Why is Trump Targeting TSMC? Expert Reveals the Hidden Agenda: "Saving Intel"

[XYang2023]

https://twitter.com/x/status/1890117529345094123

 

[jabo]

I think you will start to hear the reality of what is going on at Intel foundry now that Pat is gone. No customers, Costs way to high, SCIP partnerships becoming problematic, Intel 3/4 losing money, 18A planning to lose money if they build fabs and lose money if they dont build fabs. There is a reason that Intel BOD is looking for options

Pat has been gone for over 2 months now. Yet, the message about Foundry has gotten more encouraging, not worse. The co-CEOs have stated countless times at conferences and earnings calls that 18A is on track to deliver on Panther Lake in the 2nd of half of this year. On top of that, one of the co-CEO's, MJ, has stated she wants to improve Intel's "do to say ratio". She has stated she aims to only make claims she knows will happen. She seems like a straight shooter and not one to sugar coat. Other than some misleading (maybe on purpose) news about Broadcom having 18A yield issues, I don't see any indication that 18A is in trouble. The Broadcom thing was likely due to them having monster size die. I have seen no indication that Intel 3 or 4 are losing money, at least not due to being low yield. Foundry is in general losing money, but they are new, ramping, and building out capacity, and need to bring in customers. They have stated they expect break even in 2027. So we still have 2 to 2.5-ish years left to break even.

Bringing wafers back to Intel Foundry as opposed to outsourcing to TSMC, helps T getting to break even. The only change I see from new leaders is that Foundry capex has been slightly lowered and they will be more conservative and slower with build aheads. THe bottom line is that Intel doesn't need help or "know how" from TSMC. From a technological perspective Intel may in fact be ahead. Intel simply needs a jump start to bring on customers. Customers are happy with TSMC and do not want to risk things. Why would they? But it's not an all of nothing scenario. Nvidia or Apple wouldn't have to ditch TSMC overnight. They could build some small percentage of chips with Intel and slowly increase this. Tarrifs, CHIPs and the current political environment certainly seem they will help enticing some to try out Intel. Intel Foundry does in fact have at least two high profile 18A customers: MSFT and Amazon. Not sure of the monetary value of those deals, but it gets the pump primed and (hopefully) will show the world Intel Foundry can be a truster partner. I believe MediaTek is also a customer and this was announced a very long time ago. This could all change suddenly. All it would take is for QCOM, Nvidia, Apple or similar to make an annoucement and Intel would be the new darling.

 

[Mooredaddy]

https://twitter.com/x/status/1890117529345094123

he's an idiot. What do you expect

 

[Jure]

I think you will start to hear the reality of what is going on at Intel foundry now that Pat is gone. No customers, Costs way to high, SCIP partnerships becoming problematic, Intel 3/4 losing money, 18A planning to lose money if they build fabs and lose money if they dont build fabs. There is a reason that Intel BOD is looking for options

Semiaccurate/ does not agree about the process cost, but TSMC does have huge scale advantage. SA hints at pricey packaging.

 

[Daniel Nenni]

The tone of recent new articles about this topic are insulting and demoralizing for Intel Foundry. The articles speak of TSMC engineers sharing their "know how" and helping Intel on 3nm and 2nm nodes. They claim Intel would be the JV player in the merger and that TSMC would control operations. Intel doesn't need help with 3nm or 2nm nodes. They have EUV up and running just fine. Intel 3 has been in produciton for quite some time and is shipping Xeon chips. Intel 18A is on track to be used on Panther Lake later this year. Remember, TSMC N2 isn't up an running yet either. Recent articles on Tom's Hardware and other places show Intel 18A outperforming TSMC N2. Not only that, Intel 18A has backside power. Who would be helping who exactly? To top it all off, Intel Foundry has at least a year headstart on high NA EUV. They have at least two high NA machines up and running. Seems to me Intel Foundry would be helping TSMC. How would any of this ever be allowed anyways? Intel Foundry wasn't even able to acquire Tower. Now people really think Intel's just going hand over its Foundry operations to a foreign comapny after spending $100B and 4 years to get in the game? This is crazy.

