According to this page, sorting by cost, https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat...
the most expensive fab is 33 Giga$. TSMC is 17G$. pennies on the dollar in our 1-3Trillion infrastructure bill.
Make america fab again.
EDIT: it appears this is in the 2T infrastructure bill. https://www.eetimes.com/biden-ups-ante-to-50-billion-for-chi...
Prices in many foreign countries are wonky though, because of different cultural & economic systems. You can have a good lunch out (the equivalent of fast-casual; most eateries are like that) in Taiwan for about USD$2-4, and because the expectation is that you live with your parents until you get married, housing expenses are minimal. There's also a big oversupply of 20-30something labor, because the older engineers who built the company (and get paid significantly more) aren't retiring and so those skilled jobs aren't opening up. That's behind both the low prices for food (many young Taiwanese open small restaurants) and for entry-level fab engineer jobs.
[1] https://www.glassdoor.com/Monthly-Pay/TSMC-Process-Engineer-...
You should just adjust for PPP. The world bank estimates taiwan's GDP (nominal) to be 759M, but adjusted for PPP it's 1,403B. If we do the same adjustment for wages, we get $3165 per month, or $18.19/hr.
source: Parents are from Taiwan, and I have relatives in Taiwan who tell me similar price estimates.
Edit: my source (potus) was mistaken apparently. A better source is below and involves non poaching and different restaurants (not McDonald’s as far as I can tell)
> About 80 percent of fast-food workers are constricted by no-poaching clauses, according to Healey's office. The other fast-food chains targeted by the states' investigation are Arby's, Five Guys, Little Caesars and Popeyes Louisiana Kitchen.
https://www.washingtonpost.com/news/wonk/wp/2018/07/09/11-st...
See:
https://www.factcheck.org/2020/07/bidens-false-claim-about-m...
https://www.politifact.com/factchecks/2020/jul/28/joe-biden/...
https://focustaiwan.tw/business/202106300017
"The median employee salary at Taiwan Semiconductor Manufacturing Co. (TSMC), the world's largest contract chipmaker, reached NT$1.81 million (US$64,874) in 2020, up from NT$1.63 million a year earlier."
https://www.glassdoor.com/Salary/TSMC-Salaries-E4130.htm
Process engineer is like $73K/yr
So these, very skilled, employees probably get like ~15% less in Taiwan. No idea what Taiwan's employer tax situation looks like.
Sure, we can do small scale cutting edge prototype fabs, but the few production fabs we have here are not TSMC quality (go ask Intel), and they are already fully allocated. Many of those might not exist here in the country at all, if it weren’t for the requirements from the US military and certain other classified customers to have certain types of chips made here domestically.
We used to have a lot more fab engineers here in the US, but we outsourced those jobs and sent them overseas. And the people who used to do that work have moved on to other careers, or retired.
And the reason why Silicon Valley has so many EPA disaster area cleanup sites is because of the fabs(and related businesses) that used to be endemic in that part of the country. Fabs aren’t clean businesses to run. You’re going to have to find a place where you can run those dirty kinds of businesses, and that’s either going to be extraordinarily expensive, or even impossible, due to legal restrictions on the use of toxic chemicals, etc….
There’s a reason why those fabs are overseas. And all those reasons are why it’s going to be extraordinarily hard for us to pull that work back. There’s a reason why Steve Jobs and Tim Cook have both said “those jobs aren’t coming back”.
If staffing costs are 5% of of marginal costs, then making them 10% is fine if your residual income/profit is high enough.
I suspect you put a fab in Asia because you get higher profitability, not necessarily because you wouldn’t make a profit at all in the US.
Can anyone comment on what percentage of costs for say TSMC go on staffing?
Making chips seems to be quite a bit more complicated than making and launching rockets, I don't think even Elon Musk could do it.
Exports won't offset operating costs here in the USA, but perhaps we could apply some ingenuity to that through national research investments and the ensuing startup ecology. The ancillary benefits of funding research and small business infrastructure around the big fabs would be huge.
