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.

 

[Exist50]

The way I read it, they have lots of capacity in 16nm and maybe 14nm.

Intel isn't offering 14nm as a foundry node, only Intel 16, 3, and 18A so far. But since 16/22FFL uses some of the 14nm manufacturing flow, maybe capacity is fungible.

 

[DrMrLordX]

Sure but they have to start somewhere.

Intel launched 22FFL in 2017 specifically for its foundry business. They started years ago.

The way I read it, they have lots of capacity in 16nm and maybe 14nm.

I'm still curious about Intel 16. Is it just 22FFL rebranded? I mean probably, sounds like a 14nm-family node or something . . .

Intel isn't offering 14nm as a foundry node, only Intel 16, 3, and 18A so far.

The idea of Intel offering Intel 3 as a foundry node is a little weird. It doesn't exist yet and oh wait I don't want to start the "Intel has more EUV machines than you think" conversation so nm, maybe they will have the capacity. But that still means the foundry would be competing with Granite Rapids and who knows what else for wafers. I would not be surprised to see Intel launch a version of Arrow Lake on Intel 3 instead of 20a.

But since 16/22FFL uses some of the 14nm manufacturing flow, maybe capacity is fungible.

Yeah that's my assumption as well.

 

[JasonLD]

The idea of Intel offering Intel 3 as a foundry node is a little weird.

I don't expect Intel 3 would be available for foundry before Intel moves onto 18A with high-NA EUV. So probably after 2025.

 

[Hitman928]

I'm still curious about Intel 16. Is it just 22FFL rebranded? I mean probably, sounds like a 14nm-family node or something . . .

Intel 16 is an improved 22FFL. You could call it 22FFL+.

 

[Thibsie]

Intel 16 is an improved 22FFL. You could call it 22FFL+.

Can't wait for 22FFL+++++++

 

[jpiniero]

Hey, the Senate passed the Chips bill. The House still has to vote on it but it should be a formality and should be on Biden's desk by the weekend.

Intel will be happy with all that Free Money they get to blow... although I didn't look into where the money is actually going.

 

[Saylick]

Hey, the Senate passed the Chips bill. The House still has to vote on it but it should be a formality and should be on Biden's desk by the weekend.

Intel will be happy with all that Free Money they get to blow... although I didn't look into where the money is actually going.

Yeah, I have mixed feelings about the CHIPS bill. While I think it is important that the US has more chip manufacturing on American soil, how the Free Money is spent doesn't appear to be very well regulated. I would want to see Intel's manufacturing side be split out from their design side where each division operates independently, not unlike how Samsung has their fab and design side separate. The important thing would be that each division has their own accounting and financial goals such that Intel's design side has to compete with other fabless companies for Intel fab wafers. The way things currently stand, Intel's design side essentially gets Intel wafers at cost (read: without a markup, unlike fabless companies who pay for GF or TSMC wafers) and so a portion of the Free Money would go directly into making Intel more competitive against its fabless competitors rather than benefiting anyone and everyone who could use Intel IDM. What if Intel decides that only they get the best Intel nodes, and the IDM nodes are always going to be 6 months behind? This bill doesn't appear to have provisions to prevent using the money for anti-competitive behavior.

 

[DrMrLordX]

Hey, the Senate passed the Chips bill. The House still has to vote on it but it should be a formality and should be on Biden's desk by the weekend.

Intel will be happy with all that Free Money they get to blow... although I didn't look into where the money is actually going.

That's sad. Intel is going to soak up most of that money as its reward for destroying the diversity of American semiconductor fabrication. How utterly ridiculous.

 

[Ajay]

Intel's design side has to compete with other fabless companies for Intel fab wafers.

The way things currently stand, Intel's design side essentially gets Intel wafers at cost (read: without a markup, unlike fabless companies who pay for GF or TSMC wafers) and so a portion of the Free Money would go directly into making Intel more competitive against its fabless competitors rather than benefiting anyone and everyone who could use Intel IDM.

1) Intel having to compete with other companies for wafer allotments from their own fabs is just plain nutz. That would be a business plan designed to fail. Intel may have become fat, stupid and happy when hey had a big lead, but they are not complete morons.

2). It's a common fallacy to believe Intel gets their wafers at cost. Intel alone has to carry the fully burdened cost of building and operating its fabs. AMD shares in paying TSMC's costs with other top semiconductor design companies. The advantage for Intel was in designing a process that precisely met the needs of new CPUs - until they borked that.

 

[Ajay]

That's sad. Intel is going to soak up most of that money as its reward for destroying the diversity of American semiconductor fabrication. How utterly ridiculous.

