So basically, another billion dollars worth of business for ASML and other suppliers of equipment that TSMC needed?
What I do wonder though, given ASML is a Dutch company, why no European governments are cooperating to start a new, European-owned chip foundry on European soil when the equipment is to a large degree manufactured in Europe. Something similar to Airbus/EADS in terms of setup.
Yes, there are GlobalFoundries and Bosch investing in "Silicon Saxony" aka Dresden, but the former is a US company owned by Emirati oil money and the latter is targeted to produce automotive chips. Europe needs independence from the US and security from China going full bonkers regarding Taiwan.
Likely because it's BILLIONS of dollars of R&D with 0 customers at this point. You'd be building a factory HOPING to get design wins. If I'm Nvidia or Apple or insert company - am I going to risk an entire generation of chips on a startup in Europe with 0 experience beyond what they can poach from competitors? Whatever this European company is, they'd likely need to sell at a massive loss for a generation or 2, just to prove they can do it, and then hope to win some of the profitable contracts from an Apple or Samsung or Nvidia. I would imagine the investment would make a nuclear plant look cheap. And all of that is assuming TSMC or Intel don't build fabs in response to strangle you out of the market.
> I would imagine the investment would make a nuclear plant look cheap.
The disaster of Flamanville is projected to cost 13 billion euros. Even TSMCs biggest projects are smaller than that in numbers.
> Whatever this European company is, they'd likely need to sell at a massive loss for a generation or 2, just to prove they can do it, and then hope to win some of the profitable contracts from an Apple or Samsung or Nvidia.
Samsung, NVIDIA and AMD are currently blocked by the limitations of TSMC which is mostly caused by Apple paying upfront to reserve all of TSMCs 3nm output [1]. If Europe were to say "okay, we're putting up a 3nm fab and we will not sell to Apple at all" you can bet they would instantly join in to the collaboration, alone to have an alternative to a bidding war with Apple's unfathomably large war chest.
3nm chip production requires BEUV litoghraphy. That one is far out. The so-called 5nm nodes can currently only produce 14nm-sized features. If there is a customer that is willing to reserve the entire EUV and BEUV output of TSMC, so be it; but also keep in mind it is unlikely to see such a large project.
> If there is a customer that is willing to reserve the entire EUV and BEUV output of TSMC, so be it
A modern society cannot accept a single company gobbling up the entire supply of a constrained resource simply because it can. This stifles innovation.
>need to sell at a massive loss for a generation or 2,
Isn't this where gov't subsidies can help? Not only can gov't wave taxes which helps, but they can also just flat out buy the chips. Or they can add incentives to companies for buying these chips. Either way, it helps ease the transitional pains.
In fact that’s exactly how TSMC and Samsung got where they are - the Taiwanese and Korean governments decided to invest in semiconductor manufacturing at the national level.
If a plan like the above goes through EU governments will likely subsidize the likes of Siemens and Philips with 10s of billions of €.
Then they can do it like the LHC. Each country contributes, but it happens to physically located in one member country. Either EU feels like it is strategically important to not be dependent, or it's not. By your response, it is clear it is not an important agenda item.
> What I do wonder though, given ASML is a Dutch company […]
It should be remembered that they have an international supply chain, including lots of stuff out of the US. If the US wanted to, they could basically prevent ASML from shipping finished products.
Not only is ASML a one-of-a-kind company, but so are a lot of their suppliers/partners.
The Odd Lots podcast had a good episode on this for a general audience (click Manage Cookies > Reject all to get to the article):
FWIW, there are some initiatives, specifically there's an Important Project of Common European Interest (IPCEI) microelectronics and a European Chips Act (semi-conductors) in planning.
> According to the European Commission work programme for 2022, released on 19 October 2021,
> a proposal on a European chips act will be published in Q2 2022.
Because there is precious R&D you need. ASML does not have the technology to do what TSMC does. There is no European company that has it. There are few non-TSMC companies that have it right now, and their tech is out of reach of anyone but them.
It costs TSMC $12b to build their fab. It will cost a United European government a lot more money and a lot of time to even develop the tech and by that time semiconductors will have moved on.
Yes, those are valuable but power electronics don't need/use leading edge processes with EUV photolithography. Building more European fabs for power devices would be affordable and useful but it wouldn't compete with TSMC for advanced CPUs and GPUs.