Sandia national labs has a ~3.5 annual Giga$ budget! That's _entirely_ publicly funded and represents about half of TSMC's operating budget (https://investor.tsmc.com/english/encrypt/files/encrypt_file... , looking at 50% profit yielding ~12 G$)
Tbf, many (most?) government websites are built and maintained by private companies. Remember the ACA website debacle a few years back? The government was mad at the contractor for building a shoddy product. They claimed they would take action against them but I’m not sure anything happened. If the government should be held to task on anything it should be an inability to write good contracts or hold companies feet to the fire. I’ve heard civil servants are unwilling to to the latter because they don’t want to hurt businesses or are afraid of legal protests.
var a = 1; a = 2;
Before I got any further they blurted out “that’s false! The second line is false” and this was someone who had been writing production code for the government for the last five years.
I honestly just felt really bad for them, more than anything.
I’ve seen some orgs rely on rely bad programmers because they were relatively cheap. Programmers who make changes on the fly in meetings to live production systems or say their test plan is to push it to production and just wait for customers to send in work tickets when they encounter a bug. Or blatantly lie about feasibility because they would prefer to work in an antiquated framework rather than learn a new one.
The govt owns some responsibility for lack of technical oversight but I can’t help but think a lot of it is due to business contracting incentives.
America does fab. However not newest nodes at high capacity. It just doesn't take building an identical factory and hiring away engineers. There is a complete supply chain of specialist companies which produce or fix the part the part that goes into a machine that that ... . Not mentioning that those construction costs are at economies of scale of building factories.
Such a thing can't be build in the US as every state needs a piece of the pie. An US run copy will be more expensive even when you ignore wages. Extra spare capacity for emergencies is not sound either.
EU has lately been thinking about localizing chip production here, but they seem to be stuck in the same fallacy: we need a huge, gleaming fab on our soil. No, that is just the tip of the iceberg, unfortunately.
Running such a production is difficult.
But, if we assume they were, just double the cost as a sign-on bonus for all the engineers and poach away.
This is problem that can be solved with money, and not a lot of it in the grand scheme of things.
The US is subsidizing a TSMC plant here, which should help the situation, as it'll give us access to their engineers and cross pollination. But it's still tiny compared to the need if we want to move the majority of our chip imports to domestic.
Another part is lax environmental laws. Hillsboro Oregon is embroiled in a suit with Intel where Intel dumped 100x the fluorine into the air that they claimed when D1X was first pitched. Don't need to worry about that stuff in Asia (for now).
Also, lead time. The x-ray litho machines take years to build and test. There are only two companies that make Intel's testers, and the lead time is years. So a "quick fix" isn't possible.
Speaking opinion: in the long term, it is mostly rich people trying to get richer that caused this. If greedy CEOs and shareholders would just be humans for once and think about the future we wouldn't have this trainwreck. That ain't ever gonna happen.
And the chips in question exist in the first place because...?
It's also 3 years away from opening in AZ (2024). Then they have to fine tune the equipment and processes over x runs. New plants tend to have low yields until those processes are smoothed out and perfected. You don't just set up a plant and start churning out chips with 98% success rate at 5nm once the factory is built. It probably takes 5-6 years from construction start to churning out quality chips in numbers that are actually profitable.
ASML has engineers who go out and spend months or a year or more, just taking various measurements of the location inside the building at the site where the machine will be operating. Then those engineers go back to Eindhoven to oversee the building of that machine from the ground up, and run through proving cycles. Then the whole thing has to be taken apart and shipped and then rebuilt at the site. And then there are months afterwards as the team does further work and testing before they can finally use that machine for production purposes.
The Saturn V rocket was much easier to build and move. Nothing the rocket industry has can compare to the difficulty of building and moving these machines.
Disclaimer: I worked as a consultant at ASML for several months, helping them to rebuild their Unix infrastructure systems that they used to support the engineers who did all that design and build work.
Given that they will return
https://www.youtube.com/watch?v=TuW4oGKzVKc
That's what a president sounds like.