The original Intent included the 're-shoring' of other critical semiconductor suppliers. Not sure what the current bill says, but bringing silicon ingot manufacturing back to the US would be a positive development.

 

[Thibsie]

And it seems AVX512's state won't evolve much...

Intel Meteor Lake-P Bootlog | Coelacanth's Dream

www.coelacanth-dream.com

 

[Saylick]

1) Intel having to compete with other companies for wafer allotments from their own fabs is just plain nutz. That would be a business plan designed to fail. Intel may have become fat, stupid and happy when hey had a big lead, but they are not complete morons.

2). It's a common fallacy to believe Intel gets their wafers at cost. Intel alone has to carry the fully burdened cost of building and operating its fabs. AMD shares in paying TSMC's costs with other top semiconductor design companies. The advantage for Intel was in designing a process that precisely met the needs of new CPUs - until they borked that.

1) I'm not so sure it's a business plan designed to fail. If TSMC can be a successful business as a pure play foundry and if AMD and NVDA can be successful businesses without a foundry, why can't a split Intel be equally successful?
2) Maybe that's even more reason for Intel to split off their foundry business as an independent entity so that their design side doesn't have to be burdened by the entire cost of their foundry business.

 

[DrMrLordX]

The original Intent included the 're-shoring' of other critical semiconductor suppliers. Not sure what the current bill says, but bringing silicon ingot manufacturing back to the US would be a positive development.

Devil's in the details. We won't see the actual effects of this bill for many years to come. Outside of the initial cash splash on companies like Intel, and . . . Intel.

 

[Thunder 57]

Pretty sad that Intel had to say "ya know what, maybe we won't build a fab in Ohio after all. Let's see what Europe thinks.". But yes, the devil is in the details. It's a shame that only one company is a clear winner with this bill though.

 

[Saylick]

Pretty sad that Intel had to say "ya know what, maybe we won't build a fab in Ohio after all. Let's see what Europe thinks.". But yes, the devil is in the details. It's a shame that only one company is a clear winner with this bill though.

Yeah, I think we all know that Intel has the money to fund a domestic fab (*glares at Intel's prior stock buybacks and generous dividend*), but they choose not to because Europe is offering their own subsidy. The fact that Intel can make that statement is basically saying, "We'll build wherever people willing to pay us to build, geography location be damned". If geographic diversity was so important to Intel, they'd make sure they get a fab on every major continent and build on US soil with or without government money.

 

[DrMrLordX]

Yeah, I think we all know that Intel has the money to fund a domestic fab (*glares at Intel's prior stock buybacks and generous dividend*), but they choose not to because Europe is offering their own subsidy. The fact that Intel can make that statement is basically saying, "We'll build wherever people willing to pay us to build, geography location be damned". If geographic diversity was so important to Intel, they'd make sure they get a fab on every major continent and build on US soil with or without government money.

They need cash to offset spending on wafers from TSMC and spending to shore up their flagging R&D efforts. So with one hand they spend to build a fab in Ohio, but with the other they send cash overseas.

 

[Ajay]

1) I'm not so sure it's a business plan designed to fail. If TSMC can be a successful business as a pure play foundry and if AMD and NVDA can be successful businesses without a foundry, why can't a split Intel be equally successful?
2) Maybe that's even more reason for Intel to split off their foundry business as an independent entity so that their design side doesn't have to be burdened by the entire cost of their foundry business.

So, exactly when to you see the top buyers (Apple, AMD, Nvidia) switching over to Intels's fabs? Intel won’t get any of these companies while Intel is an IDM. Intel can’t afford to operate a leading edge fab without at least one of these customers. Apple is pretty much done with Intel. AMD and Nvidia have conflicts with Intel’s product line. That leaves Qualcomm. They have Mediatek, but on an n-2 node (or whatever). Sucks being Intel right now.

 

[moinmoin]

So does or doesn't the CHIPS act include efforts to get TSMC and Samsung to invest into the US instead of China?

 

[Doug S]

So, exactly when to you see the top buyers (Apple, AMD, Nvidia) switching over to Intels's fabs? Intel won’t get any of these companies while Intel is an IDM. Intel can’t afford to operate a leading edge fab without at least one of these customers. Apple is pretty much done with Intel. AMD and Nvidia have conflicts with Intel’s product line. That leaves Qualcomm. They have Mediatek, but on an n-2 node (or whatever). Sucks being Intel right now.

Why would Apple refuse to consider Intel?

Now I'm not suggesting they will - Apple's very close relationship with TSMC that extends even to partnering on process development means the bar would be VERY high for Intel, but if they really were able to deliver on their roadmap and take back process leadership, Apple would have to at least consider them.

