Bernstein: World needs a TSMC alternative—and this name is best positioned

[Daniel Nenni]

Samsung, not Intel (NASDAQ:INTC), is best placed to emerge as the world’s leading alternative to Taiwan Semiconductor Manufacturing Co., according to Bernstein.

In a new note examining Samsung’s reported $16.5 billion chip deal with Tesla (NASDAQ:TSLA), Bernstein said the partnership reinforces its view that “the world needs a leading-logic semiconductor producer in addition to TSMC, and that Samsung…is better positioned than Intel to be the alternative.”

Bernstein analysts noted that while the Tesla AI6 project might eventually expand beyond automotive use, “the annual revenue for Samsung (KS:005930) will peak at US$2-2.5B and the lifetime contribution will be only ~US$8B.”

The firm said that figure is “much lower than disclosed US$16.5B,” though they noted that Elon Musk has suggested the potential is “several times higher,” prompting speculation about applications in robotics or other devices.

Still, for Samsung’s struggling foundry business, the impact could be transformative, according to Bernstein.

“US$16.5B may lift Samsung Foundry revenue by 30-40% and the benefit to profitability should be much more,” the analysts said, noting the unit is currently running at low capacity, especially at its Texas-based Taylor fab.

On the broader industry impact, Bernstein stated that the project would expand the wafer fab equipment (WFE) market “by LSD% at most.”

It added: “The impact on TSMC is negligible,” estimating that Tesla’s current contribution to TSMC revenue is minimal.

“Samsung, with its comparable technologies, better cost structure and, more importantly, the support from a memory business and hence the ability to sustain investment, is better positioned than Intel,” Bernstein concluded.

https://www.yahoo.com/finance/news/bernstein-world-needs-tsmc-alternative-144305396.html

 

[Y.H]

If Bernstein is right than Lip-bu Tan is correct in shutting down IFS. From what I understand , both TSMC and Samsung does quite a bit of R&D in Asia which is likely to be much more cost effective with lower cost base, lower tax rate , more tax break and more relaxed labor requirements. TSMC might also be sharing its R&D burden with its ecosystem. So they have inherent advantages. Intel on the other hand had none of these and in fact quite the opposite. In that light they had already done quite well. Unless , there is an industry, to be more specific, US semiconductor industry consensus to have a US based second source like the early days of AMD, Lip bu is probably right. I mean some say Intel cannot survive without offering foundry services. It could be. But they won’t immediately die. I think the run rate of remaining cash flow till it disappears will likely have a present value higher than its market value.

 

[siliconbruh999]

Samsung hasn't had a good yielding process since like 8nm? that was the last time i saw them ship large dies in volume.

 

[Y.H]

I think Bernstein argument is the HBM business give significant stamina to Samsung. Foundry. Also having your R&D outside of US has a very big cost advantage and I don’t think the productivity is any lower. Intel should explored that.

 

[Daniel Nenni]

If Bernstein is right than Lip-bu Tan is correct in shutting down IFS. From what I understand , both TSMC and Samsung does quite a bit of R&D in Asia which is likely to be much more cost effective with lower cost base, lower tax rate , more tax break and more relaxed labor requirements. TSMC might also be sharing its R&D burden with its ecosystem. So they have inherent advantages. Intel on the other hand had none of these and in fact quite the opposite. In that light they had already done quite well. Unless , there is an industry, to be more specific, US semiconductor industry consensus to have a US based second source like the early days of AMD, Lip bu is probably right. I mean some say Intel cannot survive without offering foundry services. It could be. But they won’t immediately die. I think the run rate of remaining cash flow till it disappears will likely have a present value higher than its market value.

Personally I do not think IFS will be shut down. Lip-Bu set low expectations and will beat them, that is how he works. It is the complete opposite of Pat Gelsinger.

Samsung has a horrible track record in the foundry business. It is not just a yield problem, it is a company culture problem. If you put those two together it is disastrous. Samsung has been riding the NOT TSMC bus for years now but TSMC has made some changes since CC Wei took over as CEO. The NOT TSMC bus is now a short bus due to deeper customer engagements and yield superiority driven by CC Wei.