Politicians like MTG and Boebert are part of the system just to cheapen it. The crazier they are, the more off-tilt they are, the better the strategy works, the more it weakens the average Americans faith in the system. This strategy started with Trump, but it's started permeating every level of politics. The more wild they act, the more they can insist that the other side is being just as crazy.
But look what he did against racism vs Johnson (not that much compared to J.), and they both still pushed the Vietnam war. Kennedy had to work around the edges on entrenched racism to maintain his Democratic majority with southern votes. At least the parties are more honest about their proclivities today.
The population can relate just fine. It's just there's no one rich and powerful who wants to make the investment that is making it an issue.
You may say "But semiconductors are used in far more devices than phones and computers", but I'm not sure the average person realizes that. Well maybe they are now when they can't buy the car they want because it's delayed due to semiconductor availability.
Using a simple example people understand they can't get consumer electronics like Xboxs, Playstations, and Switches because of the electronics shortage. People understand there is a car part shortage due to electronics. These aren't high information people talking to me about this. It's understood there's a chip shortage right now by the general population.
People are fine with and nobody really argues that America should support a steel industry. Frame the argument the same way and most people would be ok with supporting a technology infrastructure local to America.
The other issue that hasn't been touched on is that a chip subsidy will probably only benefit 1 or 2 states, where as agriculture and other commodity level subsidies get spread out so there's less political will to make it happen. (along with a zero sum attitude in politics that if X state is getting money it means my state is not getting that money)
It's actually a pretty critical geopolitical issue, sense there's a real chance that China makes a move on Taiwan soon (soon being within the next 10 years), which will result in either the destruction of alot of the chip infrastructure or China with global control over chip infrastructure.
I'm kinda surprised countries like Japan and Korea aren't making moves to address this by starting their own fabs. (Maybe there are?)
Japan really has no presence in this business. Japanese automotive chip makers like Renesas had fabbed chips in-house, but decided to outsource everything to TSMC back in 2018. In fact, most automotive chip makers, such as Infineon and NXP also use TSMC -- the company has something like 70% of the market share.
Now, the real issue here is that the automotive chip business is fairly low-tech and low-margin business. TSMC has been around and has been making MCUs used in cars for a long time, but it seems others are less willing to jump in b/c it's going to take years to recoup their investment.
Canon is the other major vendor of chip lithography equipment, but I don’t know where their EUV processes are, or what the major chip manufacturing companies are in Japan that they would be sourcing for.
> wait, I paid thousands of dollars for my phone and computer, why are we paying them to build plants here, don't we pay them enough
This line of reasoning isn’t generally extended to any other area of production, so I don’t see why it would apply here.
https://foundersbroadsheet.com/does-the-semiconductor-indust...
Supply and demand shocks will work themselves out fairly quickly if just left to market forces without government involvement. But a large section of the US population is under the delusion that if there’s a problem, only government can fix it.
https://thehill.com/opinion/international/560338-government-...
One part of the legislation, the CHIPS Act, allocates $52 billion to subsidize the construction of new chip factories in the United States. Officials believe it could result in seven to 10 new U.S. factories. Although a national security case can be made for it, too much subsidy can do more harm than good.
War. In the event a war breaks out, a running auto plant can be re-tooled relatively rapidly into a light and heavy armor plant. It's not even a question of whether allies could supply tanks and troop transports; it's about having domestic capacity to make them because enemies could execute blockades and disrupt allied resource supply.
The funny thing about a centrally-planned dairy system is that it takes very little disruption to destabilize it.
The stuff is so specialized and automated now that even the workers probably wouldn't have much translatable knowledge as WW2 days. You would i guess have a supply of welders to lean on at least that would ramp up faster than someone fresh.
But I doubt a factory that makes modern cars can handle making M1 Abrams without so much additional tooling and extra equipment that you are halfway to making a new factory anyways. (they could probably handle making light vehicles like cars and bikes at least)
Maybe if we had mass production style designs that were built purpose first to get cranked out. But we don't, and there isn't profit in that for the Military Industrial Complex.
