News Ex-Intel CEO Gelsinger warns TSMC's $165B investment will not restore U.S. semiconductor leadership

[helper800]

Pat served for 3 years and 10 months and promised "5 nodes in 4 years", and now he asking for more time? He needs to be held to his timelines that he declared.

Yeah too bad the board was fighting him the whole way through purposefully obstructing him at times.

 

[phead128]

Yeah too bad the board was fighting him the whole way through purposefully obstructing him at times.

They typically do that after blowing 60% of stock value and cancelling nodes.

 

[helper800]

They typically do that after blowing 60% of stock value and cancelling nodes.

That was happening as a downward trend from way before his time. I really dont care too much to be honest, thats just my opinion.

 

[bit_user]

While TSMC's R&D center is certainly planned in the U.S., its focus remains to be seen. TSMC has developed its fabrication processes in Taiwan for decades, although many of its engineers come from the U.S. As manufacturing technologies get more complex and require longer pathfinding processes with each generation, TSMC may offload part of its R&D operations from its Taiwan facilities to the U.S. facility.

R&D can mean a lot of things. As mentioned in the article, it can mean refinement of existing nodes. However, I'd suggest they might stretch the definition to include one of gloried tech support. Basically, helping US-based customers adapt, prepare, and optimize their designs for fabrication in TSMC's factories. This is certainly adding value, but it's not the kind of fundamental development work which is needed for the US to be self-sufficient.

IMO, there's no way either TSMC or Taiwan will allow the US operations to have a chance at becoming truly independent. That means if the US gets cut off from Taiwan, any fabs we have here will only be capable of continuing to produce existing nodes, at best.

 

[bit_user]

The only way for the US to have leadership in the manufacturing of semiconductors is to have a US based company have that leadership. It doesn't have to be Intel, but who else is there?

I'd say don't forget about Micron. TBH, I don't actually know where most of their R&D happens. Also, they don't seem a whole lot further ahead than Global Foundries and really specialize in memories. So, that does pretty much just leave Intel in the leadership position for logic.

 

[bit_user]

All the comments about him merely being self-serving and also critical of his new role are irrationally dismissive. if you've followed him since before he was CEO, his role now is very much in line with past comments.

I think some of his comments are more self-serving than others. With the exception of the one where he called Nvidia "lucky", I generally see them as trying to support Intel staying on the path he believes is right. A comment like this seems clearly aimed at trying to save CHIPS, for instance.

Why is Gelsinger still talking so much? Is he high on himself?

I think The Financial Times (which is sort of like Britain's equivalent of The Wall St. Journal) probably asked for an interview and he obliged. You're certainly not obligated to care what he has to say.

His opinion would have more weight if he had managed to turn Intel around, during his tenure. I had high hopes he would turn Intel around, given his background, to keep competition alive and healthy, but he failed to do so. He was all talk.

That's a little too harsh, IMO. I think it wasn't possible to do a full turn-around, in that amount of time. What I think was realistic is for them not to have had quite so many seemingly-avoidable failures and for Pat to be more realistic in the kinds of promises and spin he put on Intel's situation.

If 18A gets serious and growing wins and IFS takes off over the next few years, should Tan get all the credit? Will people remember Pat's name? Outside of some in forums like this, no, they will remember Tan's. Yet Pat is rooting for that. Not seeing magnanimity in that treads beyond mere cynicism.

The engineers should get the lion's share of the credit, but I'll bet none of us know any of their names.

Pat was compensated very well and surely still owns quite a bit of Intel stock. If the share price goes up significantly, I think that should be more than enough reward for his efforts.

 

[bit_user]

If Pat actually looked at the distribution of labs vs fabs in the U. S., he would know the U.S. have significantly more research labs/fabs than production fabs, the U.S. lost manufacturing leadership to Taiwan and others, however the U.S. still has a strong R&D leadership and invents the majority of chip manufacturing technologies, including the science and technology that enabled ASML to bring Extreme Ultraviolet Lithography to the market.

It's not enough to simply have part of the supply chain. A crucial piece is the one which happens at the foundries. If you don't have that, it doesn't matter how much other R&D you do, you still cannot make leading-edge chips.