The deal with Tesla and Samsung really is a bright moment in the foundry business. I wish it had been with Intel but Samsung is in Texas right down the street from Elon Musk. Samsung is also desperate to get back into the foundry business and has pricing flexibility that Intel does not have. If Samsung was not a memory powerhouse they could not afford to be in the logic foundry business but after 20 years they need to start making some money, right?

 

[Y.H]

Personally I do not think IFS will be shut down. Lip-Bu set low expectations and will beat them, that is how he works. It is the complete opposite of Pat Gelsinger.

Samsung has a horrible track record in the foundry business. It is not just a yield problem, it is a company culture problem. If you put those two together it is disastrous. Samsung has been riding the NOT TSMC bus for years now but TSMC has made some changes since CC Wei took over as CEO. The NOT TSMC bus is now a short bus due to deeper customer engagements and yield superiority driven by CC Wei.

The deal with Tesla and Samsung really is a bright moment in the foundry business. I wish it had been with Intel but Samsung is in Texas right down the street from Elon Musk. Samsung is also desperate to get back into the foundry business and has pricing flexibility that Intel does not have. If Samsung was not a memory powerhouse they could not afford to be in the logic foundry business but after 20 years they need to start making some money, right?

Well. There is definitely a possibility that lip-bu is doing a blood bath here. I certainly hope so. Samsung is about to lose big customers like Google or had they lose already? Personally I think having Elon Musk interfering with your fab is a no no. It would be unfair and perhaps compromising to others potential business. Somehow I just wonder why can’t the industry sort of mandate a second source like during the days where they demanded AMD to be second source supplier of x86 chips? That move help the industry to have the epyc chips of now. Lastly, thought I am not in the industry I do know a bit about research activities in Asia. The grant and licensing is enticing. There are dedicated agency to help you do the heavy lifting , perhaps not at Imec level but they are still credible. Eg. There are some specialized research in silicon photonics etc. Intel should move some R&D out of US.

 

[Daniel Nenni]

Well. There is definitely a possibility that lip-bu is doing a blood bath here. I certainly hope so. Samsung is about to lose big customers like Google or had they lose already? Personally I think having Elon Musk interfering with your fab is a no no. It would be unfair and perhaps compromising to others potential business. Somehow I just wonder why can’t the industry sort of mandate a second source like during the days where they demanded AMD to be second source supplier of x86 chips? That move help the industry to have the epyc chips of now. Lastly, thought I am not in the industry I do know a bit about research activities in Asia. The grant and licensing is enticing. There are dedicated agency to help you do the heavy lifting , perhaps not at Imec level but they are still credible. Eg. There are some specialized research in silicon photonics etc. Intel should move some R&D out of US.

Right, it is about risk. Out of the three foundries who can afford to take the Elon Musk risk? Only Samsung.

 

[Y.H]

Right, it is about risk. Out of the three foundries who can afford to take the Elon Musk risk? Only Samsung.

Anyway I read your article on Intel 3.0. I guessed you have very high confidence in Lip Bu. I would be looking out for your follow up article.

 

[Artificer60]

Personally I do not think IFS will be shut down. Lip-Bu set low expectations and will beat them, that is how he works. It is the complete opposite of Pat Gelsinger.

I'm trying to remember the last time Stacy Rasgon had something positive to say about Intel and coming up short.

Am I to assume from this comment that you believe that IFS will be able to land that big customer in the next 12-18 months under Lip-Bu Tan? Because without staying on the leading edge I don't really see a niche for Intel.

 

[DanX]

Mostly, Asians have a tradition of humility.
They can accept mentally that I’m currently the #3, and that it will take at least 8-10 years to catch up with the leader;
they are willing to endure far greater hardship in order to do so.

Americans, on the other hand, are willing to endure hardship for an audacious goal like SpaceX.

 

[DanX]

Mostly, Asians have a tradition of humility.
They can accept mentally that I’m currently the #3, and that it will take at least 8-10 years to catch up with the leader;
they are willing to endure far greater hardship in order to do so.

Americans, on the other hand, are willing to endure hardship for an audacious goal like SpaceX.

As an example. Lisa Su joined AMD at 2012, AMD overtook Intel at 2022 for the first time.