I agree except for the part about Intel 18A outperforming TSMC N2, it does not. Based on the hype it might but not based on the PDKs and test chips. Intel 18A is faster but TSMC N2 is more dense, lower power, and less expensive to manufacture. The margins however are not that big so I would put them as competitive. Samsung 2nm might be in the game as well based on the hype but I have not seen the PDK in the wild yet.

The problem is that Intel Foundry comes with more risk than TSMC which is a big deal to large customers. Hopefully Intel can reduce that risk before the next design cycle starts but as it is today TSMC N2 will win the node, my opinion. It will not be another landslide win like TSMC N3 but a majority win just the same.

 

[Daniel Nenni]

Semiaccurate/ does not agree about the process cost, but TSMC does have huge scale advantage. SA hints at pricey packaging.

Surely TechInsights (Scotten Jones) has better cost information than Semiaccurate (a rumor site).

 

[kevin01]

I agree except for the part about Intel 18A outperforming TSMC N2, it does not. Based on the hype it might but not based on the PDKs and test chips. Intel 18A is faster but TSMC N2 is more dense, lower power, and less expensive to manufacture. The margins however are not that big so I would put them as competitive. Samsung 2nm might be in the game as well based on the hype but I have not seen the PDK in the wild yet.

The problem is that Intel Foundry comes with more risk than TSMC which is a big deal to large customers. Hopefully Intel can reduce that risk before the next design cycle starts but as it is today TSMC N2 will win the node, my opinion. It will not be another landslide win like TSMC N3 but a majority win just the same.

Given the N3 adoption timeline as a guide, it appears N2 and 18A won't be widely adopted for at least 4-5 years. Therefore, potential customers likely have the time to evaluate the performance of leading products built on these processes and also assess their respective costs. A wild card is tariff, which could significantly influence those cost considerations.

By the way, I used to be against tariffs, but I've come to see them in a different light. It seems to me that, in Trump's perspective, tariffs serve two main purposes: 1) raising revenue to allow for income tax cuts, and 2) revitalizing key industries. Both purposes seem like they could have positive outcomes.

 

[XYang2023]

 

[Daniel Nenni]

Given the N3 adoption timeline as a guide, it appears N2 and 18A won't be widely adopted for at least 4-5 years. Therefore, potential customers likely have the time to evaluate the performance of leading products built on these processes and also assess their respective costs. A wild card is tariff, which could significantly influence those cost considerations.

What do you mean by "widely adopted" ? I count design starts. N2 designs have already started, a couple have even taped out. We will not see N2 production start until Q4 but N2 designs are in definitely in progress. It takes 2-3 years to design big chips so I would not say at least 4-5 years from now. Apple and Qualcomm will be first for N2 from what I hear.

 

[Mooredaddy]

What do you mean by "widely adopted" ? I count design starts. N2 designs have already started, a couple have even taped out. We will not see N2 production start until Q4 but N2 designs are in definitely in progress. It takes 2-3 years to design big chips so I would not say at least 4-5 years from now. Apple and Qualcomm will be first for N2 from what I hear.

N2 is an outstanding node from what I hear. 18A we've heard about for years and still nothing concrete. And now the U.S gov wants TSMC to share tech with Intel? TSMC should tell this admin to get bent.

 

[Daniel Nenni]

Intel does not have American AI? What does that even mean?

Does Intel really have most of their manufacturing in the US? What about Ireland and Israel? Maybe when Intel builds out Ohio but I do not think that statement is true today.

 

[Mooredaddy]

Pat is a cheerleader, He is not pragmatic enough to be CEO. He is the exact opposite of Andy Grove. As CTO, he was cheerleading. He is good for providing vision when the vision is correct.

Intel has serious financial issues thanks to going big on foundry and foundry not working as predicted. Pat bet on the Patriots (Intel) to win it all based on their long history.... probably should have looked at recent history instead.

The government should not get involved IMO. that will not help the problem. The Government should partner with the Eagles, not the Patriots.

The Patriots may be great again someday. Today is not that day.

[multiple mixed metaphors and some weak analogies LOL]

A TSMC JV is perfect solution for Intel. Lets hope it happens

Why would TSMC ever agree to help incubate a competitor and share technology at that. I hate the idea

 

[XYang2023]

Intel does not have American AI? What does that even mean?

Does Intel really have most of their manufacturing in the US? What about Ireland and Israel? Maybe when Intel builds out Ohio but I do not think that statement is true today.

