Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/cc-wei-discussed-the-challenges-of-establishing-tsmcs-foundry-in-the-us.21892/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

CC Wei discussed the challenges of establishing TSMC's foundry in the US

Maximus

Well-known member
CC Wei attended the National Taiwan University EMBA Baichuan Lecture tonight and gave a speech titled "If there was no TSMC." His witty and humorous speech not only made the audience laugh, but also made the participants understand through a 30-minute speech why TSMC's latest process cannot be moved to other countries and the world. If there is no TSMC, not only will people not be able to replace their Apple phones, but even refrigerators, televisions, cars and other products will be affected.

Since TSMC officially set up a factory in the United States in 2021, many people in the market have been worried about whether TSMC will bring the latest technology to the United States, making Taiwan no longer have an advantage.

Before answering the market's concerns, CC Wei first talked about, "If you want to repeat the TSMC, how long will it take? Not long, you can build a production line in 2.5 years." However, this 2.5 years is estimated with prerequisites, "there must be enough land, water, electricity, and talent. If you don't have the technology, that's another matter."

2.5 years is just the time it takes to build a production line, but every time a new technology emerges, TSMC has to build at least three factories. CC Wei took 3nm as an example. "It will take about 5 years to build a 3nm plant to meet all customer needs. If you want to move TSMC's 3nm to any country, assuming you have good land, water and electricity, There are no problems with the talent. It will only take 5 years. "

However, the progress of the construction of the US factory has not met TSMC's original expectations. CC Wei said with a smile, "We were in Arizona, I was crying and going through all kinds of training." He said it humorously, and the whole audience burst into laughter, but he pointed out three challenges that outsiders would not imagine.

"First, we thought that the United States is so big, building a house is not a big deal. Wrong, building a house is a big deal. I want to build a production line in Arizona. First, they have never heard of what a production line is. I said, Isn't there an IDM (referring to IDM giant Intel) next door? He said, that was ten years ago, now everyone has left, so they won't build one. "CC Wei said that the first problem is the local infrastructure. The thing is, they didn't know how to build a factory, and in the end half of the construction workers were recruited from Texas.

Second, I never thought that in one place, I would encounter different laws and regulations from the local government. "TSMC is a company that is well taken care of by the government. We are ahead of the curve and have built various production lines. The government doesn't know what to do. We discuss with the Taiwanese government how to set the regulations." However, after arriving in the United States, the US government also don’t know how to make regulations, but the US government told TSMC, “Please pay first, and we will use your money to hire experts to make rules. In the end, every time we have a meeting, city government officials and our experts hired by the company and TSMC, who then spoke out and worked together to formulate the rules, and only after they were formulated did we move forward. In total, we formulated more than 18,000 rules, and it cost 35 million US dollars.

CC Wei said with a smile, "I'm not complaining, I'm just sharing my experience." He then mentioned the problem of chemicals in the United States. The United States cannot produce the level that TSMC needs. If it can be produced, the price will be 5 times higher than what TSMC originally purchased. "We ended up shipping sulfuric acid from Taiwan to the Port of L.A., and then trucking it from the port to Arizona. Even this was cheaper than doing it in the U.S."

But Intel, a giant American semiconductor manufacturer, has a factory next to TSMC. So, how does Intel deal with the problem of chemical materials? CC Wei said, "They saw this as a good deal, so they also started buying from Taiwan. It's hard to believe when you hear this, and you ask them, how did you live in the past? They say, they had a hard life before, but now they have you."

The second problem is related to the labor union. CC Wei said, "The establishment of the labor union was a noble ideal to help workers fight for better benefits, but later, the union became very strange." It is obvious that TSMC has also encountered labor-capital conflicts in the local area.

The third problem is related to the speed of local government administration. "TSMC has made such rapid progress because it can build a production line in two and a half years. However, in these two and a half years, our technology has been improving. Every time the technology is improved, the entire production line design had to be changed, including the water, electricity, and chemical lines under the production line. Every time a change was made, we worked with the manufacturer and did it immediately. But in the United States, it was not possible. Any change had to be stopped and fixed first. Once the drawing is done, we can proceed with the work after the application is approved.”

