Biden to visit Taiwanese chip manufacturer TSMC's Arizona plant on Dec 6

[Daniel Nenni]

​

[1/2] A logo of Taiwan Semiconductor Manufacturing Co (TSMC) is seen at its headquarters in Hsinchu, Taiwan August 31, 2018. REUTERS/Tyrone Siu/File Photo

WASHINGTON, Nov 30 (Reuters) - U.S. President Joe Biden plans to travel to Taiwanese chip manufacturer TSMC's Arizona facility on Dec. 6 to promote the administration's push to boost U.S. semiconductor manufacturing, the White House said.

Taiwan Semiconductor Manufacturing Co Ltd (2330.TW), , a major supplier to Apple Inc (AAPL.O) and the world's largest contract chipmaker, is constructing a $12 billion plant in Phoenix, Arizona, Reuters reported last week. The company is holding a "tool-in" ceremony in Arizona on Dec. 6.
Advertisement · Scroll to continue

Taiwan Economy Minister Wang Mei-hua said Biden's attendance showed the importance he attached to TSMC's investment.
"I think we will form a good supply relationship with the United States," she told reporters in Taipei.
TSMC founder Morris Chang said last week the company is planning to produce chips with advanced 3-nanometre technology at its new factory in Phoenix but the plans are not completely finalized yet.

TSMC, Asia's most valuable listed company, has said it was building on a site for a potential second plant in Arizona. The company began construction in mid-2021.
Taiwan's dominant position as a maker of chips used in technology from cell phones and cars to fighter jets has sparked concerns of over-reliance on the island, especially as China ramps up military pressure to assert its sovereignty claims.

 

[Daniel Nenni]

Apple sure does have a lot of influence in Washington DC! Tim Cook will be there.

 

[blueone]

Every time Morris Chang speaks lately I become less impressed with him. What an egomaniac. From the article linked above:

"Chang said people had only just woken up to how important chips were."

"There are a lot of jealous people, jealous of Taiwan's excellent chip manufacturing."

No, Morris, even our profoundly non-technical politicians have always realized how important chips are. The world is reacting to China's threats regarding Taiwan, and how they might impact the western world's globalized manufacturing strategy and national security. It's not jealousy, Morris, it's fear. And I suspect it's not going to end well. Subsidies are market distortions that seldom end well. And TSMC's successes with manufacturing locations outside of Taiwan are not broad or impressive. Egomania combined with a lack of experience is often a recipe for difficult lessons to be learned.

 

[Daniel Nenni]

Every time Morris Chang speaks lately I become less impressed with him. What an egomaniac. From the article linked above:

"Chang said people had only just woken up to how important chips were."

"There are a lot of jealous people, jealous of Taiwan's excellent chip manufacturing."

No, Morris, even our profoundly non-technical politicians have always realized how important chips are. The world is reacting to China's threats regarding Taiwan, and how they might impact the western world's globalized manufacturing strategy and national security. It's not jealousy, Morris, it's fear. And I suspect it's not going to end well. Subsidies are market distortions that seldom end well. And TSMC's successes with manufacturing locations outside of Taiwan are not broad or impressive. Egomania combined with a lack of experience is often a recipe for difficult lessons to be learned.

Morris had filter issues a long time ago and now it is even worse as he ages. Same can be said for most people in their 70s and 80s including our sitting president.

 

[cliff]

Ross Perot said in 1992 that the politicians in D.C don't know the difference between potato chips and computer chips. They seem to have realized it in the last few years. TJ Rogers said correctly about Bob Swan and the Intel culture 1.5 years ago. Some nice old lady in my neighborhood says Intel went woke from the late 80s through Bob Swan. Her husband worked at Intel. He ran some of the fabs. Something like that, but I doubt he had the in-depth technical knowledge of that Warren, or whatever his name is character (did he get fitted with a bunny suit at TSMC yet?). I don't know about the jealousy part, but I happen to agree with what Morris usually says.

