The world is probably going to be split between EUV and non-EUV lithography for a long time. I'd like to see things like the Raspberry Pi stay on the older nodes and just wring out the performance. For reference the Zen1 CPUs were on 14/12nm from GF and I believe they are still faster than the pi, so there is room for improvement from them on old nodes.
That’s true, but Zen1 TDP is a lot higher than an RPi. Hitting that level of performance in a power and heat envelope that works for a Pi probably demands a smaller process node, unfortunately.
3% to 5% more expensive every year - that is similar to or less than inflation. Given their dominance, i'm genuinely curious why they don't raise it by 20% a year. Or 50%?
I have been interested in city planning for a while and the idea of an 'urban growth barrier' is a key concept with massive benefits. Basically, when you limit a resource, like available land, it simplifies thinking, forces inward development and and very often the 'impossible' problems go away because creative solutions come out. Maybe the chip industry will see something similar if a barrier of 'just wait till the new smaller process hits' gets put in place.
A bit OT but I live on a very land constrained island with the highest population density in Europe (see Malta).
There has been major increase in demand for housing and supply cannot be built fast enough to match. Its turned most of the island into a construction site so rampant that I made an online tracker for urban planning permits so folks can get ahead on knowing whats going on around them.
Idk if you have any wisdom but there's no creative solutionising happening, just the rich able to buy whatever property they want causing prices to rise which is pricing out the middle class, causing a whole lot of grief and downstream issues (such as plummeting fertility rate because homes are too expensive).
Is there a magic toggle we missed to unlock this creativity or am I being realistic by being skeptical that limiting important resources just leads to harsher inequality?
> Is there a magic toggle we missed to unlock this creativity or am I being realistic by being skeptical that limiting important resources just leads to harsher inequality.
Creativity has been unlocked but it's not the magic kind, it's the rather mundane Robber Barron kind, where the robbers occupy the sources and hoard a bunch of goods to resell or rent at higher profits. It's true for housing, GPUs, RAM, etc. [1]
[1] Microsoft CEO says the company doesn't have enough electricity to install all the AI GPUs in its inventory - 'you may actually have a bunch of chips sitting in inventory that I can’t plug in'
Maybe Malta is just too small of a world for that magic toggle, since entrenched interests can basically just prevent the government from approving certain strategies, like building more large buildings with apartments like the Mercury Tower. But on the global stage, not even the biggest player(US) can dictate what China, India, the EU does, so if the US self-limits because of entrenched interests, that's just more juice for the squeeze for other players not aligned with the US.
So I remember urban growth barriers being all the rage in the late 00s, and Portland specifically was held up as a shining example. Many Western cities (eg. Portland, Seattle, SF & the Peninsula, Salt Lake City) followed this general pattern though.
What I actually see, living in one of those metro regions that is land-constrained, is very little infill development and actually implemented creative thinking, and lots of homelessness and unaffordable homes. Sure, being supply-constrained gets people thinking about infill development and mass transit and densification and walkable neighbors. But there's always a barrier, usually multiple ones, to actually putting that into practice. Instead, you get the obvious effects: the supply of homes is constrained, their price rises to the point where only rich people can afford them, everyone else either goes homeless or moves out of the area. Sometimes, very often, a supply constraint simply means that people do without. Even when there is multi-family infill development, people don't want to live there. Everybody wants their SFH, even though intellectually they know it's unsustainable.
I suspect it's going to be the same here. More expensive CPUs and GPUs are just going to mean more expensive CPUs and GPUs with no real silver lining.
> I suspect it's going to be the same here. More expensive CPUs and GPUs are just going to mean more expensive CPUs and GPUs with no real silver lining.
Presumably, higher profits will incentivize other players to enter the market and increase supply, and/or the company earning high profits plows at least some back into R&D to at least create better chips, which can result in eventually lower prices for the previous generation of chips.
This isn’t a possibility with land and land rights, however, so I wouldn’t expect the same dynamics to play out.
Over a long enough time horizon, yeah. Over a long enough time horizon the same happens in land too - we are finally getting decent amounts of infill development in the Bay Area, but it's 15 years after prices started to skyrocket.
Starting a semiconductor company and constructing a few fabs at cutting-edge process nodes takes at least that amount of time, so I'd expect semiconductors to have similar dynamics.