At this rate, CC Wei bluntly said, "Do you think it is possible for the latest technology to go to the United States? Just applying for everything and then coming in will take at least twice as long as Taiwan." Obviously, for TSMC, the latest technology will only remain in Taiwan and will be difficult to move to other countries.

 
Kind of embarrassing how little homework was done and also claiming things that aren't true. As an example Fujifilm is right in Phoenix and makes a huge variety of semiconductor grade chemicals for all manner of local customers. Now if TSMC didn't want to qualify another supplier (be it because it would be too much effort or out of a desire to be CE! with fab 18), or a desire to only buy Taiwanese when possible is their business. Another is the nonsense that all of the IDMs left Phoniex and that there is no talent. ON, NXP, Microchip, infineon, and Intel. Intel even had built a new Arizona expansion back in the 2010s and filled it with tools in that 2019-2020 timeframe. You also don't see Intel delaying fab42 or fab52. One would hope TSMC is taking these issues seriously going forward and takes a good hard internal look at what needs to be improved. Between some of their interactions with tool vendors and how they talk publicly you can certainly see plenty of arrogance (which I suppose isn't exactly new) and some level of compliancy setting in (which does seem new and is the most dangerous thing of them all).
 
Kind of embarrassing how little homework was done and also claiming things that aren't true. As an example Fujifilm is right in Phoenix and makes a huge variety of semiconductor grade chemicals for all manner of local customers. Now if TSMC didn't want to qualify another supplier (be it because it would be too much effort or out of a desire to be CE! with fab 18), or a desire to only buy Taiwanese when possible is their business. Another is the nonsense that all of the IDMs left Phoniex and that there is no talent. ON, NXP, Microchip, infineon, and Intel. Intel even had built a new Arizona expansion back in the 2010s and filled it with tools in that 2019-2020 timeframe. You also don't see Intel delaying fab42 or fab52. One would hope TSMC is taking these issues seriously going forward and takes a good hard internal look at what needs to be improved. Between some of their interactions with tool vendors and how they talk publicly you can certainly see plenty of arrogance (which I suppose isn't exactly new) and some level of compliancy setting in (which does seem new and is the most dangerous thing of them all).

No delays in F42? 🤔
Started in 2011 and fully operational in 2020. Hardly light speed. Fabs in Taiwan can go from shovel to silicon in half the time it takes in AZ.
 
No delays in F42? 🤔
Started in 2011 and fully operational in 2020. Hardly light speed. Fabs in Taiwan can go from shovel to silicon in half the time it takes in AZ.
F42 was finished long before sitting empty because BK was too cheap to fill it with equipment and would rather outsource chipsets to Samsung when intel started to run out of 14nm capacity. Per Intel, it was built to be ready for the 2014 14nm ramp; but was of course never actually filled until 2019 to support the 2020 TGL launch.
 
I lost interest in everything Wei said after this statement. TSMC had built TSMC Washington (was WaferTech) decades ago, and TSMC doesn't know that building a house in the US is a major PITA? Seriously?
Debby downer crew you two always. Why are you two always SO down on TSMC. C.C is clearly being flippant in an informal talk. There is nothing to suggest TSMC is becoming complacent. Rather they are consistently executing at a high level in the most complex business there is. Have there been stumbles in Arizona? Surely but getting mass production up and going by the end of 2024 is pretty remarkable considering how this is TSMC's first real step out of Taiwan at scale. They got Japan done on time, Arizona now up and humming and starting in Germany soon. All while executing at the highest level financially and technically.
 
Debby downer crew you two always. Why are you two always SO down on TSMC. C.C is clearly being flippant in an informal talk. There is nothing to suggest TSMC is becoming complacent. Rather they are consistently executing at a high level in the most complex business there is. Have there been stumbles in Arizona? Surely but getting mass production up and going by the end of 2024 is pretty remarkable considering how this is TSMC's first real step out of Taiwan at scale. They got Japan done on time, Arizona now up and humming and starting in Germany soon. All while executing at the highest level financially and technically.
Can we exclude Japan Cause Japanese and Taiwan's Work Culture are pretty similar that plus the support from government
 