One scary fact that I learned is that, in war games between China and the USA, China almost always wins due to superior production capacity. Those $100B American battle carriers and $40MM advanced fighter jets can be taken out by $50,000 rockets/missiles produced at a rate of hundreds or thousands per day.
The US loses many war-games against China in the short term as China has local numerical superiority of weapons. US armed forces are spread all around the world while almost all of the PLA is near mainland China. In the opening salvos of a Taiwan conflict China would flatten US bases in the region which would be devastating.
But long term China is effectively an Island. Its geography means that it imports nearly everything it needs by Sea. It imports 10 Million barrels of oil a day. It is not food secure. It imports nearly all of the raw materials it needs. While the PLAN could dominate their near shore, they cannot escort super tankers from the Middle East all the way back to China. The USN has had a carrier strike group in the Persian Gulf for decades.
The missile argument is also just nonsense.
> The issue is that China needs energy and other natural resources to sustain production which the US and allies can effectively blockade.
You just answered your own question.
They are so hostile, and aggressive exactly because of that immense insecurity, and fear.
It's the mental model you pickup growing in any red country: you never acknowledge your weaknesses, lest you want them being instantaneously exploited.
On other hand, you feign a polar opposite. It's everywhere in China:
— Far from rich people in small towns buying fake Ferraris
— Dumb companies hiring fake "big name foreign executives" to mumbo jumbo their plans to investors
— "Advance to retreat" tactic
— Chinese businessmen hiring a service of "fake thugs" and tatooing themselves to appear "tough mafia men" to scare off actual mafias
Also, US China conflict depends on entirely 'where' and 'over what' and what each nation is holding up it's sleeve.
China does not have $50K rockets that it can use to down 'reasonably stealthy fighters' and as we have seen before, tactics combined with good tech gives overwhelming odds. For example, if (big 'if' but entirely plausible) the US can maintain air superiority in a particular region ... then it will basically maul whatever is before it. As just one example.
China is building 100 subs that are a pretty big threat to any navy ... but we also don't know about advanced US tech that may render them completely moot.
If there is no huge leverage by one side, it happens on land, and over a long time ... the then 'home team' will win.
Stealth is a lot easier in water than it is in the air.
US has enough nuclear munitions to outright defeat 1 equal superpower in a first strike scenario.
I wouldn't call half the population of the US dying in 30 minutes, and most of the other half over the next 30 days to be much of a 'victory', and I don't want the DoD to even consider employing anyone who would.
Plus, any destroyer launching these at targets deep in Chinese territory will be within range of Chinese missiles. And, as we saw in Syria, Russian air defense systems are pretty effective against Tomahawks, and we should expect Chinese systems to be at least as effective.
No they aren't. They were designed with the assumption that GPS wouldn't be available and instead will rely on terrain tracking.
>Plus, any destroyer launching these at targets deep in Chinese territory will be within range of Chinese missiles. And, as we saw in Syria, Russian air defense systems are pretty effective against Tomahawks, and we should expect Chinese systems to be at least as effective.
What are you talking about? We had satellite imagery with holes in the ground that disproved the Russian claims that they shot down most of it.
Nothing you said is factual.
> whether light and heavy armor is actually of military worth in an era of drones and air power is a question I don't have the military training or knowledge to answer.
On this point, it obviously depends on what kind of war we're fighting. The British used unarmoured and open top Land Rovers with 50cals mounted on the rear for the first few years in Afghanistan until the government got its act together (they looked cool, but were IED death traps). There's a similar story with the Humvee.
Nonetheless, civilian style vehicles had their use.
Car factories might not make good tanks, but the mechanical engineers and tooling knowledge is invaluable and can't be scaled up overnight.
Same with solar panels, there's a reason both Obama and Trump imposed protective tariffs. One of the few areas both parties agree to. PV panels are the future of energy, so you need to have domestic factories no matter the cost.
Another benefit of domestic factories is domestic innovation. It's much harder for an engineer coming out of university and going through their career working at an office computer terminal to ever make a breakthrough innovation, if they can't step inside the factory to see how things are really done.