Just look how hard it's been for Intel to regain leadership. Do you seriously think that if China locked up access to TSMC's new nodes and Intel had to cease development of new nodes, that we could just turn the key and rev back up to a leadership position overnight? I see no evidence that would be possible, or else why didn't Intel have 18A ready last year? If they had the equipment needed for 20A, they should've been able to do 18A product, no? So, likely it wasn't a matter of simply not having the equipment.

Node development takes a lot of time, resources, and expertise! Consider the > 30k wafers they've gone through on high-NA, which isn't yet even production-ready!

https://www.tomshardware.com/tech-i...0-000-wafers-with-high-na-euv-chipmaking-tool
​

 

[TerryLaze]

It's not enough to simply have part of the supply chain. A crucial piece is the one which happens at the foundries. If you don't have that, it doesn't matter how much other R&D you do, you still cannot make leading-edge chips.

Just look how hard it's been for Intel to regain leadership. Do you seriously think that if China locked up access to TSMC's new nodes and Intel had to cease development of new nodes, that we could just turn the key and rev back up to a leadership position overnight? I see no evidence that would be possible, or else why didn't Intel have 18A ready last year? If they had the equipment needed for 20A, they should've been able to do 18A product, no? So, likely it wasn't a matter of simply not having the equipment.

Node development takes a lot of time, resources, and expertise! Consider the > 30k wafers they've gone through on high-NA, which isn't yet even production-ready!

​

https://www.tomshardware.com/tech-i...0-000-wafers-with-high-na-euv-chipmaking-tool​

​

The US doesn't need leadership, they want it for e-peen but they don't need it.
As long as they can produce technology that is usable it's fine.
Just look at other countries, russia just made 30 year old tech work and that would be enough for them to at least keep going in case they would get shut off from everybody.
Intel 7 would we usable for plenty of years to come even if they would completely fumble on 18A. It wouldn't if they had to be competitive and sell them but "US being self sufficient" isn't about selling, it's about being able to make the electronics the country needs to work.

 

[bit_user]

The US doesn't need leadership, they want it for e-peen but they don't need it.

For some definition of "need", yes it does. The USA certainly needs continued access to current nodes. I think we could make do with that, but we honestly have no idea whether TSMC's US-based fabs could continue operating, if they got cut off from the mothership. I somewhat doubt it.

Just look at other countries, russia just made 30 year old tech work and that would be enough for them to at least keep going in case they would get shut off from everybody.

Heh, no. Russia can't get by with 30-year-old tech. Basically, no country can. All countries' businesses and economy depends on machines and infrastructure that probably can't be sustained by anything older than maybe 10 years, max. 30-year-old fab tech is certainly good enough for some peripheral chips and industrial automation, but that's about it.

Intel 7 would we usable for plenty of years to come even if they would completely fumble on 18A.

I pretty much agree that 7nm-class tech would probably be good enough to keep the country from grinding to a halt. Being limited to that would certainly trigger an economic downturn, as AI is already yielding measurable productivity gains. Going back to 7 nm-class tech would mean GPUs like RDNA1 and Nvidia's A100.

 

[TerryLaze]

For some definition of "need", yes it does.

Yes, with some definition of any word you get any other word.....
The words needs and wants are well defined even in american english.

The danger of being cut off from outside tech is total collapse of society and to prevent that you just need enough to keep your military government and infra structure working.

What the US (or any country) wants is to have business go on as usual.
One is the need the other is the want.

 

[bit_user]

The danger of being cut off from outside tech is total collapse of society and to prevent that you just need enough to keep your military government and infra structure working.

Just keeping the economy from complete meltdown requires reasonably modern tech. Like I said, certainly within the last 10 years. Anything less and lots of businesses and industries would no longer be able to function as they do. If we had to go back to 1990's era tech, there'd be mass disruption and displacement, as entire industries completely failed. I classify that as a need.

For others, like politicians, they need even less disruption to the economy. For that, it definitely takes continued access to current nodes, at a bare minimum.