F42 was finished long before sitting empty because BK was too cheap to fill it with equipment and would rather outsource chipsets to Samsung when intel started to run out of 14nm capacity. Per Intel, it was built to be ready for the 2014 14nm ramp; but was of course never actually filled until 2019 to support the 2020 TGL launch.
A delay in getting equipment in for 6 years is still a delay. Tools depend on what's being produced and Intel wasn't set on what to produce at Fab 42. If TSMC let the Phoenix plant shell sit empty for 6 years, you'd call it a delay.
I lost interest in everything Wei said after this statement. TSMC had built TSMC Washington (was WaferTech) decades ago, and TSMC doesn't know that building a house in the US is a major PITA? Seriously?
Exactly, the last time TSMC built something in the US was 25+ years ago. Semiconductor manufacturing best practices have changed quite a bit in that time. No need to be so dismissive; he's obviously just trying to tell a story here.
 
Debby downer crew you two always. Why are you two always SO down on TSMC.
I don't see why you would say that. I have never discredited TSMC's technical or service capability. They simply have no peers on this full spectrum capability. They have a sterling reputation worth more than gold. Total capacity only surpassed by Samsung. A high degree of efficiency, and they run one mean high performance manufacturing org. But just because all of that is true, doesn't mean I will pretend everything is sunshine and rainbows when it ain't. A learning organization needs to recognize when something can be improved and improve it. This sort of thinking is what separates a good process engineer from a mediocre one. Saying "TSMC perfect, they do nothing wrong" is A) not true, and B) not constructive.
C.C is clearly being flippant in an informal talk.
So it is ok for CC but not for others? Of course, the answer is no. It is stupid when others do it, and it is equally stupid when CC is flippant or says things that are widely known to be factually incorrect. It is also TSMC who officially that talks all this big game about how "nobody can match them on fab construction speed" or how their logic fabs are "2x the size of the average logic fab". Even in Taiwan, this isn't really true (see Micron/UMC/Samsung for point 1 or see every advanced logic fab built in the past 15 for point 2). But whatever; marketing will do what marketing does.
There is nothing to suggest TSMC is becoming complacent.
You should talk to the engineers at the tool vendors then. One counterpoint against TSMC's starting to become more complacent is that they do seem to take intel foundry with a level of seriousness that feels appropriate. But either way, I don't think they are anywhere near the complacency of old intel. Not even by a country mile. But the only company I have ever seen with that level of pure concentrated arrogance is NVIDIA (but at least over the past 10-20 years, I suppose they at least have had the chops to earn it). Even though, any complacency at TSMC seems to be mild. I see hints of it in how they work/talk to and about others (be they OSATs, IDMs, competitor foundries, or their vendors). A little is no big deal, but left unchecked it can become a problem (see Intel). But who knows, maybe I'm wrong, and it is fine as long as it doesn't start leaking into how they interact with their customers.
Rather they are consistently executing at a high level in the most complex business there is. Have there been stumbles in Arizona? Surely but getting mass production up and going by the end of 2024 is pretty remarkable considering how this is TSMC's first real step out of Taiwan at scale. They got Japan done on time, Arizona now up and humming and starting in Germany soon. All while executing at the highest level financially and technically.
All of the above is all fine and dandy; but that isn't my point. For one, the standard TSMC needs to meet is higher. Meeting and exceeding high standards is how you build the rapport that TSMC has, and how you get to market caps like TSMC has. Failing to meet them is how you build a reputation like intel or Samsung. Two, my concern is not so much that AZ had stumbles but TSMC seemingly blaming everyone else but themselves and seemingly refusing to learn from their mistakes. Blaming others rather than looking at what YOU could have done better is how you stay stagnant. Three, blaming others is not how TSMC acts with its customers. Even if TSMC has a tool error that is due to some hardware flaw or software bug from their vendor, and this scraps a wafer. Does TSMC tell their customer "Well it was AMAT's fault your wafer was scraped, don't blame us!"? No TSMC tells their customer "This is OUR fault and OUR problem. WE will fix this issue and WE will make you whole.". Meanwhile, in the background, their engineers would work with vendor engineers and make sure this never happens again. Either by driving a root cause solution, or changing to a vendor who will not have said problem. In the foundry business, nobody wants or accepts excuses. You are trusted with a design house's livelihood, and they expect first time right.
A delay in getting equipment in for 6 years is still a delay. Tools depend on what's being produced and Intel wasn't set on what to produce at Fab 42. If TSMC let the Phoenix plant shell sit empty for 6 years, you'd call it a delay.
No I wouldn't. If the fab was done, but TSMC willingly chose to not fill it with equipment, the fab is still "done". Were the later phases of fab 18 delayed because N3 was late? No, the process was late, and they had nothing to run. Def not an issue with the construction team. Another example is GF 7LP. All indications are the process was fine and healthy, but there was insufficient demand or capital to ramp it; so GF canceled it. Assuming the process was healthy, I don't know why I would call the process "late" or "delayed". It never came to production, but that was by choice/intended, whereas a delay implies something unintentional/unscheduled.
Exactly, the last time TSMC built something in the US was 25+ years ago. Semiconductor manufacturing best practices have changed quite a bit in that time. No need to be so dismissive; he's obviously just trying to tell a story here.
Yes, but the regulatory situation in the US isn't fundamentally any different. Permitting is more or less the same. It should come as a shock to absolutely nobody that AZ/the FED won't bend over backwards for TSMC the same way the Taiwanese federal government does. In the US, TSMC is a small employer in a VERY big job market. Over there they dictate the terms, over here Uncle Sam holds all of the cards. This was the case 25 years ago, 50 years ago, and 100 years ago. Additionally, any learning organization would have worked to gather BKMs from someone already in the region or get a JV partner (See Toyota entering the US manufacturing market, or TSMC in Japan). His "story" also just has factually incorrect information about what the semiconductor infrastructure is like in AZ. For Pete's sake, every major tool vendor has R&D and or customer engagement sites down there already with field service and process engineers holding decades of semiconductor experience.
 