What I don't understand is why ASML/TSMC doesn't start a more aggressive monetization model, taking e.g. 30% of the profits of their customers like Apple does.
why Westerners always thinking about aggressive increase price? it may give you a short term profit but in the long run. your customers will start to look else where and try new technology to by pass you.
Apple stop paying Intel's tax. Apple stop paying Qualcomm's tax. Apple stop paying Samsung's tax.
i just don't understand why you want to kill your golden goose and its not like TSMC doesn't raise price. (they increase it by 20% during covid shortage). why create a lose-lose situation instead a win-win for everyone.
> Imagine thinking profit-seeking is just a Western thing. Lol.
Seems pretty clear to me that the comment you're replying to is talking about short-term profit maximization, and not "profit seeking" in general.
The argument is that excessive short-term profit maximization hurts you in the long run by driving your customers to look for alternatives. Which you'll notice only makes sense as an argument if you do care about profits.
And I agree that it's debatable whether this is specifically a Western failing or not, but it's sure become a common approach over at least the last few decades in the US of A.
Isn't this a natural end state of a quasi-monopolist traded in the stock market? At some point the market share cannot go up by much anymore but your share holders want to see growth. So the only way to show the needed growth is by increasing margins.
Increasing prices seems fairly safe for TSMC though as long as they keep their technological edge and are ready to lower prices again once a viable competitor emerges.
The Intel example is particularly interesting as AFAIK Jobs wanted to go with Intel for the iPhone. The problem wasn't the different architecture but that Intel wanted significant margins. Seems definitely short sighed for a strategic deal like that.
Because Intel struck back and NVidia managed to punch upwards by an entire node.
The moment TSMC can say "our node makes the king -- start bidding" their margins will expand exactly as you describe. They got close. Very close. Desperate, extreme action by the other players prevented them from fully capitalizing on the trophy. This time.
In truth, the 20% price ramp still reflects a partial victory.
NVidia is a customer of fab services, and it's customers, not nanometers, that decide which processes are competitive. Nobody would say Samsung 8nm was on equal quality footing to TSMC N7. NVidia, however, proved that they could ship products on Samsung 8 that successfully compete with products shipped on TSMC N7. That's what "punching up" means. It obviously wasn't easy, but it gives them leverage in the next round of negotiations with TSMC, because it proves that TSMC isn't an exclusive kingmaker. Not yet.
Yes, Intel did "strike back" -- they shipped competitive products and earned money. Like NVidia on Samsung 8nm, they had to punch up.
Once TSMC is a kingmaker, they will charge a king's ransom -- but they aren't. Not yet. Competitors are still holding on, though they certainly have the disadvantage at the moment. The 20% price hikes (which are actually more than 20%) reflect this dynamic almost exactly: TSMC is the champ, but not the undisputed champ, so they can start hiking prices but they can't actually squeeze out their customers' margins yet.
In the case of Apple, I would presume that they directly invest under very favourable terms in the technologies, the manufacturing equipment and the fabs to lock in a price. You don't raise the price on the bank giving you your loans.
i.e. Europe does not care if 24 million Democratic Chinese are suddenly forced into an authoritative society against their will as long as they receive their cheap goods?
Personally, I'd support more European backing of Taiwan, and if it needs to be by deploying a bunch of troops there, then that's fine by me.
The problem is that the German automotive industry has only China as market to grow (Europe/US/Canada are saturated, South America, most of Asia and Africa don't have enough rich elites, Arabia wants ultra-luxury/sports cars), and the automotive industry is extremely well connected to our politics - even if the CDU/CSU aren't in government any more, liberal FDP now is, and they'll bow to the whims of the industry... meaning Germany will not consent to anything that threatens the automotive industry. Italy and Greece have been forced to sell off harbors, airports and other infrastructure to China as a consequence of the 2008ff crises, Croatia has sold a large and important bridge project to China, and Hungary's Orban is effectively China's puppet.
The result of that, combined with the 100% consensus requirement on European foreign policy decisions, is why Europe will likely rather stand aside than show China the well-deserved boot. Especially disgusting given that the European Union's foundation was the lessons-learned of the NS dictatorship and we all swore "never again" - and now we're staying silent as China is building concentration camps for Uyghurs and Tibetans.
China is a whole different league to fight against than some backwater former colony and current dictatorship that struggles to feed its army.
It's no big deal to bomb someone like Ghaddafi's Libya to hell and back... China will fuck over your entire economy if you dare to establish relations with Taiwan, such as Lithuania currently experiences.
