rolling basis starting September 1, 2023, and full applications will be accepted on a rolling basis starting October 23, 2023.
For all potential applicants: The Department continues to accept statements of interest on a rolling basis to further inform the program of interest and make for efficient application review.
Forthcoming CHIPS manufacturing incentives applications processes and timelines:

For smaller materials and manufacturing equipment supplier facility projects below $300 million: The Department will release an additional funding opportunity in the fall with details on the application process and timeline. For more information about this forthcoming funding opportunity, see here.
For commercial R&D facilities: The Department will subsequently release a separate funding opportunity with details on the application process and timeline.

International Coordination with U.S. Partners and Allies
As CHIPS for America invests across the supply chain, the Department of Commerce will prioritize robust international engagement. Through bilateral and multilateral dialogues, and business-to-business and government-to-business forums, the Department will work with likeminded partners to strengthen and diversify the global semiconductor supply chain.

The Department's CHIPS-related international engagement to date has included engagements with the Republic of Korea, Japan, India, and the United Kingdom, and through the Indo-Pacific Economic Framework, European Union-United States Trade and Technology Council, and North America Leaders' Summit. The Department will continue coordinating closely with U.S. partners and allies to advance these shared goals, advance our collective security, and strengthen global supply chains.

So, I think there is better support for building a safer supply chain - which I really wanted be the case. I'd like the see more in North America, so we'll see how this goes...

 

[moinmoin]

So, I think there is better support for building a safer supply chain - which I really wanted be the case. I'd like the see more in North America, so we'll see how this goes...

Certainly sounds way more well thought out and thorough than the knee jerk funding Intel managed to squeeze out from the German government.

 

[Doug S]

https://twitter.com/x/status/1673503824865300480

Not sure if N1.4 has had a tentative production date published before. A 2 year window is not much of a 'date', but its a beginning for a long engineering effort.

Looks like TSMC has taken back the lead from Intel in roadmap claims years from realization. Look for Intel to announce 14A and 12A soon to take it back

 

[H433x0n]

TSMC N2 pricing

If those prices are anywhere close to accurate that’s insane.

 

[Doug S]

TSMC N2 pricing

If those prices are anywhere close to accurate that’s insane.

So if you figure a square 100 mm^2 die and a much worse than acceptable 80% yield you get about 500 chips per wafer, which comes out to $50 per chip. That's an increase to be sure but an extra $10 per smartphone or laptop is hardly an issue. For the low end stuff where it is they already aren't using lead edge processes for their CPUs/SoCs.

 

[H433x0n]

So if you figure a square 100 mm^2 die and a much worse than acceptable 80% yield you get about 500 chips per wafer, which comes out to $50 per chip. That's an increase to be sure but an extra $10 per smartphone or laptop is hardly an issue. For the low end stuff where it is they already aren't using lead edge processes for their CPUs/SoCs.

I’ve got $26 to $48 for the same die when going from an iPhone 14 to a hypothetical iPhone 16/17 with an SoC on N2 node (using same dimensions). Those dollar figures aren’t accurate (I doubt these companies are actually paying anywhere near the prices listed in this article) and the die wouldn’t stay the same size either but the ~80% increase in price from N5 -> N2 definitely far outpaces the increased density and performance.

 

[jerrytw]

https://semiwiki.com/semiconductor-...141-samsung-foundry-on-track-for-2nm-in-2025/
'Samsung will begin mass production of the 2nm process for mobile applications in 2025, and then expand to HPC in 2026 with backside power delivery, and automotive in 2027. Samsung’s 2nm (SF2) process has shown a 12% increase in performance, a 25% increase in power efficiency, and a 5% decrease in area, when compared to its 3nm process (SF3). Mass production of 1.4nm is slated for 2027.'

 

[Doug S]

I’ve got $26 to $48 for the same die when going from an iPhone 14 to a hypothetical iPhone 16/17 with an SoC on N2 node (using same dimensions). Those dollar figures aren’t accurate (I doubt these companies are actually paying anywhere near the prices listed in this article) and the die wouldn’t stay the same size either but the ~80% increase in price from N5 -> N2 definitely far outpaces the increased density and performance.

No way N2 will be ready for iPhone 16 next year, so that's at best a three year gap from a very mature N5 family node to N2 - two node jumps (and it isn't clear N2 will be ready for the 2025 iPhone...) All that for a ~$20 increase in a ~$400 BOM for an iPhone Pro - i.e. 5%. That's nothing, they have taken bigger jumps in the BOM when they introduced Retina, when they introduced OLED, probably at various times as the cameras have improved / increased in number. Heck 10 years ago the BOM was more like $200, it has grown over time as they added the plus/pro lines and ASPs have increased.

Other things that aren't improving often come in down in price to compensate. i.e. Retina cost more when new but over time came down in price. Then it jumped again for the OLED displays, but those are getting lower over time. The contribution to BOM of the base flash has gone down over the years despite that base increasing. Ditto for LPDDR, which is really only increasing as dies get denser but there is more RAM than is really needed now so only "that's the minimum efficient package size" will cause further increases.

