News South Korea Unveils $450 Billion Semiconductor Investment Plan

[Ajay]

LINK

Samsung will be spending over $151 billion through 2030 in expanding its manufacturing facilities, while SK Hynix will spend $97 billion to expand its existing facilities; in addition to $106 billion planned to build four new fabs in the Yongin. Both Samsung and SK Hynix are predominantly memory companies, manufacturing DRAM and NAND flash products. This means that while Korea is globally competitive in semiconductor manufacturing overall, it is relying mainly on memory dies, and not logic dies (chips such as ASICs, CPUs, GPUs, SoCs, FPGAs, etc). The two could put in efforts to change this, so their foundry capacity attracts fabless logic IC companies away from Taiwan's TSMC, which specializes in logic over memory.

 

[Ajay]

Countries are really ramping up the spending to remain competitive. Lots of high paying jobs, secondary manufacturers and of course, local service jobs to feed, clothe and entertain the well paid employees.
This will be interesting to watch over the next decade.

 

[DrMrLordX]

Interesting to see that the public money is going to established players rather than to potential startups. It makes a certain amount of sense in the short term since semiconductor manufacturing is now so insanely expensive that it's difficult for a new player to enter the field. In the long term, though, it may stifle innovation. And it's not just South Korea thinking that way either.

One wonders if the large-scale subsidy of semiconductor research and production will lead to companies like Samsung et al becoming more like Airbus and Boeing. Scary thought really.

 

[Doug S]

It would be stupid to give public money to a startup foundry. People were complaining about public money being invested in Solyndra a decade ago , and that cost only a half billion which is a tiny fraction of what it would cost to get a foundry startup off the ground.

It is ridiculous not only in the face of the tens of billions that would be required to build a state of the art foundry from scratch, but even something like a startup trying to build a trailing edge fab.

The only case I've heard of someone trying to ENTER the business in the past decade or two is Apple's purchase of that former Maxim fab in San Jose for $18 million back in 2015. That was clearly not purchased to operate as a going business, but to allow them to do R&D and prototyping without having any non-Apple employees in the loop as would be the case if they have a foundry fab the prototypes.

That $18 million included not only the building and land, but all the equipment in the fab (Apple brought in some new equipment so who knows how much of the original equipment has avoided the scrap heap) and Maxim had been trying to sell it for at least a year so even though it would have been very cheap for a startup to acquire it is pretty clear it would not work economically as a going concern without constant subsidy.

 

[Mopetar]

If the government wanted to support startups they could just offer a massive tax break which would spur private investment into such companies.

Much like aviation though it's a difficult industry to get into since the initial capital costs are so high. Unless a company has something new or a specific underserved niche to fill they'll probably fail or just be a bit player that gets bought up by a larger existing company later when the investors want a big payday.

 

[Doug S]

If the government wanted to support startups they could just offer a massive tax break which would spur private investment into such companies.

Much like aviation though it's a difficult industry to get into since the initial capital costs are so high. Unless a company has something new or a specific underserved niche to fill they'll probably fail or just be a bit player that gets bought up by a larger existing company later when the investors want a big payday.

Tax breaks only have value if you make a profit. Given how long it take to see a return on investment for a startup fab, even offering a 100% tax break wouldn't be enough to persuade VCs that it is worth investing tens of billions and hoping it can compete with TSMC and Samsung.

Once they were operating they'd have billions in losses from startup costs and unrealized depreciation and amortization zeroing out their taxes before the got to the point where they owed taxes. VCs don't like waiting a decade for their investment to pay off.

 

[Mopetar]

It almost certainly won't result in a new TSMC overnight or even in a decade. Even TSMC didn't reach their own prominence in the industry in such a period of time.

However billions in losses aren't too much of a concern to some investors if the long term gains are their. Take a look at Uber sometime and tell me someone who has invested in that company isn't okay with bleeding a fair bit.

Even with such incentives a lot of startups will die. That's the nature of the things and most large scale investors (often referred to as angel investors) are quite fine with 95% of their multi-million dollar investments failing because the one that succeeds will more than cover that. That winner having no taxes is actually quite substantial. Even at the lowered corporate tax rates it's an extra ~25% which is a lot of money when you get into the billion dollar territory.

 

[Ajay]

The US has plans to put 37B into semiconductor manufacturing companies. And what’s Intel spending? ~35B? We are really kicking butt, as usual .

 

[aigomorla]

The US has plans to put 37B into semiconductor manufacturing companies. And what’s Intel spending? ~35B? We are really kicking butt, as usual .

*Trying to prevent this thread from going political*

 

[Ajay]

*Trying to prevent this thread from going political*

Just a factual comment about the difference in the two countries priorities. But, consider me well warned.

 

[DrMrLordX]

It would be stupid to give public money to a startup foundry.

Again, in the short term . . . absolutely! If you're going to spend public money, you want predictable results. It's true in any country. But what if your established players become accustomed to fattening themselves at the public expense? You need a credible threat to keep them honest.

 

[moonbogg]

I'd think finding a capable army of engineers would be the biggest challenge in starting a foundry, right? Being smart and educated isn't good enough. You need people with good, functional experience, right?

 

[Doug S]

Again, in the short term . . . absolutely! If you're going to spend public money, you want predictable results. It's true in any country. But what if your established players become accustomed to fattening themselves at the public expense? You need a credible threat to keep them honest.

That would theoretically be a job for antitrust - though that mechanism is notoriously unfit for fixing competition in markets with very high barriers to entry, which is the very definition of modern leading edge foundries.

So while I'm not sure exactly how to fix things in your scenario, letting the government choose a startup to win the lottery and get tens of billions in public subsidy to be able to compete with Intel, TSMC and Samsung is definitely not the way!

 

[ronss]

TSMC building a chip factory in phoenix ,az,....that will employ 1600 people...and very possible 3 more coming to arizona..i was hoping cost of living would level off in housing , but its going up here

 

[DrMrLordX]

So while I'm not sure exactly how to fix things in your scenario, letting the government choose a startup to win the lottery and get tens of billions in public subsidy to be able to compete with Intel, TSMC and Samsung is definitely not the way!

Simple: don't choose one. Choose many, and spread out the spending. Hand out enough to get private money interested in joining the fray. Assuming, of course, you have more than one credible party.

 

[gdansk]

Given the time frame, inflation, and predicted increase in demand over the next decade the figures do not seem much higher than expected. Nor comparable to other recently announced figures.

 

[Roland00Address]

Interesting to see that the public money is going to established players rather than to potential startups. It makes a certain amount of sense in the short term since semiconductor manufacturing is now so insanely expensive that it's difficult for a new player to enter the field. In the long term, though, it may stifle innovation. And it's not just South Korea thinking that way either.

One wonders if the large-scale subsidy of semiconductor research and production will lead to companies like Samsung et al becoming more like Airbus and Boeing. Scary thought really.

Well that is because certain sectors of the economy learn things via doing them. Efficiency gains are found at the practical industrial level not just at the university level. While other sectors you can learn things and create new things at the industrial level, for example pharmaceutical drugs.

If you want a few firms not to dominate in those type of sectors that learn via doing things you need to promote pooling of knowledge and this is stuff such as require companies to be able to source X chip from 2 different or more suppliers, require cross licensing of patents and IP, prevent worker exclusivity periods and so on.

Some stuff there is no "ideal" good answers yet knowledge on how to promote the "good" is still possible it is not at the ideal / contemplative stage of cognition but the more hands on active stage of doing.