 

[Y.H]

Mostly, Asians have a tradition of humility.
They can accept mentally that I’m currently the #3, and that it will take at least 8-10 years to catch up with the leader;
they are willing to endure far greater hardship in order to do so.

Americans, on the other hand, are willing to endure hardship for an audacious goal like SpaceX.

I think the biggest issue for Intel , other than their own execution,is they operate in an environment where a lot is promised but very few or anything is delivered and those policies are highly uncertain.
TSMC and Samsung on the other hand had enjoyed decades of reliable, consistent support from their home government. For them , it remains to be seen if these 2 can operate just as well outside of their home countries.
PG seems to bet too much on relying on government support. He is right that Asian government are, to a point , unfairly supporting their enterprises. But he is wrong to expect US could do the same for Intel.
I think Lip-Bu is just being realistic. If the best we can do is be number 3, let’s face it and make do with that, at least for next few years

 

[Barnsley]

Mostly, Asians have a tradition of humility.
They can accept mentally that I’m currently the #3, and that it will take at least 8-10 years to catch up with the leader;
they are willing to endure far greater hardship in order to do so.

Americans, on the other hand, are willing to endure hardship for an audacious goal like SpaceX.

Which particular group of Asians are you talking to?

Asia a big place

 

[hist78]

View attachment 3411

Samsung, not Intel (NASDAQ:INTC), is best placed to emerge as the world’s leading alternative to Taiwan Semiconductor Manufacturing Co., according to Bernstein.

In a new note examining Samsung’s reported $16.5 billion chip deal with Tesla (NASDAQ:TSLA), Bernstein said the partnership reinforces its view that “the world needs a leading-logic semiconductor producer in addition to TSMC, and that Samsung…is better positioned than Intel to be the alternative.”

Bernstein analysts noted that while the Tesla AI6 project might eventually expand beyond automotive use, “the annual revenue for Samsung (KS:005930) will peak at US$2-2.5B and the lifetime contribution will be only ~US$8B.”

The firm said that figure is “much lower than disclosed US$16.5B,” though they noted that Elon Musk has suggested the potential is “several times higher,” prompting speculation about applications in robotics or other devices.

Still, for Samsung’s struggling foundry business, the impact could be transformative, according to Bernstein.

“US$16.5B may lift Samsung Foundry revenue by 30-40% and the benefit to profitability should be much more,” the analysts said, noting the unit is currently running at low capacity, especially at its Texas-based Taylor fab.

On the broader industry impact, Bernstein stated that the project would expand the wafer fab equipment (WFE) market “by LSD% at most.”

It added: “The impact on TSMC is negligible,” estimating that Tesla’s current contribution to TSMC revenue is minimal.

“Samsung, with its comparable technologies, better cost structure and, more importantly, the support from a memory business and hence the ability to sustain investment, is better positioned than Intel,” Bernstein concluded.

https://www.yahoo.com/finance/news/bernstein-world-needs-tsmc-alternative-144305396.html

"Samsung, with its comparable technologies, better cost structure and, more importantly, the support from a memory business and hence the ability to sustain investment, is better positioned than Intel,” Bernstein concluded."

Is it an illusion or just wishful thinking?

1. Comparable technologies — for what?

2. A better cost structure?
Maybe. But Samsung just can’t get the yield right, which makes the so called “better cost” meaningless.

3. Samsung has internal memory manufacturing.
I don’t see any foundry suffering because they don’t have that. In fact, foundries like TSMC and UMC have the freedom to choose the best memory partners for their customers. So what’s the problem?

 

[osnium]

I'm trying to remember the last time Stacy Rasgon had something positive to say about Intel and coming up short.

Am I to assume from this comment that you believe that IFS will be able to land that big customer in the next 12-18 months under Lip-Bu Tan? Because without staying on the leading edge I don't really see a niche for Intel.

I mean Dr. Rasgon has generally been right about Intel.