Intel being an IDM wouldn't matter to them, what would matter would be an ironclad commitment to get the number of wafers they need; the same commitment as TSMC is currently giving them. I think it is more likely that Apple's demands would cause Intel to turn them away than for Apple to refuse to deal with Intel because they're an IDM (and at best a secondhand competitor to Apple) or because they're "done with Intel".

Apple wouldn't have to go all in with a high volume product like iPhone SoCs, they could start with something lower volume like AS Pro/Max chips or Apple Watch SoCs.

 

[DrMrLordX]

So does or doesn't the CHIPS act include efforts to get TSMC and Samsung to invest into the US instead of China?

Good question. The existing expansion plans for both companies don't seem reliant on the CHIPS act.

 

[DrMrLordX]

So I guess TSMC and GF, and possibly others, will get some money too? Specifically ,any company building a fab in the US gets a 25% tax credit (details unknown, I'm just reading summaries).

Here's a slightly-old article on competing versions of the CHIPs act:

Chip industry splits over U.S. CHIPS Act benefits for Intel

Several U.S. chip firms are reportedly debating whether to oppose $52 billion in congressional subsidies for chip fab construction if the final deal disproportionately favors Intel and others. | Chip designers like AMD, Nvidia and Qualcomm favor a federal design credit, not just benefits to new fabs

www.fierceelectronics.com

Not sure which version will go into law, but if you look one version essentially had ~$30 billion going to Intel (potentially) without anything going to their design competitors (AMD, Qualcomm, NV, etc.).

I haven't seen much indication of how much cash or what incentives will go to anyone that is not Intel, aside from the aforementioned tax credit that will help companies like Samsung and TSMC to build here.

 

[pakotlar]

So I guess TSMC and GF, and possibly others, will get some money too? Specifically ,any company building a fab in the US gets a 25% tax credit (details unknown, I'm just reading summaries).

Here's a slightly-old article on competing versions of the CHIPs act:

Chip industry splits over U.S. CHIPS Act benefits for Intel

Several U.S. chip firms are reportedly debating whether to oppose $52 billion in congressional subsidies for chip fab construction if the final deal disproportionately favors Intel and others. | Chip designers like AMD, Nvidia and Qualcomm favor a federal design credit, not just benefits to new fabs

www.fierceelectronics.com

Not sure which version will go into law, but if you look one version essentially had ~$30 billion going to Intel (potentially) without anything going to their design competitors (AMD, Qualcomm, NV, etc.).

I haven't seen much indication of how much cash or what incentives will go to anyone that is not Intel, aside from the aforementioned tax credit that will help companies like Samsung and TSMC to build here.

The overall bill is ~250B, most of it is going to companies besides Intel.

 

[oak8292]

The overall bill is ~250B, most of it is going to companies besides Intel.

Here is a link the overall bill. The CHIPs Act is still only 52 billion and the rest is supporting a lot of the science programs, e.g. NSF. NIST and NASA.

View the CHIPS+ Legislation

www.commerce.senate.gov

Here is an excerpt from the CHIPs part of the bill.

"$39 billion in financial assistance to build, expand, or modernize domestic facilities and equipment for semiconductor fabrication, assembly, testing, advanced packaging, or research and development, including $2 billion specifically for mature semiconductors. Within the incentive program, up to $6 billion may be used for the cost of direct loans and loan guarantees."

I believe Micron and others will also trying to get some of this funding.

There was a study done in 2020 on what should be done and the impact of some different funding scenarios.

https://www.semiconductors.org/wp-content/uploads/2020/09/Government-Incentives-and-US-Competitiveness-in-Semiconductor-Manufacturing-Sep-2020.pdf

The $50 billion funding was expected to increase the number of fabs between 2020 and 2030 by about 10 through incentives and increase the global percentage of fabs in the U.S. from about 10% (do nothing) to 14% (+$50 billion) of global capacity. The current state is the U.S. has about 12% of global capacity (page 23).

The report also includes a comparison of the current incentives that are already in place. All of the countries already have incentives, including the U.S. China has the most incentives and Korea and Taiwan have higher levels of incentives than the U.S.

 

[MadRat]

Written to not exclude China, and its going to benefit them more than Intel.

 

[moinmoin]

Meanwhile in China's fledgling chipmaking industry:

Corruption is sending shock waves through China’s chipmaking industry

The arrests of several top semiconductor fund executives could force the government to rethink how it invests in the sector.

www.technologyreview.com

I posted about this topic before in this thread, so it's overdue there are some more visible correction moves done.

Money quote: “This is the least surprising corruption investigation I’ve heard of for a while”