Last edited:
Kind of embarrassing how little homework was done and also claiming things that aren't true. As an example Fujifilm is right in Phoenix and makes a huge variety of semiconductor grade chemicals for all manner of local customers. Now if TSMC didn't want to qualify another supplier (be it because it would be too much effort or out of a desire to be CE! with fab 18), or a desire to only buy Taiwanese when possible is their business.
I'm trying to understand the discrepancy between CC Wei's statements and your comments. Not looking to start a war here. 😅

Fujifilm is a key supplier for TSMC, particularly for Cu slurry. If Fujifilm can provide the sulfuric acid needed for TSMC's 4nm process locally, with the required quality and at a reasonable cost, I don't see why TSMC wouldn't use it. TSMC's specifications for sulfuric acid in their 4nm process might be stricter than those used by Intel AZ, making it hard to judge without specific details. I don't think TSMC has a strong preference for Taiwanese vendors in this case.

Additionally, if TSMC's efforts to source sulfuric acid aren't driven by quality or cost concerns, why is Intel also starting sourcing sulfuric acid from Taiwan, as CC Wei suggested?

Another is the nonsense that all of the IDMs left Phoniex and that there is no talent. ON, NXP, Microchip, infineon, and Intel. Intel even had built a new Arizona expansion back in the 2010s and filled it with tools in that 2019-2020 timeframe. You also don't see Intel delaying fab42 or fab52.
I don't think CC Wei was implying that there's no semiconductor talent in Arizona. His comment was about that it's been a decade since the last major (logic?) fab construction in Arizona.

As you mentioned, Intel expanded in Arizona in the 2010s, so that construction was indeed about 10 years ago. Microchip's Fab 2 was built over 10 years ago. ON Semiconductor has its headquarters in Arizona but no fab there (I get it from Google, correct me if I'm wrong).

NXP did open a 6-inch GaN fab in 2020, though I'm not sure if the construction of a 6-inch GaN fab is very different from a 12-inch advanced logic fab. Similarly, Infineon opened a GaN cleanroom in 2016. (Get these from Google too.) If TSMC hired many construction workers from Texas, it suggests they prefer workers with advanced logic fab construction experience, like those from Samsung's Taylor facility.

I'm not trying to say TSMC is perfect. Honestly, I think TSMC's biggest challenge in Arizona isn't construction or chemicals, but how to run the fab with a majority American workforce. From what I've heard, TSMC is still relying on Taiwanese employees to sustain operations in the initial stage, but this isn't sustainable. They need to find a solution soon.
 
