[dmeeker]

Samsung has a long-term capital and capacity planning process. Isn't it possible (if not likely) that they would pre-order equipment as part of capacity planning without having already decided where it's going to be deployed?

[ksec]

>Samsung has a long-term capital and capacity planning process.

This. Samsung is another company like Amazon that takes long view and flywheel to another level. And to answer question to comment below. Yes at one point Samsung was trying to make their own EUV Scanner, if you consider their NAND and DRAM scale it should comes as no surprise. And Samsung has their own Chemical Subsidiary along with dozen of others for their Foundry. They are taking vertical integration to an extreme.

And that is speaking as someone who dont particularly like Samsung. But Credit where Credit's due.

[to11mtm]

Let's also not forget their digital imaging line, specifically their foray into Mirrorless cameras

Sure, they shelved it (Hard to compete with established players who can spread R&D over volume), but people were pretty amazed at how well Samsung did entering that market from a quality standpoint.

Why does that matter? Because that indicates they have plenty of skill in lens manufacture, which is another critical component to a good fab.

[totalZero]

Makes you wonder why one of the major semiconductor equipment companies doesn't team up with Canon or Nikon and compete with ASML/Zeiss.

[MarkSweep]

Canon and Nikon already makes their own lithography machines:

https://global.canon/en/product/indtech/semicon/

https://www.nikon.com/products/semi/

[totalZero]

That's right. Deep UV. They all but threw in the towel on EUV.

But the difference is that ASML + Carl Zeiss absolutely run the market.

Nikon has about 10% market share, and Canon 5%. The rest is basically ASML.

The thought on my mind is that any one of the major American semiconductor equipment manufacturers could partner with either of the two Japanese optics houses. Canon is worth about US $25B, and its semiconductor business generates about a fifth of its revenue. It's big enough to be useful, but small enough to be acquirable.

[avs733]

it's possible but it's going to take at least a year to build the building, another year or so to install, and another year to qual (these are ballparks).

Not to mention they have to staff the thing, train the staff, etc.

[babelfish]

All of those things can be parallelized. It's one thing to say "opening by the end of 2023 is unlikely", but to say it's impossible is unnecessarily dismissive.

[avs733]

You can’t parallelism those... you need a building to install tools in it. You need the tools installed to qual them. You need the tools quaked to qual a process...

From personal experience I would posit scaling semiconductor manufacturing as 10x to 50x harder than getting a process flow to work in a development fab. Ramps are absolutely killer.

[singhrac]

While I'm also optimistic on Samsung's ability to execute some version of this plan, those things can't be parallelized. You can't set up a complex water delivery mechanism until after the building is completed and the cleanroom has been built to spec and tested, and you can't calibrate the gigantic ASML machines until after that.

[jedberg]

But you can train the staff in Korea (or at any other running fab) and have them ready to go the day it's qualified.

[shard]

That seems like a reasonable timeline, which means the fab will be online by 2023, just like the article says. Mass production won't start until 2024, though.