I worked with several engineers who worked at Intel in completely different business units (a cellphone division, a CPU division, and another). They all used TSMCs fabs... while they were working at Intel! Can anybody here explain why they would do that?

I am glad TSMC has moved to Arizona and are bringing the Taiwanese engineers over. I share Mr. Blue's concerns. In parallel, I wish Mr. Gunslinger the best. Hopefully he has some Mr. Grove in him.

 

[blueone]

I worked with several engineers who worked at Intel in completely different business units (a cellphone division, a CPU division, and another). They all used TSMCs fabs... while they were working at Intel! Can anybody here explain why they would do that?

I am glad TSMC has moved to Arizona and are bringing the Taiwanese engineers over. I share Mr. Blue's concerns. In parallel, I wish Mr. Gunslinger the best. Hopefully he has some Mr. Grove in him.

When I was at Intel I was in a division for a while that used TSMC as a foundry. There have been multiple reasons why some Intel-designed chips were and are fabricated at TSMC. First, and probably most importantly, many of these chips have projected sales volumes in the millions, not many tens or hundreds of millions, like Intel CPUs and chipsets. Intel's fabs and their organizations are most efficient for very large volumes. (As an aside, you would think that would have made Intel best for fabricating iPhone and iPad CPUs years ago, but Intel said "no" when Apple came calling, which was in retrospect dumb.) Fitting a relatively low volume fab run into processes optimized for volumes an order of magnitude or two larger made little sense to the fabs.

Then there's the issue that Intel's fab processes were highly tuned for circuits CPUs needed. Like digital logic and SRAMs. Some of these other chips, like networking and wireless chips, FPGAs, etc., have broader needs, and tuning processes for those needs at low volumes made little economic sense.

Another issue was that these other chips were often dependent on external IP, and porting them to Intel process was relatively expensive for the business cases of lower volume or less expensive chips, and they probably already existed for TSMC processes.

An economic reason was that Intel's fab capacity allocation has been historically done by the projected total gross margin of the products. Nothing could compete with CPUs and chipsets for capacity allocation by that measure.

There are other reasons too, but I think these are the top ones I can think of.

 

[cliff]

Thanks for that information. Do you think that IFS can handle transceiver chips now on 16nm?

 

[blueone]

Thanks for that information. Do you think that IFS can handle transceiver chips now on 16nm?

I don't have that information.

 

[Daniel Nenni]

Thanks for that information. Do you think that IFS can handle transceiver chips now on 16nm?

Intel IFS 22nm and 14nm both have that ability.

 

[cliff]

Thanks.

I would like to "ask the experts" and take a poll. If Intel would have bought out Infineon's and Freescale/Fujitu's cellphone division today, and you are the GM/VP of the cellphone division with a choice of your own Intel fab using 22-14, or choosing the external TSMC 16-12, which do you pick?

Mr. Blue, the answer will tell us if Mr. Chang is justifiable arrogant. Daniel is on the spot here, so I don't think he will give an answer.

Note: JMS, you cannot choose MCHP 270. We aren't making toaster ovens

 

[Daniel Nenni]

Thanks.
I would like to "ask the experts" and take a poll. If Intel would have bought out Infineon's and Freescale/Fujitu's cellphone division today, and you are the GM/VP of the cellphone division with a choice of your own Intel fab using 22-14, or choosing the external TSMC 16-12, which do you pick?
Mr. Blue, the answer will tell us if Mr. Chang is justifiable arrogant. Daniel is on the spot here, so I don't think he will give an answer.
Note: JMS, you cannot choose MCHP 270. We aren't making toaster ovens

Since the designs were originally on TSMC I would have stayed there. New designs I would do a competitive analysis of internal and external processes for that design (same as fabless companies do) and may the best fab win. Process nepotism is a bad thing as Intel has shown. Some people call it drinking your own bath water which is why I prefer showers.

 

[blueone]

More whining and complaining from TSMC.