If that factory is down the street, it's much easier to do an apprenticeship, get a tour, or chat with the manager about their pain points. If it's in another country, you'll have to schedule a formal visit and you'll probably need to be a very important customer for that to happen.
Grad students coming out of a research lab would tend to focus on getting an extra 0.2% cell efficiency, but it's more likely that the innovation that makes PV competitive is something like reducing the scrap rate, or figuring out how to run cells through the QC machines more quickly.
Worst? Really? You think corn fields survive nuclear war? An asteroid the size of you know what?
What would the land be doing without subsidies?
And on a bribes and corruption basis, to some extent, too?
For example, I never thought about what are the strategic industries, when voting.
You're clearly not a dairy producer in Wisconsin, a military contractor in Virginia/Maryland, or a property owner in Southern Pines, NC.
But he also saw the job as a lifetime career, which is partly generational and partly just a different mindset because a lot (most?) of the engineers there were former military.
I will say, though, that his mindset generally made for a very friendly and nurturing team experience.
I was just giving it as an example of how whole tertiary industries that aren't directly part of the military-industrial complex will none-the-less vote (R) like their lifestyle depends on it, and for the same reason as contractors.
West Virginia has "Friends of Coal" license plates. Indiana has IUOE license plates. etc.
Wow, I had no idea
I work with software, maybe in a way then I'm in a happy bubble
For auto - companies do not exist in isolation. They're a huge pillar for other things, especially the local economy.
If you were the CEO of a 'region' that actually owned all of the businesses, this would show up in your balance sheet: it would be perfectly acceptable to lose money in your auto-making 'division' if it meant that the 'education division', 'healthcare division', 'civic division' were made profitable.
Japan also makes a ton of money exporting it's cars so it's valuable.
Finance, telecoms, farming and entertainment all have different reasons for being supported but the impetus in most cases is rational.
If anything we probably need more high end fabs and this whole 'single source with 24/7 operations' is maybe not worth it, we might just have to be paying a little more for our chips.
Especially in capital intensive industries, say cars, all of the competing countries are propping up or bailing out their auto industries. So even if companies should theoretically die and get replaced, there is a structural disadvantage as the competition in other countries has an insurance policy against companies failing (bailout / subsidy / tariffs / strategic protection / etc).
I figure there are also different definitions of strategic depending on the size and capability of an economy. The US is large and geographically fortunate enough that it theoretically "could" be almost entirely self-sustaining if it set its mind to it. Fuel, semiconductors, transport, lumber, food, guns.
Did you consider that it's because people need an income to survive? Employment is objectively a basic need for anyone who doesn't own sufficient capital being that they have nothing else to sell but their labor.
I think this time is quite a bit far off. But eventually will have sourcing price of just the raw materials. That said and regarding your next point -
"Ferrari and Rayban? Near-zero market value tomorrow."
On this I disagree strongly. Current trend is that IP becomes constantly more important. So that while manufacturing costs for said items may tumble, IP laws likely protect the sale of items branded as such. And the price difference does not likely result in cheaper prices, but in a bigger revenues.
I read some year ago that there were laws in US which at least somehow discouraged the pharma from shipping manufacturing abroad, and it was the big pharma itself which lobbied these laws out.
This is how China held US at gunpoint in the early days of COVID — "Stay put, or we pull 90%+ of your antibiotics supply"
In mine, dairy is a strategic industry that is coddled by government so that we can have a local source. But every country thinks that so there’s foes and allies that can supply us UHT or powder milk on a moment’s notice. And if our milk supply disappeared tomorrow, everyone over 1y could substitute with 10000000 different things for nutrition.
Cars are another. Every country props up its auto industry, so why worry about a domestic supply?
Meanwhile, there are critical components, pharma ingredients or other inputs built in 1 factory in a country we could likely end up in a big dispute with.
Just goes to show that industry protection has nothing to do with risk/dependence on that product, and everything to do with picking and choosing which industries are important voting bases.