It's nothing to do with "Europe" or anywhere else, or consumers.
The reason the Chinese government is uninterested with the "bad press" around repression of minorities in their country is because the know corporations will always take profit over ethics, and will happily contract Chinese factories in the name of being "competitive" -- which boils down to bigger payouts for the executive class.
Hopefully their discussion to setup plant in India also materialise.
Considering they are going for 22nm is Japan India also will get something similar.
Everyone start somewhere, Indian government is almost obsessed with getting a fab in India.
India already screwed up changes of getting big fabs twice.
Edit: India has announced $10 billion package for semiconductor/display, which is lot of money. There will definitely companies investing, big or small.
>India has announced $10 billion package for semiconductor/display, which is lot of money. There will definitely companies investing, big or small.
Unless India only wants to attract low end, mature node Foundry capacity. $10B is absolutely peanuts for anything that is mainstream or leading edge. Especially for India when they have very little infrastructure and Foundry Ecosystem.
And this is something I notice with India. They dont seems to have a long term strategic plan with anything. It would be much better if it was $50B over 10 years.
> Unless India only wants to attract low end, mature node Foundry capacity. $10B is absolutely peanuts for anything that is mainstream or leading edge.
I don't think this is a good "leapfrog" opportunity. Start with a larger node, sell a bunch of commodity parts, develop some expertise (and business and physical) infrastructure. This is the path Japan and China followed.
> And this is something I notice with India. They dont seems to have a long term strategic plan with anything.
Mmm…maybe. Same was said about Japan in the 1980s. Let’s look again and 25 years and see.
USSR had enormous growth in the early years (well, after adopting the NEP) all through the rural electrification period and were favorably contrasted with the 1930s depression in the west. It didn’t work out though in the long term.
Having a solid long term vision is plan, only if the solid long term plan is any good.
You trade stability for flexibility. It takes time to steer a big ship, even if your current plan has clearly extremely negative effects (e.g. killing all the sparrows resulting in a locust population boom during the Great Leap Forward). A command economy is about as big as a ship can get.
They've been making some hard left turns lately that could very well drain the gas tank from the bottom. Or usher in a utopia, although it seems very unlikely. I guess we'll see!
The vast majority of the chips we use are perfectly fine on 22 and 28 nm nodes. Not all applications require the high densities a phone or laptop CPU does.
I agree, that's why I added the last sentence. On the other hand, I do not expect this announcement to belay any of the fears over dependence on Taiwan by politicians since this fab is not on the cutting edge.
Yes. The politicians constantly talk about the advanced process and Apple. They talk about cutting edge and not the logistical chain. The fear is denial of the advanced process fab.
This should still be able to cover a lot of the shortages we see in manufacturing. I suppose a lot of the concerns motivating this are in that space more than ensuring a continuous supply of bleeding edge silicon.
For most of their stuff 60nm provides enough computing power and as for actual electricity usage, that's obviously dwarfed by the "move 2 tons of steel at 150kmph" part :-D
There’s plenty of Malaysian fabs to handle those node sizes and products, no need to site them in Japan. Unless we collectively think that c-19 is going to be a problem in SEA for several more years.
Build time for fabs is quite long, and build time for a mature semiconductor industry (in India, for example) will be longer.
That said: I have stopped trying to think any COVID timeline is too long. Shit's wack. Shit will continue to stay wack for a while. Maybe a very long while.
Exactly. MOST companies are priced out of deep nanometer nodes these days. So they go for these nodes instead because many are still selling 50-100 nm products so there's cheaper room to improve merely to 22-28 nm.
Taiwan has a long history with Japan so "culture" isn't really the biggest problem with setting up a fab in Japan by TSMC. It's probably the last problem.
I didn't mean they aren't, but said not as much as TSMC.
TSMC has already announced US, Japan new Fabs and China fab expension. They are also discussing with Germany and India govt.
Because Samsung's Foundry are not as mature and profitable as TSMC. You dont built Foundry and expect customer to come. You wait until your customer confirm their orders years in advance before you stamp that $20B cost in building a new fab.
There are always companies investing in news fabs, the difference right now is that governments are subsidising them for strategic and/or political reasons.
For example “If we choose a left-wing Education Minister, what will test scores
be in five years?” vs. “If we choose a right-wing Education Minister, what will test scores be in five years?” and then we have a good guess as to whether the
left-wingers or right-wingers have better education policy on this axis.