 

[hist78]

I think the biggest issue for Intel , other than their own execution,is they operate in an environment where a lot is promised but very few or anything is delivered and those policies are highly uncertain.
TSMC and Samsung on the other hand had enjoyed decades of reliable, consistent support from their home government. For them , it remains to be seen if these 2 can operate just as well outside of their home countries.
PG seems to bet too much on relying on government support. He is right that Asian government are, to a point , unfairly supporting their enterprises. But he is wrong to expect US could do the same for Intel.
I think Lip-Bu is just being realistic. If the best we can do is be number 3, let’s face it and make do with that, at least for next few years

"PG seems to bet too much on relying on government support. He is right that Asian government are, to a point , unfairly supporting their enterprise."


In my opinion, Pat Gelsinger misled both the public and himself regarding Asian government subsidies. His typical talking point claims that semiconductor industries in Asian countries, like Taiwan, are getting stronger every day, and that countries such as the PRC and Japan provide substantial subsidies, making fabs there XYZ% cheaper to build.

The sneaky part is that Pat Gelsinger never explicitly explains how or what exactly the Taiwanese government subsidizes for companies like TSMC, UMC, VIS, or PSMC. He simply assumes there are subsidies—but if he actually knew, he could have spelled them out during his years at Intel.

Isn’t it strange that Pat Gelsinger treats TSMC (headquartered in Taiwan) as his biggest foundry competitor, yet doesn’t study Taiwan’s policies, regulations, or strategy? Instead, he cites subsidy practices in the PRC or Japan as justification for demanding similar actions from the US government.

 

[tomatoma]

I think the biggest issue for Intel , other than their own execution,is they operate in an environment where a lot is promised but very few or anything is delivered and those policies are highly uncertain.
TSMC and Samsung on the other hand had enjoyed decades of reliable, consistent support from their home government. For them , it remains to be seen if these 2 can operate just as well outside of their home countries.
PG seems to bet too much on relying on government support. He is right that Asian government are, to a point , unfairly supporting their enterprises. But he is wrong to expect US could do the same for Intel.
I think Lip-Bu is just being realistic. If the best we can do is be number 3, let’s face it and make do with that, at least for next few years

Well, it's been said for a long time that Foundry takes time to see the results.

 

[Artificer60]

I mean Dr. Rasgon has generally been right about Intel.

No, with Stacy Rasgon, this is his M.O. where Intel is concerned. This is from a 2012 news report.

Intel Corp cut its third-quarter revenue estimate more than expected on Friday due to a decline in demand for its chips as customers reduce inventories and businesses buy fewer personal computers.

Bernstein analyst Stacy Rasgon said the size of the Intel cut was surprising and was a worrying new sign that the company is seeing weakness in PC sales to businesses and governments, known as enterprise sales.

"They also have weakness in enterprise PCs in emerging markets. In the last six to eight quarters, consumers have been weak but the enterprise was strong. Now the enterprise is weak," Rasgon said.

Gelsinger was absolutely right in calling him a "perma-bear". He consistently rates Intel more harshly than other analysts, so yeah he looks good now, but go back more than 15 years and not so much.

 

[MKWVentures]

I would be nice for customers if there was a leading foundry like TSMC. But TSMC moved ahead based on skill and execution, So you cannot mandate a competitor. Everyone is free to choose Intel (no one chose them) or Samsung or UMC or GF. TSMC is a much better company. Just like Nvidia is a much better computer company.

Intel is fully enabled to be a competitor as is Samsung. UMC and GF could use some help to pay for leading nodes but they probably will not execute much better.... plus the equipment ecosystem likes one leader.

If a company is just outperforming everyone, they are not going to have competition.

 

[Y.H]

I would be nice for customers if there was a leading foundry like TSMC. But TSMC moved ahead based on skill and execution, So you cannot mandate a competitor. Everyone is free to choose Intel (no one chose them) or Samsung or UMC or GF. TSMC is a much better company. Just like Nvidia is a much better computer company.

Intel is fully enabled to be a competitor as is Samsung. UMC and GF could use some help to pay for leading nodes but they probably will not execute much better.... plus the equipment ecosystem likes one leader.

If a company is just outperforming everyone, they are not going to have competition.

I get your point. But I remember reading about there was a time the industry ( or is it IBM)mandating a second source x86 manufacturer and that’s AMD. I mean if the second source weren’t mandated, we would probably have a lot less choice now.