If you want to look for something big to drive down the BOM between iPhone 14 and iPhone 17, look at the transition from Qualcomm's modem to Apple's. That's going to seriously cut down their BOM. Inevitably lead to lawsuits too knowing Qualcomm, but even if they end up winning a few billion in court sometime in the 2030s, Apple will still have come out way ahead because of the way Qualcomm does their licensing when you buy their chips versus how they are licensing for other companies making modems.

 

[A///]

No way N2 will be ready for iPhone 16 next year, so that's at best a three year gap from a very mature N5 family node to N2 - two node jumps (and it isn't clear N2 will be ready for the 2025 iPhone...) All that for a ~$20 increase in a ~$400 BOM for an iPhone Pro - i.e. 5%. That's nothing, they have taken bigger jumps in the BOM when they introduced Retina, when they introduced OLED, probably at various times as the cameras have improved / increased in number. Heck 10 years ago the BOM was more like $200, it has grown over time as they added the plus/pro lines and ASPs have increased.

Other things that aren't improving often come in down in price to compensate. i.e. Retina cost more when new but over time came down in price. Then it jumped again for the OLED displays, but those are getting lower over time. The contribution to BOM of the base flash has gone down over the years despite that base increasing. Ditto for LPDDR, which is really only increasing as dies get denser but there is more RAM than is really needed now so only "that's the minimum efficient package size" will cause further increases.

If you want to look for something big to drive down the BOM between iPhone 14 and iPhone 17, look at the transition from Qualcomm's modem to Apple's. That's going to seriously cut down their BOM. Inevitably lead to lawsuits too knowing Qualcomm, but even if they end up winning a few billion in court sometime in the 2030s, Apple will still have come out way ahead because of the way Qualcomm does their licensing when you buy their chips versus how they are licensing for other companies making modems.

Qualcomm's lawsuits are like Intel's back in the day when they swung their muscle around only to be crapped on by everyone else now. I don't see n2 being ready either but see apple reusing N3 and optimizing their new design?

I want my damn m3 mac book pro damn it.

 

[moinmoin]

the ~80% increase in price from N5 -> N2 definitely far outpaces the increased density and performance.

The general trend, increasing cost for decreasing amounts of improvements, should be a surprise to nobody at this point. So I'm not seeing the news in your statement, it's just a confirmation what had been the saying for quite some time already. Cost outpaces improvements, thus Moore's law is dead.

 

[Ajay]

No way N2 will be ready for iPhone 16 next year, so that's at best a three year gap from a very mature N5 family node to N2 - two node jumps (and it isn't clear N2 will be ready for the 2025 iPhone...) All that for a ~$20 increase in a ~$400 BOM for an iPhone Pro - i.e. 5%. That's nothing, they have taken bigger jumps in the BOM when they introduced Retina, when they introduced OLED, probably at various times as the cameras have improved / increased in number. Heck 10 years ago the BOM was more like $200, it has grown over time as they added the plus/pro lines and ASPs have increased.

Other things that aren't improving often come in down in price to compensate. i.e. Retina cost more when new but over time came down in price. Then it jumped again for the OLED displays, but those are getting lower over time. The contribution to BOM of the base flash has gone down over the years despite that base increasing. Ditto for LPDDR, which is really only increasing as dies get denser but there is more RAM than is really needed now so only "that's the minimum efficient package size" will cause further increases.

If you want to look for something big to drive down the BOM between iPhone 14 and iPhone 17, look at the transition from Qualcomm's modem to Apple's. That's going to seriously cut down their BOM. Inevitably lead to lawsuits too knowing Qualcomm, but even if they end up winning a few billion in court sometime in the 2030s, Apple will still have come out way ahead because of the way Qualcomm does their licensing when you buy their chips versus how they are licensing for other companies making modems.

From the AT N2 article:


Seems much more likely to me Apple will use a late phase N3, especially given the experience of staying the course with N3'B' - having to switch libraries for all follow-on SoCs. IMHO.

 

[Geddagod]

From the AT N2 article:
View attachment 82351

Seems much more likely to me Apple will use a late phase N3, especially given the experience of staying the course with N3'B' - having to switch libraries for all follow-on SoCs. IMHO.

What's N5A? Never saw that name so far

 

[Curious_Inquirer]

What's N5A? Never saw that name so far

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.

 

[Doug S]

Seems much more likely to me Apple will use a late phase N3, especially given the experience of staying the course with N3'B' - having to switch libraries for all follow-on SoCs. IMHO.

Pretty sure I read somewhere that N2 was H2 2025, i.e. the same bad timing for Apple as N3. Of course it is two years out so maybe they were just being conservative when they said H2 so it isn't seen as a delay if it doesn't hit in time for Apple.

I'm thinking more and more that if M3 is built on N3E from day one (meaning no M3 Macs this fall) that A17 will have been originally designed on N3E, and the N3B version would be the port not the other way around.

No way Apple would use an N3 variant if N2 is available to them for A19. They have always gone with TSMC's best available process and I see no reason why they would prefer N3P over N2 unless N2 isn't ready on time.

 

[jpiniero]

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.