That's just delaying progress for everyone as TSMC basically doesn't have alternatives for the processes that are actually good. Definition of cutting off your nose to spite your face.
The Nvidia 3000 series were on Samsung's process since it was quite a bit cheaper than the TSMC process at the time. Samsung has already said it intends to undercut TSMC so assuming they are fairly equivalent it doesn't seem unreasonable for consumer CPUs and GPUs to be made on their process, it gives a price advantage to anyone that makes the move.
I agree. Even my quad core i7 from 2011 is still fast enough for everything average consumer does. Heck, even average i5 laptop from 2015 or so is more than sufficient.
I only partially agree. I recently got nostalgic and bought a 2nd-gen Core-i7 ThinkPad X220 from ebay (similar to what i had back in my university days).
It's generally very usable util you open firefox and browse a modern website. Chrome/chromium aren't any better.
The issue with old hardware is web browser and web pages, particularly web pages. Modern websites are incredible resource drains, it's unbelievable.
I don't notice any issues with an i5-6600k or an n100. Both are 4c/4t and extremely snappy. The only ways to get them above like 0.5 load are CPU intensive tasks like compiling, video editing, or games. The only thing I've found I can't do is real time viewing of 200 Mbit/s 4k60 10 bit 4:2:2 h265 off my camera because it's software decoded (software can only do ~half speed).
I disagree. I would submit that the average consumer utilizes a lot of computing power and that you might actually be using less than that as a more technical power user with more discerning tastes in applications and a viewpoint that sees the computer as more of a tool than an entertainment and lifestyle device.
I would also sumbit that the person who is still using a computer from 2011 would never have bought an i7 in the first place. They’re not going to follow an upgrade path where they say “I had a core i3 in 2011 that is too slow now so I’ll replace it with something that had a 2011 i7 that I couldn’t justify purchasing at the time.” Instead they’re more likely to say “I bought this laptop 15 years ago and I’ve battered the hell out of it and I’ve been frugal and it’s time for a new one,” and when they buy that new one it will also be a very modest configuration.
So the fact that a top of the line processor from 15 years ago is still serviceable is not really all that relevant. It’s going to not even be worth it on power consumption alone if you’re running it 24/7.
In other words, there’s more to a chip than raw performance.
decent chunk of RAM maybe but my parents are happily browsing away on decade+ old chrome books. Ordinary web browsing you can do on a potato machine from 2008.
Even if you're a gamer performance on the desktop has eclipsed demand so much that you can play anything on a few years old machine. It used to be way worse in the "olden" days when only 10% of people had a system that was able to play Crysis.
This is what I don't get about laptop power consumption.
I would love an x86 laptop that can operate at 5W or less, even if it isn't blazing fast. If they can squeeze more performance out of a CPU for the same power budget, why can't they make a CPU with the same capability with less power use?
5W for a laptop is raspberry pi territory, and we're adding a screen to it. I hope we ever get there for a 14" fully usable laptop, but that sounds bleeding edge territory at this point.
A laptop running below 5W needs to turn down its storage, RAM, GPU, LAN, WiFi, USB controllers, etc. The OS needs to act more like a smartphone, put everything to sleep, and use consolidated polling and interrupts for async behaviors.
I think this is the real advantage Apple has with their vertical integration. They can play a lot of complementary games between hardware and OS.
The steam deck's chip runs at 15w maximum and has I believe a 4, 6, and 8w mode. Maybe not 5w but pretty close. The linux desktop on it is a perfectly unimpressive but usable daily driver kind of thing. The problem is consumers dont want this. The failure of the netbook era proved it. People dont want to deal with slow speeds so this kind of thing wont be a mainstream mass produced item.
ARM faking x86 is probably the way to go for this case.
You should look into chinese laptops using the intel n150/n250 cpus. Those have a 6W TDP.
The issue with those cpus is that they're pretty much never used in laptops meant to do anything serious. They usually are employed into cheap plastic laptops using ridiculous batteries.
I would really like the form factor of 2011-2013 thinkpads, with the chonky 90WHr battery on the back and the 12" form factor.
With a modern cpu (low tdp and low power draw) that would be unbeatable everyday carry.
Absolutely everything in the modern economy stops him from forming a competing company, this naive stupid market is the solution idea needs to stop being regurgitated, it's toxic to society.
