News Taiwan Government Looks to Protect TSMC Tech from US — Report

[Admin]

TSMC's fabs in the U.S. will be a generation behind those in Taiwan, say sources close to the company.

Taiwan Government Looks to Protect TSMC Tech from US — Report : Read more

 

[lmcnabney]

What technologies? Aren't the key lithographic equipment all imported from Europe?

 

[peachpuff]

Tim Apple: damnit!

 

[helper800]

What technologies? Aren't the key lithographic equipment all imported from Europe?

I would liken it to the following. Having / owning a CNC machine is vastly different than knowing how to make use of it.

 

[anonymousdude]

What technologies? Aren't the key lithographic equipment all imported from Europe?

ASML is Dutch. Their major partner is Zeiss which is German. Then throw in whatever other companies they also contract or subcontract from.

TSMC technologies really just means their know-how. So their process engineering, R&D, etc. Just because you have the same equipment doesn't you know how to use it and use it well. That's the exact issue that any semiconductor fabrication firm will run into. See Intel, Samsung, or the most apt example SMIC.

SMIC has DUV machines, but it took them a long while to build up the knowledge on how to use them to make a 7nm class node. Reports say that it has similarities to the original 7nm node used by TSMC, that also used DUV. That's not a coincidence at all and something that TSMC,like any company, would like to avoid in the future if at all possible.

 

[lmcnabney]

TSMC technologies really just means their know-how.

That know-how is hardly a secret. There are people with the exact same know-how all over the world. The only thing that Taiwan has is a whole bunch of fabs making and etching wafers. This is not a secret.

 

[Co BIY]

Technology transfer is a tricky business. Some of the keys to good performance may not be able to be protected as intellectual property.

OTOH in the US TSMC will have access to the full power of global IP protection. The Taiwanese government and TSMC management may have different ideas about what is wise to move "overseas" because they could have quite different goals.

 

[Co BIY]

That know-how is hardly a secret. There are people with the exact same know-how all over the world. The only thing that Taiwan has is a whole bunch of fabs making and etching wafers. This is not a secret.

Actually building, running and producing at world leading economic scale indicates that they have knowledge others don't.

 

[lmcnabney]

Actually building, running and producing at world leading economic scale indicates that they have knowledge others don't.

They have dedicated and lower-cost labor and are comfortable handling the environmental impacts of microprocessor fabrication. Western nations have to dole out huge incentives to compete because of the higher labor cost and environmental restrictions.

 

[anonymousdude]

That know-how is hardly a secret. There are people with the exact same know-how all over the world. The only thing that Taiwan has is a whole bunch of fabs making and etching wafers. This is not a secret.

By your logic, then Intel, Samsung, and TSMC who all have effectively the same equipment should have node parity right now. And yet they don't. Guess what the difference is? R&D, process engineering, etc. What people would call the know-how. You're discounting how truly complicated the fabrication process is. It's likely the most complicated thing humankind has ever done.

They have dedicated and lower-cost labor and are comfortable handling the environmental impacts of microprocessor fabrication. Western nations have to dole out huge incentives to compete because of the higher labor cost and environmental restrictions.

Then what about China? Russia? NK? Any foreign adversary to the US? You think they care about money or would be unwilling to bear the environmental impact if it meant they could have their own chip fabrication? No, it's cause they can't cause they don't have the equipment nor the know-how.

 

[DavidLejdar]

TSMC is not really like a sweat-shop, which could be easily replicated at that level of sophistication, as some here seem to be implying. To expand a bit on what some other commenters already pointed out, at that level, there are world-wide only 3 players, TSMC, Intel, and Samsung, who are producing or about to produce with the 3nm process. In comparison, e.g. GlobalFoundries has 12 nm "at best". And for a range of electronics, it doesn't need the most sophisticated semiconductors. But talking e.g. mobile phones, tablets and GPUs, there is a reason why companies such as Apple, AMD, Nvidia and also e.g. Qualcomm are customers of in particular TSMC. The high transistor density simply does a lot for such consumer products. And entering the market with something first, such as TSMC did with 7 nm and 5 nm, that helps a lot to get more orders in, such as e.g. by Apple for the A14 and M1 SoC.

So, while it is technically possible to "catch up", realistically that would require to invest a lot in (successful) R&D for own patents including having the qualified staff for it, then it would also require software development (in particular for the machines), and then about $20 billion for the newest-gen foundry. All in all, quite a different dimension than to simply get a simple sewing machine to start producing T-Shirts with common know-how.

Also, the worldwide median annual compensation for TSMC employees, excluding pensions and benefits, was $56,505 two years ago, and has since been raised reportedly. This median is something even most in the U.S. and Europe wouldn't mind earning. Which isn't to say that some staff in the U.S. wouldn't expect a bit more than that, working e.g. as principal engineer. But it is a quite different situation to a sweat-shop, where the minimum wage e.g. in Bangladesh is way below $1 per hour.

 

[logainofhades]

I would liken it to the following. Having / owning a CNC machine is vastly different than knowing how to make use of it.

But anyone can learn how to use a CNC machine.

 

[spongiemaster]

But anyone can learn how to use a CNC machine.

Stand by the self checkout registers at a Walmart for a couple of hours and you will not agree with yourself.

 

[bit_user]

But anyone can learn how to use a CNC machine.

A better analogy would be that ASML is like an oven maker. You can buy the same ovens that a 3 Michelin Star restaurant uses, but if you don't know their recipes, ingredients, assembly techniques, and cooking times, just having the same ovens won't do you a whole lot of good.

 

[logainofhades]

Stand by the self checkout registers at a Walmart for a couple of hours and you will not agree with yourself.

I work with people that probably fall into this category, but are still able to run a CNC. They might not be able to understand how to program it, but they can at least learn how to use it

 

[edzieba]

They might not be able to understand how to program it

Well ,there you go. Expecting anyone to be able to buy an ASML TwinScan and pump out leading-edge process dies makes about as much sense as dropping a fresh out of the box Okuma at the feet of somebody who can generally hold the right end of a hammer and expecting them to pump out Hastelloy turbine casings.

Operating a fab is about far more than just owning some hardware. That is why TSMC, Intel, Samsung, GloFo, etc all have radically different processes despite buying litho hardware from the same vendor.

 

[systemBuilder_49]

Intel fell behind tsmc for two reasons.

First intel was arrogant and refused to become a foundry even though the economics indicated that every chip maker had to become a foundry or they would die because no single semiconductor maker has enough business to pay for next generation foundary costs! Intel was greedy and tried to buy a CDMA chipmaker and expand into cellular (twice!) and into many other markets! ALL FAILURES!

Second through the mistake which is American Business Practice the executive suite became staffed with marketing morons and not phds in chemical engineering! These marketing morons "rested" and "vested" at the company from 2013 until 2020! Big Data center companies like Google and Amazon were pissed at hell because Intel slacked off during this time and ripped them off for Xeon chips! During this time many parts of the company made zero progress in tech - integrated gpus got no faster from 2013 until 2020 - 10nm took 6Y to develop instead of 2! That is where they lost their lead!