The media just doesn't know. Pat should add more colour to that.

https://twitter.com/x/status/1890157060740076024

Also what I wrote previously:

Intel will ship the world's first "2nm class" GPU?

Intel will introduce Panther Lake on the 18A "2nm class" process node featuring the next-gen Celestial (Xe3) GPU with a purported 12 Xe Cores (current top Battlemage, the discrete B580, has 20 Xe Cores). With the latest quarterly earnings update, Intel confirmed PTL for 25H2 "release" though I'm...

semiwiki.com

 

[kevin01]

What do you mean by "widely adopted" ? I count design starts. N2 designs have already started, a couple have even taped out. We will not see N2 production start until Q4 but N2 designs are in definitely in progress. It takes 2-3 years to design big chips so I would not say at least 4-5 years from now. Apple and Qualcomm will be first for N2 from what I hear.

By wide adoption, I meant large volumes of real products on the market. Earlier in this thread, someone counted N3 products on the market now, still not that many, more than 2 years after N3B entered high volume production for Apple.

And when will N2 start high volume production? Probably in the end of 2025. If the same timeline follows, N2 won't get into its prime time anytime soon.

 

[Mooredaddy]

By wide adoption, I meant large volumes of real products on the market. Earlier in this thread, someone counted N3 products on the market now, still not that many, more than 2 years after N3B entered high volume production for Apple.

And when will N2 start high volume production? Probably in the end of 2025. If the same timeline follows, N2 won't get into its prime time anytime soon.

Kevin you really don't know what you are talking about so I suggest you stop . Look at the N3 revenue numbers

 

[kevin01]

Kevin you really don't know what you are talking about so I suggest you stop . Look at the N3 revenue numbers

Yes, revenue wise, apple alone could be really huge. But that revenue number could be even higher in the coming years, as more customers start adopting N3.

 

[Mooredaddy]

Yes, revenue wise, apple alone could be really huge. But that revenue number could be even higher in the coming years, as more customers start adopting N3.

Companies pay in advance. Don't think of it as launched products. I can assure you nvidia is probably already paying for N3 in some capacity even as blackwell is on N4

 

[kevin01]

Companies pay in advance. Don't think of it as launched products. I can assure you nvidia is probably already paying for N3 in some capacity even as blackwell is on N4

I acknowledge your points. However, we seem to have differing views on "wide adoption" and "prime time." Should subsequent process nodes, such as A16, encounter difficulties, N2's lifespan could be considerably extended. As it stands, N2 is far from its prime.

 

[jabo]

I agree except for the part about Intel 18A outperforming TSMC N2, it does not. Based on the hype it might but not based on the PDKs and test chips. Intel 18A is faster but TSMC N2 is more dense, lower power, and less expensive to manufacture. The margins however are not that big so I would put them as competitive. Samsung 2nm might be in the game as well based on the hype but I have not seen the PDK in the wild yet.

The problem is that Intel Foundry comes with more risk than TSMC which is a big deal to large customers. Hopefully Intel can reduce that risk before the next design cycle starts but as it is today TSMC N2 will win the node, my opinion. It will not be another landslide win like TSMC N3 but a majority win just the same.

We will have to wait and see how they each perform. The Tom's Hardware article has many caveats. As one might expect, TSMC and Intel are both claiming leadership here. It looks like 18A will have products on the market faster (assuming Panther Lake really hits retail stores in late 2025)....so at least for a few months Intel can claim leadership. In general, I am just happy Intel is back in the game. I'm sure over the next decade they will make for some exciting competition...with competition we all win.

 

[Maximus]

By wide adoption, I meant large volumes of real products on the market. Earlier in this thread, someone counted N3 products on the market now, still not that many, more than 2 years after N3B entered high volume production for Apple.

And when will N2 start high volume production? Probably in the end of 2025. If the same timeline follows, N2 won't get into its prime time anytime soon.

Apple: A17/17Pro (N3B), A18/18Pro (N3E), M3 series (N3B), M4 series (N3E)
Qualcomm: Snapdragon 8 Elite (N3E)
MediaTek: Dimensity 9400 (N3E)
AMD: Zen5c ccd (N3E)
Intel: Arrow Lake, Lunar Lake (N3B)
----
Not many??