Freedom refers to a potentiality. You are conflating a potential (freedom) for a hoped-for benefit (highly competitive). A free market is a means towards a competitive market because it lowers the artificial barriers to entry into the market. Those barriers imposed by the universe simply are and while we can invest into R&D and education, market forces cannot change the laws of physics.
> A free market is a means towards a competitive market because it lowers the artificial barriers to entry into the market. Those barriers imposed by the universe simply are...
Are you talking about tariffs or about banning the sale of semiconductor equipment to China, so they can't compete in the market for high performance chips? Is that "a means towards a competitive market" or is it "barriers imposed by the universe"?
High barriers to entry have nothing to do with free markets.
>In economics, a free market is an economic system in which the prices of goods and services are determined by supply and demand expressed by sellers and buyers.
> A free market does not directly require the existence of competition; however, it does require a framework that freely allows new market entrants. Hence, competition in a free market is a consequence of the conditions of a free market, including that market participants not be obstructed from following their profit motive.
The barriers to entry being referred to in the Wikipedia article are presumably government related. No government is stopping Intel or Samsung or any of the other chip makers from selling what TSMC sells, hence the “framework that freely allows new market entrants” is there.
And the subsequent paragraph seems to be exactly what is happening in the market of high end chips:
> Hence, competition in a free market is a consequence of the conditions of a free market, including that market participants not be obstructed from following their profit motive.
You don't see a discrepancy between your claim that free markets have nothing to do with high barriers to entry and your own link claiming free markets require free entrance to the market?
And no, it is not specifically referring to government related barriers, although those certainly count. Government is very often the thing that generates barriers to entry, but it is by no means the only barrier to entry. Any barrier to entry is, from a strict theoretical sense, a departure from a pure free market, although that's more in the "spherical cow" sense of the idea.
You are using "free market" as a political term. In that sense, it refers to little more than to the absence of unnecessary regulations.
Economics is more concerned about whether "free market" is a useful description of the system. If there is nothing in principle to prevent the emergence of new competition, but barriers to entry (e.g. availability of capital / talent / machinery / raw materials, long-term contracts, or the time required to set up a new business) make it difficult in practice, the system may not behave like a market. Then you need to focus more on politics, both between and within governments and companies, to understand how the system is actually working.
Judging by the response to the parent come back, evidently TSMC deserves the high prices it will be charging. Why?
It's unrealistic to expect just anyone to start up a new fab. But if not one person among billions will start up a new fab it points to intrinsic difficulty or unpleasantness or lack of prestige in the task. A correctively high price has all kinds of advantages for the society.
I haven't seen any argument that difficult tasks shouldn't be priced high. Only name calling.
In this case its more its a 100+ billion investment to possibly produce a competing fab and its encumbered with many thousands if not more patents the likes of which will produce a minefield for any entrant. So far the silicon product market has just shrunk over the years with more and more failing to keep up with the latest process technology as the price of entrance and the patent warheads has risen.
Entering a multi billion risky market is no easy task even if you have investira and teams with deep knowledge. That's the reason monopolise should be avoided.
TSMC isn’t a monopoly, they are just better than all their competitors. Nothing forces you to buy TSMC; you can buy Intel, Samsung, SMIC, GlobalFoundries and various others.
This is getting faded out, but this is absolutely right. For very few use cases do you truly need the bleeding edge. So many things do not have such strict requirements, and will meet all necessary requirements on an older node. An ATTiny85 is still an incredibly useful microcontroller even today.
This monopoly is unavoidable due to the limited number of people in the world with the know-how to make these products, with the second constraint being that you need to risk $10B+ and 10+ years with a risk that you fail.
Edit: per link below, seems like you need $100B+ and 10+ years
This business is expensive, making and running fabs is way way more expensive than most anybody thinks. There are very few companies able to do it profitably.
Piddling sum in this domain, that amount won't even buy you enough for 1 production line, let alone where to put in, staff it and get all the necessary collateral to allow people to design to it.
And no amount of money will get you one of ASML EUV machines any time soon as the entire production line for the next handful of years has already been spoken for.
The world is probably going to be split between EUV and non-EUV lithography for a long time. I'd like to see things like the Raspberry Pi stay on the older nodes and just wring out the performance. For reference the Zen1 CPUs were on 14/12nm from GF and I believe they are still faster than the pi, so there is room for improvement from them on old nodes.