If the measure of good education policy is higher test scores, in a world where the tests are updated at regular intervals (and they have to be, because in a changing world the things you need to learn will change), just win by changing the tests. That's the problem with people who try to reduce everything to maximizing a utility function. Each side can drop or add things to the tests based on what they consider to be important.
Japan has a huge number of fabs. It's a problem that can be managed, short of a truly catastrophic seismic event hitting Japan. It's definitely not fishy at all.
Using existing infrastructure is a solid point, but gets defeated by history that told us how often they needed to stop production due to such natural events
Not wise considering such investment is supposed to counter the current chip shortage
Fighting chip shortage with more chip shortage? hmm not wise, nor is a future proof investment..
There isn't really a good place to build a fab when a short disruption can cause huge delays. The next best criteria is a place that's a short plane ride from HQ and politically friendly with people who won't take kindly to the country you're in being bombed by the country next to your HQ.
Talent? that doesn't make any sense, talents can build things everywhere in the world, talents wouldn't choose an island with dense population and high seismic activity
Investment? i agree, you need some investment to build things
Wich begs the question: why would anyone decide to invest there? if not to waste money, or perhaps get easy government's money?
Talent ecosystem. Graduates from nearby universities could intern and cut their teeth. Local professors could have coffee/tea chats with everyone, ensuring high throughput of relevant research. The best of the best from nearby related fabs / companies / customers could be poached with high salaries, or less boredom.
As pointed out, Japan has lots of fabs and top-rate high tech companies.
It matters to have a pipeline of high-skill workers and access to high-tech industries for a business like this.
> talents wouldn't choose an island with dense population and high seismic activity
Said talent might have been born on that island, have their family on that island, and speak that island's native language which is not used anywhere else in the world.
I mean, Japan’s had a solid history of electronics manufacturing in spite its seismic activity. Doesn’t it seem a bit contrary to suggest the country wouldn’t be a good bet?
ON Semi had/has factories in Malaysia and the Czech Republic in addition to having a bunch in the US. They aren’t cutting edge 3nm fabs but they still crank out critical parts just the same.
Things will get wild for TW anyway, and it's better to concentrate semi in East Asia which incentivizes US to commit future security. It's less effective silicon shield than fabs in TW proper, but not as suicidal as setting up plants in Arizona. And at end of the day, JP / US / EU is using as much sticks as carrots to wrestle fabs out of TSMC outside of TW soil. There's too much dependencies for TW to say no.
It’s not like putting TSMC fabs in Japan is making things any safer. A China that makes a military move on Taiwan is going to feel emboldened to right more perceived historical wrongs.
Yeah except most of them are in water scarce region. Texas and Arizona are going to be in trouble in the next 10 to 15 years WRT water supply, the mighty Colorado isn't delivering enough. The fossil aquifers are being depleted.
"Counterintuitively, the famously thirsty industry can even improve the local water supply due to a focus on reclamation and purification—Intel has funded 15 water restoration projects in the Grand Canyon State with a goal of restoring 937 million gallons per year, and it expects to reach net positive water use once the projects are completed."
"What Arizona lacks in water, it makes up for with overall stability—the state is very seismically stable and does not suffer from hurricanes, with low risks of other natural disasters such as tornadoes to boot. Building chip fabs without such guarantees is possible—for example, Intel has a large presence in Oregon—but chip fabricators on the West Coast must take extreme isolation measures, which Arizona plants don't require."
This is something I've wondered about myself. Like, if I was going to be building a foundry in North America that is supposed to operate 24/7 producing silicon for potentially many decades and required billions of dollars of ongoing investment, I'd probably try to pick somewhere in the upper Midwest like Indiana, Illinois, Ohio, or even Missouri. You might potentially get a little lake weather depending on where you pick, but the big risks like drought, earthquakes, wildfires, hurricanes, floods, and tornadoes would be much less likely, there's not exactly a shortage of space, and there's a number of mid-sized cities and some elite universities in the region that could both help attract and keep talent happy. Are the tax credits in AZ really that much more attractive?
I suspect it is a tradeoff between "cheap real estate" and "nearby California."