Debby downer crew you two always. Why are you two always SO down on TSMC. C.C is clearly being flippant in an informal talk. There is nothing to suggest TSMC is becoming complacent. Rather they are consistently executing at a high level in the most complex business there is. Have there been stumbles in Arizona? Surely but getting mass production up and going by the end of 2024 is pretty remarkable considering how this is TSMC's first real step out of Taiwan at scale. They got Japan done on time, Arizona now up and humming and starting in Germany soon. All while executing at the highest level financially and technically.
I don't think TSMC has become complacent. However, since they started the AZ construction projects they've made public statements about all of the obstacles and challenges they've run into which they position as surprises or issues they didn't plan on. They didn't know about permitting, environmental laws, the local environmentalists and their concerns, unions, the requirements of the CHIPS Act requiring union construction labor, the local availability of supplier partners... to name the ones I remember hearing surprise about. Speaking as a shareholder, this is not how world-class companies act, and I don't care where they're from or what they're used to. It's called diligent research and detail-oriented planning. I think my impatience has more to do with my regard for TSMC as being absolutely world-class, or they couldn't achieve what they have, and yet they're publicly talking about all of these foreseeable issues as "surprises"... well, I can't believe they do such incompetent planning. I can't avoid the impression that they're trying to make their projects in the US look more like miracles of some sort, than the typical bureaucratic nonsense everyone who builds anything in the US puts up with.

As for Wei's flippant remarks, I didn't have high regard for Gelsinger's flippant remarks when he made them either.
 
I really have mixed feelings about this. CC Wei has a very good sense of humor but sometimes it does not translate well. "I had a lot of snot and tears!?!?!? :ROFLMAO: CC Wei is also a great negotiator so this could be part of the speech as well, to get better responses from AZ politicians and Labor Unions.

Here is the translated article:

Talking about the American factory Wei Zhejia: I had a lot of snot and tears and endured all kinds of training

What would happen to the world without TSMC? Wei Zhejia, chairman and president of TSMC, pointed out, "Without TSMC, the world economy will stagnate. Without TSMC, of course my opponents will be happy, but they will not be happy for long, because without TSMC, after 2026, all autonomous driving will All gone."

Wei Zhejia attended the National Taiwan University EMBA Baichuan Lecture tonight and gave a speech titled "If there is no TSMC". His witty remarks and humorous speech not only kept the audience laughing, but also made the participants understand through a 30-minute speech why it is impossible for TSMC’s latest manufacturing process to be moved to other countries and the world. Without TSMC, not only would people not be able to replace their Apple phones, but even refrigerators, TVs, cars and other products would be affected.

Since TSMC officially set up a factory in the United States in 2021, many people in the market have been worried about whether TSMC will bring the latest technology to the United States, so that Taiwan will no longer have an advantage.

Before answering the market's doubts, Wei Zhejia first said, "If we want to repeat the production of TSMC, how long will it take? It's not long. You can build a production line in 2.5 years." However, this 2.5 years is There are prerequisites: "You must have enough land, water, electricity, and talent. If you don't have the technology, let's talk about it separately."

2.5 years is only the time to build one production line, but every time a new technology comes out, TSMC has to build at least three factories. Wei Zhejia took 3 nanometers as an example, "Building a 3 nanometer factory must meet all customer needs, which will take about five years. If you want to move the entire TSMC 3 nanometers to any country, assuming that your land, water and electricity are not a problem. If there are no problems or talent, then it will only take five years.”

So far, the progress of U.S. factory construction is not in line with TSMC’s original expectations. Wei Zhejia said with a smile, "We were in Arizona. I had a lot of snot, tears, and all kinds of training." He said it humorously, and the audience burst into laughter, but he pointed out three things that no outsider could imagine. problem.

"First, we thought that since the United States is so big, building a house is not a big deal. Wrong, building a house is a big deal. I want to build a production line in Arizona. First, they have never heard of what a production line is. I said, Isn’t there an IDM (referring to the major IDM manufacturer Intel) next door? He said, that was ten years ago, and now everyone has left, so they can’t build it.” Wei Zhejia said that the first problem is the local infrastructure. The relevant thing is that they don't know how to build factories. In the end, half of the construction workers were recruited from Texas.