That’s true, but Zen1 TDP is a lot higher than an RPi. Hitting that level of performance in a power and heat envelope that works for a Pi probably demands a smaller process node, unfortunately.
> The world is probably going to be split between EUV and non-EUV lithography for a long time.
How long before China develops indigenous EUV capability?
I think it will split between TSMC (sold out forever) and Samsung/Intel (please please be our customer).
3% to 5% more expensive every year - that is similar to or less than inflation. Given their dominance, i'm genuinely curious why they don't raise it by 20% a year. Or 50%?
I have been interested in city planning for a while and the idea of an 'urban growth barrier' is a key concept with massive benefits. Basically, when you limit a resource, like available land, it simplifies thinking, forces inward development and and very often the 'impossible' problems go away because creative solutions come out. Maybe the chip industry will see something similar if a barrier of 'just wait till the new smaller process hits' gets put in place.
A bit OT but I live on a very land constrained island with the highest population density in Europe (see Malta).
There has been major increase in demand for housing and supply cannot be built fast enough to match. Its turned most of the island into a construction site so rampant that I made an online tracker for urban planning permits so folks can get ahead on knowing whats going on around them.
Idk if you have any wisdom but there's no creative solutionising happening, just the rich able to buy whatever property they want causing prices to rise which is pricing out the middle class, causing a whole lot of grief and downstream issues (such as plummeting fertility rate because homes are too expensive).
Is there a magic toggle we missed to unlock this creativity or am I being realistic by being skeptical that limiting important resources just leads to harsher inequality?
> Is there a magic toggle we missed to unlock this creativity or am I being realistic by being skeptical that limiting important resources just leads to harsher inequality.
Creativity has been unlocked but it's not the magic kind, it's the rather mundane Robber Barron kind, where the robbers occupy the sources and hoard a bunch of goods to resell or rent at higher profits. It's true for housing, GPUs, RAM, etc. [1]
[1] Microsoft CEO says the company doesn't have enough electricity to install all the AI GPUs in its inventory - 'you may actually have a bunch of chips sitting in inventory that I can’t plug in'
https://www.tomshardware.com/tech-industry/artificial-intell...
Maybe Malta is just too small of a world for that magic toggle, since entrenched interests can basically just prevent the government from approving certain strategies, like building more large buildings with apartments like the Mercury Tower. But on the global stage, not even the biggest player(US) can dictate what China, India, the EU does, so if the US self-limits because of entrenched interests, that's just more juice for the squeeze for other players not aligned with the US.
So I remember urban growth barriers being all the rage in the late 00s, and Portland specifically was held up as a shining example. Many Western cities (eg. Portland, Seattle, SF & the Peninsula, Salt Lake City) followed this general pattern though.
What I actually see, living in one of those metro regions that is land-constrained, is very little infill development and actually implemented creative thinking, and lots of homelessness and unaffordable homes. Sure, being supply-constrained gets people thinking about infill development and mass transit and densification and walkable neighbors. But there's always a barrier, usually multiple ones, to actually putting that into practice. Instead, you get the obvious effects: the supply of homes is constrained, their price rises to the point where only rich people can afford them, everyone else either goes homeless or moves out of the area. Sometimes, very often, a supply constraint simply means that people do without. Even when there is multi-family infill development, people don't want to live there. Everybody wants their SFH, even though intellectually they know it's unsustainable.
I suspect it's going to be the same here. More expensive CPUs and GPUs are just going to mean more expensive CPUs and GPUs with no real silver lining.
> I suspect it's going to be the same here. More expensive CPUs and GPUs are just going to mean more expensive CPUs and GPUs with no real silver lining.
Presumably, higher profits will incentivize other players to enter the market and increase supply, and/or the company earning high profits plows at least some back into R&D to at least create better chips, which can result in eventually lower prices for the previous generation of chips.
This isn’t a possibility with land and land rights, however, so I wouldn’t expect the same dynamics to play out.
Over a long enough time horizon, yeah. Over a long enough time horizon the same happens in land too - we are finally getting decent amounts of infill development in the Bay Area, but it's 15 years after prices started to skyrocket.
Starting a semiconductor company and constructing a few fabs at cutting-edge process nodes takes at least that amount of time, so I'd expect semiconductors to have similar dynamics.