TSMC’s Arizona Chip Plant, Awaiting Biden Visit, Faces Birthing Pains

High costs, lack of trained personnel and unexpected construction snags are among the issues cited by the Taiwanese company as it rushes to get the $12 billion factory ready to start production in December 2023.

www.wsj.com

My favorite part of the article: TSMC is having trouble hiring US-based engineers. Newly-hired US-based engineers must relocate to Taiwan for a year to 18 months for training. I wonder... how much does that requirement reduce the pool of applicants willing to work at TSMC? I'd bet a lot.

Reiterating what I said a while back, I can't believe TSMC is so incompetent as to not have researched these cost and hiring factors before they committed to building.

 

[Mooredaddy]

More whining and complaining from TSMC.

TSMC’s Arizona Chip Plant, Awaiting Biden Visit, Faces Birthing Pains

High costs, lack of trained personnel and unexpected construction snags are among the issues cited by the Taiwanese company as it rushes to get the $12 billion factory ready to start production in December 2023.

www.wsj.com

My favorite part of the article: TSMC is having trouble hiring US-based engineers. Newly-hired US-based engineers must relocate to Taiwan for a year to 18 months for training. I wonder... how much does that requirement reduce the pool of applicants willing to work at TSMC? I'd bet a lot.

Reiterating what I said a while back, I can't believe TSMC is so incompetent as to not have researched these cost and hiring factors before they committed to building.

You seem to put alot of stock in this article being accurate huh?

 

[nghanayem]

You seem to put alot of stock in this article being accurate huh?

You can just look at all of the Arizona job postings. That, and also, where is TSMC going to train it's engineers other than the ROC? I can guarantee you they aren't getting sent to their 150mm fab in Oregon (one that TSMC has always complained about being more expensive to run than a fab in the ROC).

As I have pointed out before; the awareness for the semiconductor industry is low among MEs, chemist, and chemical engineers. Most EEs I've known would rather be a designer than a process engineer (to say nothing of many people who would have been EEs picking up higher paying CprE or CS). TSMC also has to fiercely compete for talent (and not just with intel, samsung, GF, TI, Micron, ect). Companies like DuPont, 3M, BP, P&G, Chevron-Philips, ExxonMobil, and many more have comparable wages, have plants in low cost of living areas, and likely aren't as demanding as a semiconductor company would be. As opposed to the ROC where TSMC and it's suppliers are a large portion of the economy and TSMC can leverage this scale/prestige to dictate better labor terms.

Bottom line the semiconductor industry isn't for everyone. My opinion you have to be passionate to survive - doubly so for fab workers. Unless you offer software developer pay, you will struggle to encourage more people to study EE, physics, ChE, ME, chemistry, MatE and the like: as well as struggling to pull those folks from the myriad of other American companies that are competing for the same minds. One positive prospect for TSMC is they should be able to find plenty of mechanically savy and mentally tough techs from the legions of American veterans. My advice; if they aren't already working with the relevant organizations to recruit retiring servicemen/women then they need to get on that ASAP.

 

[blueone]

You can just look at all of the Arizona job postings. That, and also, where is TSMC going to train it's engineers other than the ROC? I can guarantee you they aren't getting sent to their 150mm fab in Oregon (one that TSMC has always complained about being more expensive to run than a fab in the ROC).

The "Oregon" fab is actually in Camas, Washington. Morris keeps calling it the Oregon fab, but it isn't. The rest of your post is accurate.

WaferTech Limited Liability Company

As I have pointed out before; the awareness for the semiconductor industry is low among MEs, chemist, and chemical engineers. Most EEs I've known would rather be a designer than a process engineer (to say nothing of many people who would have been EEs picking up higher paying CprE or CS). TSMC also has to fiercely compete for talent (and not just with intel, samsung, GF, TI, Micron, ect). Companies like DuPont, 3M, BP, P&G, Chevron-Philips, ExxonMobil, and many more have comparable wages, have plants in low cost of living areas, and likely aren't as demanding as a semiconductor company would be. As opposed to the ROC where TSMC and it's suppliers are a large portion of the economy and TSMC can leverage this scale/prestige to dictate better labor terms.

