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.

 

[DrMrLordX]

Ericsson silicon is on Intel 4 and was introduced in November.

Yes, and?

 

[moinmoin]

Yes, and?

IFS is slightly beyond vaporware status now.

 

[DavidC1]

Yes, and?

That your claim that they won't have 3 and 18A is invalid since they already have it on Intel 4, something that isn't even official IFS process.

 

[Ajay]

IFS is slightly beyond vaporware status now.

Sure feels that way. I wonder If Gelsinger will be at the helm in 2-3 more years. He either has a rabbit up his sleeve - or IFS becomes a minor part of Intel's business. I haven seen any signs that Intel has gained any significant design wins - and slowing FAB buildouts is a terrible indicator of that. They really need to be shipping as much silicon from IFS as from their own silicon products. IMHO.

 

[trivik12]

OpenAI CEO Sam Altman seeks $5 to $7 trillion to build a network of fabs for AI chips

The world needs more chips, and Sam Altman wants to find money to build them.

www.tomshardware.com

With 5-7 trillion, Altman can buy TSMC itself

Rumors about Sam Altman's chip production ambitions have been circulating for a while, and now they are finally taking shape. Apparently, Sam Altman is not going to establish a company that would develop and build processors for artificial intelligence (AI). What he is working on is a very ambitious effort: He wants to raise $5 trillion to $7 trillion to build a network of fabs run by foundries that would produce chips not only for OpenAI, but for others, too, reports the Wall Street Journal. If he succeeds, the project will reshape the whole semiconductor industry.

Sam Altman is reportedly exploring a collaborative venture that would connect OpenAI, various investors, contract chip manufacturers, and power providers to build a network of fabs that would make ample use of chips for OpenAI (the company will commit to become a significant client for these chip plants) and other clients. The fabs will be operated by existing foundries (the WSJ specifically mentions TSMC, which is reportedly a company that Altman had met in recent weeks), whereas power plants for these fabs will be run by other companies. The whole project will require investments between $5 trillion to $7 trillion, according to the WSJ.

 

[Tigerick]

Is the amount of money correct?

You do realize that OpenAI could buy Microsoft AND Apple with $6 trillion dollars ?

 

[Doug S]

Its probably crazy ideas like "I need $5 to $7 trillion to do what I'm thinking of, which includes replicating the entire worldwide semiconductor production supply chain" that led to his (temporary) firing last year.

But yeah, for much less money than that he could buy ASML, TSMC, Nvidia, and Micron and have all the major pieces, though the full supply chain is much larger.

 

[CakeMonster]

Altman has been scaring journalists with talking about AI taking over the world to hype his company "Look at this powerful tech!", so yeah this kind of hubris wouldn't actually surprise me.

 

[DrMrLordX]

That your claim that they won't have 3 and 18A is invalid since they already have it on Intel 4, something that isn't even official IFS process.

I never claimed that. I stated that we haven't seen any products released on those nodes.

Intel 4 is not exactly Intel 3 either, plus we won't be able to guess any of Intel 4's node characteristics (nor Intel 3's) based on the chip since it isn't going to be in any consumer devices that can be tested or compared to anything else. We're talking low-voltage, likely low-clockspeed custom chips going into comm gear.

 

[trivik12]

https://twitter.com/x/status/1755294743565930726

Sam is doubling down on AI infrastructure. May be there is some truth to the ridiculous news above.

 

[Saylick]

IFS is slightly beyond vaporware status now.

Anyone got a Semiaccurate subscription?

Who is the first big customer for Intel’s foundry efforts?

The question on everyone’s mind is what company is Intel’s marquee foundry customer?

www.semiaccurate.com

 

[trivik12]

Has to be NVidia for packaging.

 

[clemsyn]

Anyone got a Semiaccurate subscription?

Who is the first big customer for Intel’s foundry efforts?

The question on everyone’s mind is what company is Intel’s marquee foundry customer?

www.semiaccurate.com

DOD?