But if I were in charge, I'd try to put some fabs as close to the Northeast megalopolis as possible while still being cheap. So probably like... Pennsylvania? Upstate NY? Hell, why not Maine? Better weather in the NE (which isn't much of a statement, Arizona is barely human habitable) and it might be interesting to experiment with some alternative tech culture to Silicon Valley's (more down to Earth, IMO).
This is not a significant risk factor. It’s the naïve model: “they use water, therefore they will be in trouble if water becomes scarce”. However through recycling to reuse, the net use is low enough to not be a significant risk factor.
Think about it. They’re building a massive plant. This is a basic input factor. What model of the world reconciles the idea that these guys are very good at making semiconductors for decades but unable to model changes in the input factors over that period? Does that model sound like the highest probability model to explain their behavior?
the water talking point is never going to die. Taiwan doesn't have enough water for all its industry and semis industry taking up the most water yet Taiwan managed and TSMC keep building new fab in Taiwan.
"Counterintuitively, the famously thirsty industry can even improve the local water supply due to a focus on reclamation and purification—Intel has funded 15 water restoration projects in the Grand Canyon State with a goal of restoring 937 million gallons per year, and it expects to reach net positive water use once the projects are completed."
Exactly! Fabs use lots of hazardous chemicals, it's not like they can just dump the water contaminated with these into the next river. And if they clean it, why not clean it a little more and reuse it?
I think the interesting question is for what alternative purpose you could use that water for in Texas or Arizona. While processing a wafer might require quite a bit of water, it is also the case that the economic value of that finished wafer is very substantial. How comparable are other industries in the region when it comes to dollars generated per cubic meter of water consumed?
The US needs more fabs in more diverse geographical locations. AZ seems like a poor choice given the water situation. TX may have similar issues. Maybe some fabs in Canada would also be a good idea?
As I understand it, the Fabs in AZ are a net positive for the water situation there. They are incredibly responsible users (especially compared to agriculture)
The major reason for them being in AZ is constant(ish) weather, cheap land, workforce, and cheap power.
The US has Intel and China is highly unlikely to make a move on Taiwan at the peak of Taiwan's technological power. They're far more likely to wait 10-20 years to when TSMC is no longer the same powerhouse and then slowly come in as the benevolent helping hand.
I doubt the CCP taking the Republic of China will be slow, economical, and political like we've seen them do for their other puppet countries.
It will be abrupt and bloody. The ROC knows these strategies, the only way it will fall is by force and hopefully the US and other sovereign countries stand up and help against that when the day comes.
It seems that the CCPs plan is to whittle away support from all of the ROC's allies first, then crush them.
>hopefully the US and other sovereign countries stand up and help against that when the day comes.
Genuine question, what do you think that help would look like? Because every time I try and game out scenarios for China re-absorbing Taiwan by force I come to the conclusion it would essentially be an unwinnable 'war' by western nations and likely a further erosion of the United State's position as the global military superpower (on the back of the Afghan pullout). China can afford to play the long, slow game when it comes to their proximity and civilian toll in/around Taiwan, much of the West can not.
First off, China wouldn't be re-absorbing Taiwan, the ROC is the original and the PRC is the communist spin-off.
Second, I would assume heavy heavy sanctions, no western nation to deal with China at all. Military force if foreign troops stationed in Taiwan were injured or killed.
The West can play the slow game with Taiwan because it's solely the CCP's problem and the current status quo is what the West wants.
I'm talking about current affairs, not stories from five decades ago, unless you want to talk about Indians, Vietnam, Philadelphia bombing, or Kent State. The alleged Uighur "cultural genocide" is nothing compared to actual genocide in Middle East, with one million random civilians dead because of American invasion. The fact that you're even mentioning Tibet, which, before Chinese invasion, was literally a feudal state with the ruling caste treating everyone else as slaves doesn't really help your point either.
But please, continue to shill for the murderous American regime with your whataboutism.
One could very similarly argue that the US needs to have a great many things - most of them far, far simpler than a fab - on-shore. Unfortunately, the US seems uncaring. If not oblivious.
What I do wonder though, given ASML is a Dutch company, why no European governments are cooperating to start a new, European-owned chip foundry on European soil when the equipment is to a large degree manufactured in Europe. Something similar to Airbus/EADS in terms of setup.
Yes, there are GlobalFoundries and Bosch investing in "Silicon Saxony" aka Dresden, but the former is a US company owned by Emirati oil money and the latter is targeted to produce automotive chips. Europe needs independence from the US and security from China going full bonkers regarding Taiwan.