Second, I never thought that in one place, I would encounter different laws and regulations of the local government. "TSMC is a company that is well taken care of by the government. We were ahead of the curve and built various production lines, but the government didn't know what to do. We discussed with the Taiwanese government how to formulate regulations." However, after arriving in the United States, although the U.S. government also I don’t know how to formulate regulations, but the U.S. government told TSMC, “Please pay first, and we will use your money to hire experts to formulate the rules. Finally, every time there is a meeting, there are city officials, using our The experts hired by Qian also included TSMC, and then TSMC took the floor and jointly formulated the rules before moving forward. The total number of rules drawn up was only over 18,000, and it cost 35 million US dollars.”

Wei Zhejia said with a smile, "I'm not complaining, I'm just talking about my experience." Then he mentioned the local chemical problem in the United States. The United States cannot produce the grade that TSMC needs. If it does, the price will be the same as what TSMC originally purchased. Five times the price, "How do we solve the problem in the end? We transport sulfuric acid from Taiwan to the LA port, and then truck it from the port to Arizona. Even this is cheaper than doing it in the United States."

But it is clear that Intel, a major American semiconductor manufacturer, has a factory next to TSMC. So, how does Intel face the problem of chemical materials? Wei Zhejia said, "They thought it was a good deal, so they started buying from Taiwan. It's hard to believe when you hear this. Ask them, how did you live before? They said, it was very hard before, but now you are here."

The second problem is related to the labor union. Wei Zhejia said, "The establishment of the labor union was a lofty ideal to help workers fight for better benefits. However, after that, the labor union became very strange." It is obvious that TSMC has also encountered problems with labor and management locally. problem.

The third problem is related to the administrative speed of the local government. "TSMC has made such rapid progress because it can build a production line in two and a half years. However, in these two and a half years, our technology has been improving. Every time the technology is improved, the entire production line The design needs to be changed, including the water, electricity, and chemical lines connected under the production line. Every change is made together with the manufacturer, but it is not possible in the United States. Any changes must be stopped first. The drawing is good, please apply and get approved before proceeding.”

Judging from this speed, Wei Zhejia bluntly said, "Do you think it is possible for the latest technology to go to the United States? Just apply for everything, and it will take Taiwan at least twice as long to come in after applying." Obviously, TSMC The latest technology will only stay in Taiwan and will be difficult to move to other countries.

TSMC Chairman Wei Zhejia was invited by the National Taiwan University EMBA Student Association to deliver a speech tonight. Picture/provided by readersTSMC Chairman Wei Zhejia was invited by the National Taiwan University EMBA Student Association to deliver a speech tonight. Picture/provided by readers
 
I don't think TSMC has become complacent. However, since they started the AZ construction projects they've made public statements about all of the obstacles and challenges they've run into which they position as surprises or issues they didn't plan on. They didn't know about permitting, environmental laws, the local environmentalists and their concerns, unions, the requirements of the CHIPS Act requiring union construction labor, the local availability of supplier partners... to name the ones I remember hearing surprise about. Speaking as a shareholder, this is not how world-class companies act, and I don't care where they're from or what they're used to. It's called diligent research and detail-oriented planning. I think my impatience has more to do with my regard for TSMC as being absolutely world-class, or they couldn't achieve what they have, and yet they're publicly talking about all of these foreseeable issues as "surprises"... well, I can't believe they do such incompetent planning. I can't avoid the impression that they're trying to make their projects in the US look more like miracles of some sort, than the typical bureaucratic nonsense everyone who builds anything in the US puts up with.

As for Wei's flippant remarks, I didn't have high regard for Gelsinger's flippant remarks when he made them either.

The easiest way to sabotage the CHIPS act is for benefactors to badmouth the political arrangements behind it, and shout corruption, which is what these handouts given to union bosses, lobbied NGOs, connected chemical suppliers, etc.

I am hearing already, someone in Phoenix started to spin "microchip factories give you cancer", and you know how touchy Americans are about cancer, as if it is leprosy or something.
 
The easiest way to sabotage the CHIPS act is for benefactors to badmouth the political arrangements behind it, and shout corruption, which is what these handouts given to union bosses, lobbied NGOs, connected chemical suppliers, etc.