 

[H433x0n]

Has to be NVidia for packaging.

That’s just packaging, that’s not what this article is talking about.

DOD?

Of course the DoD is a customer but that’s tiny volume.

Anyone got a Semiaccurate subscription?

Who is the first big customer for Intel’s foundry efforts?

The question on everyone’s mind is what company is Intel’s marquee foundry customer?

www.semiaccurate.com

You can look at the guest list and get a good idea.

MediaTek, Broadcom & OpenAI. OpenAI is supposedly using IFS foundry + Intel IP to aid in their first SoC design. There’s also Faraday taping out on 18A but they’re a small company so I don’t know how much revenue that’s bringing.

Both MediaTek & Broadcom are pretty big customers. I think Nvidia will probably fab some things on 18A but that’s not confirmed yet.

 

[DavidC1]

Has to be NVidia for packaging.

They've been doing packaging for some time now and it's only worth few hundred million dollars at best. It's using process technology that will have real material benefits.

 

[maddie]

Is the amount of money correct?

You do realize that OpenAI could buy Microsoft AND Apple with $6 trillion dollars ?

Is he even rational? Sure a genius, but possible Savant?

Global Semiconductor Sales Hit $526.8 Billion in 2023. This is sales, not profit, so what's the return on investment and is this still even relevant in today's economic paradigm?

Global Semiconductor Sales Hit $526.8 Billion in 2023

www.anandtech.com

 

[igor_kavinski]

His evil plan is to create an even bigger GPT network than the one he currently has. Then he can use that to create a more advanced human DNA, inject his consciousness into that and rule the world.

 

[Tigerick]

His evil plan is to create an even bigger GPT network than the one he currently has. Then he can use that to create a more advanced human DNA, inject his consciousness into that and rule the world.

The Matrix GPT Edition

 

[FlameTail]

Rumor: Apple may raise 3nm orders 50% to TSMC this year as it boosts specs on iPads, MacBooks and iPhones for AI, and puts AI computing cores in next generation M4 and A18 processors, media report, adding Apple will also reserve a big amount of advanced packaging, including InFO and CoWoS, and potentially even SoIC.

x.com

x.com

Apple already uses InFo for M Ultra chips.

What will they use CoWoS and SoIC for?

 

[FlameTail]

It is the end.

 

[Doug S]

x.com

x.com

Apple already uses InFo for M Ultra chips.

What will they use CoWoS and SoIC for?

Pretty sure they are using CoWoS or something like it for Ultra. InFO is what they use for the iPhone, they can't use it to make the Ultra.

 

[DrMrLordX]

It is the end.

Nah, BoM has never been the deciding factor in the console wars. MS and Sony both use the hardware as a loss leader. The could probably raise licensing fees per copy sold/per microtransaction by a few percent to offset increases to BoM. People are already having to pay $70 for a AAA game now, the will not notice much or complain if they wind up paying $72-$75. If they pull a stupid and jack up console prices again, yes, people will notice and complain.

 

[FlameTail]

Nah, BoM has never been the deciding factor in the console wars. MS and Sony both use the hardware as a loss leader. The could probably raise licensing fees per copy sold/per microtransaction by a few percent to offset increases to BoM. People are already having to pay $70 for a AAA game now, the will not notice much or complain if they wind up paying $72-$75. If they pull a stupid and jack up console prices again, yes, people will notice and complain.

What about everything else other than consoles?

In 5 years we gonna have $5000 GPUs?

 

[FlameTail]

Pretty sure they are using CoWoS or something like it for Ultra. InFO is what they use for the iPhone, they can't use it to make the Ultra.

Sorry, got InFo and CoWoS mixed up.

 

[SteinFG]

What about everything else other than consoles?

In 5 years we gonna have $5000 GPUs?

We already have 5000 dollar professional GPUs. As for gaming, I doubt it, as the market for such gaming GPU will be infinitesimally small. I think 2500 is the upper limit for now (Titan RTX price).