TSMC Doesn't See Intel as Long-Term Customer, Unlikely to Build Additional Capacity for It

[AleksandarK]

TSMC has been the backbone of silicon designers for a long time. Whenever you question where you can use the latest technology and get some good supply capacity, TSMC got everyone covered. That case seems to be similar to Intel and its struggles. When Intel announced that its 7 nm semiconductor node is going to be delayed a full year, the company's customers and contractors surely became worried about the future releases of products and their delivery, like the case is with Aurora exascale supercomputer made for Argonne National Laboratory, which relies on Intel's 7 nm Ponte Vecchio graphics cards for most of the computation power.

To manage to deliver this, Intel is reportedly in talks with TSMC to prepare capacity for the GPUs and deliver them on time. However, according to industry sources of DigiTimes, TSMC is unlikely to build additional capacity for Intel, besides what it can deliver now. According to those sources, TSMC does not see Intel as a long-term customer and it is unknown what treatment will Intel get from TSMC. Surely, Intel will be able to make a deal with TSMC and secure enough of the present capacity for delivering next-generation processors.


View at TechPowerUp Main Site

 

[Noztra]

"TSMC is unlikely to build additional capacity for Intel, besides what it can deliver now. According to those sources, TSMC does not see Intel as a long-term customer and it is unknown what treatment will Intel get from TSMC. Surely, Intel will be able to make a deal with TSMC and secure enough capacity for delivering next-generation processors."

This doesn't make any sense?

 

[john_]

It's in TSMC's interest Intel to just go away in 10-15 years. It's not only a (possible future) competitor to TSMC in manufacturing, but just imagine how much more market there would be for TSMC if ARM catches up with X86 and starts replacing it in even more markets, for exampe even where high performance is the priority.

 

[dj-electric]

Intel: So here's money
TSMC: idk about that fam
Intel: Here's more money
TSMC: .... fine.

 

[fynxer]

Intel, earlier this decade, bragging their tech was so advanced that there where no chance ever for the competition to catch up.

Now TSMC is sticking it to them they wont trust Intel, will only let them have leftover capacity.

Karma is a b!tch

 

[TheLostSwede]

Intel, earlier this decade, bragging their tech was so advanced that there where no chance ever for the competition to catch up.

Now TSMC is sticking it to them they wont trust Intel, will only let them have leftover capacity.

Karma is a b!tch

I think you mean last decade?

 

[Calmmo]

I mean technically speaking 2020 is the final year of the decade We start from year 1.

 

[Vya Domus]

It's in TSMC's interest Intel to just go away in 10-15 years. It's not only a (possible future) competitor to TSMC in manufacturing

A competitor without a proper leading node ? Nah, at best they would want them to lose their fabs so they have to buy wafers from them.

Surely, Intel will be able to make a deal with TSMC and secure enough of the present capacity for delivering next-generation processors.

For Intel that's code word for "We're gonna pay you to reduce the amount of wafers given to AMD or anyone else".

 

[Shatun_Bear]

If I were TSMC I would charge a massive premium to Intel per wafer, to really drain that war chest Intel has.

We know Intel's game here; they're trying to stop AMD CPUs from taking marketshare before they're even released by taking production capacity away from their rival. They know that for AMD to really damage Intel marketshare and thus their bottom line, AMD needs to produce far more CPUs in the market.

 

[Patriot]

Intel is taking whatever spare wafers TSMC has, that is why they are on 6nm despite its inferiority to other nodes

 

[watzupken]

It is logical that TSMC should not see Intel as a long term client. They are coming to them because they are in trouble for now. So who knows when Intel's 7nm is finally ready, its possible that they may not stick around with TSMC. In this case, there is no point expanding for Intel because they run into the risk of having spare capacity down the road. The uneasy collaboration between Intel and TSMC is likely a win win situation given that TSMC have just lost a major client (Huawei), and thus, the spare capacity can be put to use to "help" Intel.

 

[ODOGG26]

If I were TSMC I would charge a massive premium to Intel per wafer, to really drain that war chest Intel has.

We know Intel's game here; they're trying to stop AMD CPUs from taking marketshare before they're even released by taking production capacity away from their rival. They know that for AMD to really damage Intel marketshare and thus their bottom line, AMD needs to produce far more CPUs in the market.

That wont happen. AMD already secured their orders and much of the spare capacity left from Huawei. They cant pay them enough money to reduce someone elses already secured capacity. TSMC is not stupid and Intel (their rival) is certainly not worth it over a long time customer both present and future.