I am hearing already, someone in Phoenix started to spin "microchip factories give you cancer", and you know how touchy Americans are about cancer, as if it is leprosy or something.

No problems reported at the TSMC fab in Japan which was completed in less than 2 years? Let's see what Dresden does. Either way TSMC clearly was not fully prepared to build a fab in the US, learning from their mistakes. TSMC, however, is a very quick learner, absolutely.

Some companies play tic-tac-toe or checkers, TSMC plays 3D chess...
 
Here are the slides:

No TSMC Speech.jpg



TSMC Timeline 2025.jpg


TSMC Experience in AZ 2025.jpg


TSMC Ststua Quo 2025.jpg
 
I think Wei needs to start asking himself, "What would Jensen do?"

Mr. Huang had decades to build himself an image in USA, he is a major persona. Everyone knows what nVidia is about, people say Geforce.

Mr. Wei likely has to often present to people who don't know him, nor can tell apart microchips from potato chips. In reality, image is an almost tangible, tradeable commodity in USA. You can exchange flattery, if you yourself is somebody, if not, badmouthing, and hostility towards you is completely risk/cost free. Mr. Wei doesn't have the "armour" which the image provides in USA.

If doing business in the country depends on someone influential not being hostile towards you, 1. it's a dumb endeavour to begin with, 2. but if you went there, forget about treating anyone nicely. If accommodating every special snowflake is obviously not possible, why even bothering treating them nicely to begin with.

If people who now spin "microchip factories give you cancer" in Arizona will prevail, and block the fab from meaningful functioning, it will mean that TSMC has lost out hugely on accommodating all these interest groups for nothing, and then it will lose out a ton on still having to wrestle them in the courtrooms, and smoke filled rooms of political lobbyists. If you bend over, you lose twice: first, when you do bend over, second when that doesn't help, and backfires.

The history of California High Speed rails is a very illustrative https://en.wikipedia.org/wiki/California_High-Speed_Rail. They were exceptionally accommodating to basically every big man with a lawyer there, thinking of it being the optimal strategy, and they got nowhere. Habitual avoidance of confrontation is exceptionally costly.
 