Completely agree. We have a chemical engineer in the family, and he had multiple industry choices when he graduated. I'd add biotech to your list of industries, and biotech companies are formidable competition for US high-tech for graduates in many of the fields you listed.

 

[blueone]

You seem to put alot of stock in this article being accurate huh?

I know it's accurate. As an investor it's embarrassing to read TSMC's constant whining. It does not reflect well on them. I still can't believe they didn't do this due diligence.

 

[Daniel Nenni]

You seem to put alot of stock in this article being accurate huh?

I know it's accurate. As an investor it's embarrassing to read TSMC's constant whining. It does not reflect well on them. I still can't believe they didn't do this due diligence.

If you want to see birthing pains just wait until Intel opens up in Ohio. And what about all of the other fabs around the world? The semiconductor industry as a whole is resource constrained, not just TSMC.

Just a couple of other things: Building and ramping fabs is resource intensive but once they are up and running it is much less so. I have toured fabs all over the world and can tell you that TSMC has very few people actually working inside the fabs in comparison to others. We are talking about complete automation and AI/ML is making things even easier.

Another thing is that doing an 18 month tour in the US is a vacation in comparison to working at fabs in Korea, Taiwan, or China. TSMC also has first pick from the semiconductor centric Taiwanese Universities. You are a rockstar if you get hired into TSMC, absolutely.

And your constant whining comment is way out of line so you had better check that at the door.

 

[blueone]

Another thing is that doing an 18 month tour in the US is a vacation in comparison to working at fabs in Korea, Taiwan, or China. TSMC also has first pick from the semiconductor centric Taiwanese Universities. You are a rockstar if you get hired into TSMC, absolutely.

The 18 month tour I was referencing from the article was the one US hires have to make in Taiwan:

Another challenge is personnel. TSMC has invested more in recruiting after struggling to find new engineering graduates in the U.S., said people familiar with its efforts. Engineers hired in the U.S. are sent to Taiwan for a year or a year and a half of training, they said.

This requirement would seem to severely limit interest from US-based applicants.

And your constant whining comment is way out of line so you had better check that at the door.

Your site, your rules.

 

[nghanayem]

Just a couple of other things: Building and ramping fabs is resource intensive but once they are up and running it is much less so. I have toured fabs all over the world and can tell you that TSMC has very few people actually working inside the fabs in comparison to others. We are talking about complete automation and AI/ML is making things even easier.

Another thing is that doing an 18 month tour in the US is a vacation in comparison to working at fabs in Korea, Taiwan, or China. TSMC also has first pick from the semiconductor centric Taiwanese Universities. You are a rockstar if you get hired into TSMC, absolutely.

1) How recent was your last tour of a TSMC fab? And how recent were your last tours to American fabs? Were these older 200 or 300mm fabs, or were they newer Samsung, intel, or Micron fabs? I feel like the fab I work at is pretty well automated as is, and while there are room for improvements (there almost always are), I have a hard time even imaging how TSMC has managed to bring automation to another level. If so then I guess I better to start wracking my brain on what things we could automate that aren't already .

2) I wouldn't doubt it. I love this industry; but from what I hear Samsung Korea is like, I wouldn't touch being a process engineer there with a 10' pole. However I would have to imagine that even working at a US fab is harder than working at most US food or chemical companies.

 

[cliff]

Can one of you guys provide an estimated cost per wafer (flat and percentage), say in 3nm, that an additional $100K/year per worker of salary will come out to be? A rough estimate is fine. You guys know how many people need to work at the fab for, say 100K WPM (or whatever you estimate).

I suspect the USG will keep subsidizing it. We have a printing press. Accept 10% inflation from now on. These foundries are a must-have. This capability picks the winners and losers. I think everybody on this forum understand this.