It is logical that TSMC should not see Intel as a long term client. They are coming to them because they are in trouble for now. So who knows when Intel's 7nm is finally ready, its possible that they may not stick around with TSMC. In this case, there is no point expanding for Intel because they run into the risk of having spare capacity down the road. The uneasy collaboration between Intel and TSMC is likely a win win situation given that TSMC have just lost a major client (Huawei), and thus, the spare capacity can be put to use to "help" Intel.

Its probably not much capacity left for them to be honest. Its been in the news for some time now that AMD snatched up a majority of the left over capacity left by Huawei. All money isn't good money

 

[Dave65]

Poor Intel just can't get a break with all that ADVANCED technology they make

 

[Vayra86]

I'm about halfway through my box of popcorn but it just keeps on giving.

 

[Ashtr1x]

Intel won't make CPUs with TSMC for sure, they poured billions into the Fab expansion and 7nm technology R&D and EUV as well, how can people think Intel will sell their Fabs to a foreign company without pressure, USA has Intel as it's last stand in Semiconductor engineering. Intel squandered it all away for appeasing the Investors and profit % for their DC market monopoly, AMD is starting to chip it away slowly. AMD had less than 2-3% marketshare in the Server market now they hold almost 5%, ARM and others are just 1%. Rest is Intel's arena, their Xeon is already feeling the heat of the 14nm++ ageing vs the AMD's 2x efficiency and latest I/O with Gen 4. Their LGA3647 is already EOL almost now, LGA 4677 and 4189 will take over it's place for the Xeon.

Intel has to stick to their guns which is their Fab technology for their processors all this Xe GPU nonsense is not going to take them anywhere IMHO, AMD could never challenge Nvidia in that even after making R&D and going expensive HBM for their GCN, and Console market. And TSMC is 100% correct, Intel's volume is way too high and their R&D won't be sitting ducks for sure, even Samsung after failing 10nm mainstream attention is trying hard, they already secuire Nvidia's chips on 8nm EUV, perhaps same as N7+ and not waved white flag like GloFo. Intel must not stop their 7nm and 5nm R&D for profit bs demands by investors. Xeon is the main source of money, if Intel loses more money like how Rome started to penetrate into powerful CSPs like AWS it's going to be worse for them, already damage is done.

 

[Patriot]

Intel is taking whatever spare wafers TSMC has, that is why they are on 6nm despite its inferiority to other nodes

It seems unclear what nodes they will be using and what products will be on them ... TSMC said 6nm initially, Intel said TSCM 7nm for intel 10nm products and TSMC 5nm for intel 7nm products... but as both have said this is a stop gap and TSMC will not be making extra capacity for intel... Intel gets what huawei would have, and any other spare capacity.

 

[medi01]

It is worth remembering that Intel's 10nm has higher transistor density than the best 7nm node TSMC has.

 

[Patriot]

It is worth remembering that Intel's 10nm has higher transistor density than the best 7nm node TSMC has.

Intel 10nm is 10% denser than Intel 10nm+ due to dropping of COAG (Contact Over Active Gate) that preventing was one of the things limiting manufacturing ability. Intel 10nm+ most likely not more dense than TSMC's original 7nm nor is it more dense than the follow up 7nm+ 7nmp 6nm but it still has worse yields.

So Intel continues to hold an imaginary density advantage as they have yet to discuss transistor density of 10nm+ as well release ice-lake server chips using it.
Intel seems to be keeping their details rather hidden despite claiming advantage... or rather not mentioning the loosening of specs for 10nm+.

Sauce

Is Intel 10nm really denser than TSMC 7nm?

In a recent article on Wikichip they claim that Intel's 10nm process is denser than TSMC's 7nm. (https://fuse.wikichip.org/news/2408/tsmc-7nm-hd-and-hp-cells-2nd-gen-7nm-and-the-snapdragon-855-dtco/) However, I think there were 2 different Intel 10nm nodes. The first was used on Cannonlake and...

semiwiki.com

7 nm lithography process - WikiChip

The 7 nanometer (7 nm) lithography process is a technology node semiconductor manufacturing process following the 10 nm process node. Mass production of integrated circuit fabricated using a 7 nm process began in 2018. The process technology will be phased out by leading-edge foundries by...

en.wikichip.org

 

[R0H1T]

We don't know what's TSMC best 7nm node, AMD apparently hinted(?) last year that they'd go with 7nm EUV at some point this year but the last official announcement from Lisa SU mentioned enhanced 7nm node IIRC. Also in case you forgot Intel's overambitious density targets on 10nm got them into this mess in the first place, as you'd know high density is a direct enemy to high(er) clocks, this is why IIRC 14nm++ is less dense than 14nm+ & clocks relatively sky high at least compared to AMD.

So Intel continues to hold an imaginary density advantage as they have yet to discuss transistor density of 10nm+ as well release ice-lake server chips using it.

Yeah basically what you said