I'm trying to understand the discrepancy between CC Wei's statements and your comments. Not looking to start a war here. 😅
I wouldn't be worried about that. I am pretty difficult to truly anger. On the contrary, asking an honest question when someone doesn't know something is a great way to get on my good side.
Fujifilm is a key supplier for TSMC, particularly for Cu slurry. If Fujifilm can provide the sulfuric acid needed for TSMC's 4nm process locally, with the required quality and at a reasonable cost, I don't see why TSMC wouldn't use it. TSMC's specifications for sulfuric acid in their 4nm process might be stricter than those used by Intel AZ, making it hard to judge without specific details. I don't think TSMC has a strong preference for Taiwanese vendors in this case.
I don't know exactly what the situation is with sulfuric acid. But what I do know is that Fujifilm does make wet chemicals and gasses that are suitable for advanced CMOS and advanced memory production. You also have folks like MilliporeSigma that also makes materials suitable for leading edge semiconductors. As for why TSMC doesn't use them, Occam's Razor would suggest that they just don't want to go through the effort of qualifying a different supplier/chemical plant. It is hard work that TSMC might not have wanted to do. Doubly so since Fab21 is COPY EXACTLY! I mean "Global Gigafab process replication"™️ of Fab18, and they just clearly wanted to stick with what they know works for sure.
Additionally, if TSMC's efforts to source sulfuric acid aren't driven by quality or cost concerns, why is Intel also starting sourcing sulfuric acid from Taiwan, as CC Wei suggested?
The only way for CC to know that is through somebody at or formally at a supplier or intel breaking an NDA as supplier information would be considered a high level secret by any fab operator. So maybe he does know, but I wouldn't be completely shocked if he was also talking out of his rear. Assuming Intel is buying this specific chemical from Taiwan, so what? The semiconductor value chain is spread across the whole world. Almost none of TSMC's wafer making equipment comes from the ROC. Most of the electronics grade wet chemicals and electronics grade industrial gases come from Japan. To my knowledge, 100% of metal oxide photo resist is from the US. EUV lasers are from the US. EUV optics all come from Germany. Bulk wafer substrates predominately come from the ROK and Japan. I believe the largest source of 200mm SiC wafers is in the US. Alot of the assembly/test equipment is made in Germany, Singapore, or Taiwan, and like all of the advanced assembly tools come from Germany or Singapore. AMAT ion implanters and directional etch tools all come from the USA. And the list goes on. So maybe for sulfuric acid the best supplier is in Taiwan. But most of the inputs for a semi fab come from outside of Taiwan, so IMO I feel this argument is at best disingenuous from CC Wei. Based on the status of Fujifilm, MiliporeSigma, and others I haven't heard of or don't remember the names of, I feel it is completely untrue to say there are zero American chemical plants capable of meeting the purity requirements necessary for a healthy 5"nm" class node.
I don't think CC Wei was implying that there's no semiconductor talent in Arizona. His comment was about that it's been a decade since the last major (logic?) fab construction in Arizona.
I suppose that is fair. His prior comments sure made it seem like he meant the former rather than the latter, though.
NXP did open a 6-inch GaN fab in 2020, though I'm not sure if the construction of a 6-inch GaN fab is very different from a 12-inch advanced logic fab. Similarly, Infineon opened a GaN cleanroom in 2016. (Get these from Google too.) If TSMC hired many construction workers from Texas, it suggests they prefer workers with advanced logic fab construction experience, like those from Samsung's Taylor facility.
TSMC started their construction 2 years before Samsung started construction in Taylor and 1 year before Intel started their simultaneous Arizona builds. I could be wrong, but I am pretty sure that back in 2020 when TSMC started the first phase of Fab21, the newest fab in Texas was Samsung S2 line fab 2 (finished back in 2007). TI Sherman Fab 1 started in 2022 and is scheduled to finish qualification and start HVM in 1H'25. RFAB 1 went to HVM in 2009, and RFAB 2 started in 2020 and finished up in 2022. And yes you read that right TI only needs 2-2.5 years to build, do tool installs, qualify, and start ramping a fab in the US. Granted, these fabs are like 2/3 the size of a modern TSMC fab, and don't have monster EUV tools to install/qualify. But to me the math ain't mathing on why it would be impossible to have built an AZ greenfield fab in less than 5 years.
I'm not trying to say TSMC is perfect. Honestly, I think TSMC's biggest challenge in Arizona isn't construction or chemicals, but how to run the fab with a majority American workforce. From what I've heard, TSMC is still relying on Taiwanese employees to sustain operations in the initial stage, but this isn't sustainable. They need to find a solution soon.
The trouble isn't so much finding the talent, but making it so they don't just quit and work somewhere better. The one nice thing for TSMC is that they don't need to worry about people jumping ship to intel for a little while. But they do need to do much better if they want to attract new hires or experienced hires to jump ship from some well paying job at LyondellBasell, 3M, P&G, or Marathon. As an example, P&G is willing to pay like 130K for a process engineer fresh out of school with a BS. To make that kind of money at most American semiconductor firms (not in California) you need to have a fresh PhD or MS with a few years of experience. Alot of petrochemical companies do 10hr standard work days with every Friday off, as well as offering a pension in addition to generous 401K packages. And all of this is without having to deal with the toxic reputation TSMC AZ has built. There is also the elephant in the room that most engineering students aren't freaks like me who studied Chemical engineering. If you like so many others studied software or electronics, you can get a cushy job at some software company like Google with exorbitant pay or land a gig at a chip design house like Qualcomm and also make way more money than a Process Engineer ever would and free shares of rapidly growing firms. Even at a software company where you actually have to work for your money like EPIC systems. An entry level BS software engineers if memory serves makes like 160K a year in Wisconsin! They also pay for employees vacations once every like 3 or 4 years, as long as it is to a country they have never visited. If they want to attract the best talent, they need to pay not just a premium over intel/Micron and especially GF but a premium vs other US employers (especially companies that hire lots of Physics/ME/EE/Mat-Sci/Chemistry/Chemical Engineering/experienced technicians/military veterans). If they want them to stay, they need to fix their culture. Even the best talent means nothing if they get run out because of a toxic work environment back to other employers.
 
Back
Top