That's just delaying progress for everyone as TSMC basically doesn't have alternatives for the processes that are actually good. Definition of cutting off your nose to spite your face.
Yes. They should open a FabStore and make Apple pay 30% of their revenue.
>post decides to use the most crusty GenAI image possible
Man, what a cancer. Straight up using the bare TSMC logo here would work just fine.
The Nvidia 3000 series were on Samsung's process since it was quite a bit cheaper than the TSMC process at the time. Samsung has already said it intends to undercut TSMC so assuming they are fairly equivalent it doesn't seem unreasonable for consumer CPUs and GPUs to be made on their process, it gives a price advantage to anyone that makes the move.
Looks like I’ll need to put that home server together sooner rather later
Most consumers would do fine with like .25 Apple M1's worth of compute.
Price of new hardware increasing will stop old hardware from deprecating as much, meaning higher prices across the board.
I agree. Even my quad core i7 from 2011 is still fast enough for everything average consumer does. Heck, even average i5 laptop from 2015 or so is more than sufficient.
idk about that. I had an i5 2019 macbook and that thing was so incredibly slow and hot. When the M1 came out it felt like jumping 10 years foward.
I only partially agree. I recently got nostalgic and bought a 2nd-gen Core-i7 ThinkPad X220 from ebay (similar to what i had back in my university days).
It's generally very usable util you open firefox and browse a modern website. Chrome/chromium aren't any better.
The issue with old hardware is web browser and web pages, particularly web pages. Modern websites are incredible resource drains, it's unbelievable.
I can easily read documentation from old-school cool websites like https://pubs.opengroup.org/onlinepubs/9699919799/ but anything fancier than that gets the fan wheels spinning...
I don't notice any issues with an i5-6600k or an n100. Both are 4c/4t and extremely snappy. The only ways to get them above like 0.5 load are CPU intensive tasks like compiling, video editing, or games. The only thing I've found I can't do is real time viewing of 200 Mbit/s 4k60 10 bit 4:2:2 h265 off my camera because it's software decoded (software can only do ~half speed).
Does it have the Ultrabook version of the i7?
I disagree. I would submit that the average consumer utilizes a lot of computing power and that you might actually be using less than that as a more technical power user with more discerning tastes in applications and a viewpoint that sees the computer as more of a tool than an entertainment and lifestyle device.
I would also sumbit that the person who is still using a computer from 2011 would never have bought an i7 in the first place. They’re not going to follow an upgrade path where they say “I had a core i3 in 2011 that is too slow now so I’ll replace it with something that had a 2011 i7 that I couldn’t justify purchasing at the time.” Instead they’re more likely to say “I bought this laptop 15 years ago and I’ve battered the hell out of it and I’ve been frugal and it’s time for a new one,” and when they buy that new one it will also be a very modest configuration.
So the fact that a top of the line processor from 15 years ago is still serviceable is not really all that relevant. It’s going to not even be worth it on power consumption alone if you’re running it 24/7.
In other words, there’s more to a chip than raw performance.
Even the mythical "average user" uses tons of processing. Just "a few tabs" is like having several large applications open.
Hasn't been true for about a decade. Modern browsers suspend most processor activity of background tabs, so it's mostly just memory pressure.
This could finally force the developers of cross-platform GUI and webview libraries, drunk on Moore’s Law, to sober up.
decent chunk of RAM maybe but my parents are happily browsing away on decade+ old chrome books. Ordinary web browsing you can do on a potato machine from 2008.
Even if you're a gamer performance on the desktop has eclipsed demand so much that you can play anything on a few years old machine. It used to be way worse in the "olden" days when only 10% of people had a system that was able to play Crysis.
Most use phones.
Alongside presumably more RAM that came on a laptop with an M1. Certainly more GPU than came on a laptop with an M1.
This is what I don't get about laptop power consumption.
I would love an x86 laptop that can operate at 5W or less, even if it isn't blazing fast. If they can squeeze more performance out of a CPU for the same power budget, why can't they make a CPU with the same capability with less power use?
5W for a laptop is raspberry pi territory, and we're adding a screen to it. I hope we ever get there for a 14" fully usable laptop, but that sounds bleeding edge territory at this point.
The CPU alone can operate below 5W:
https://www.phoronix.com/review/lunar-lake-profiles/6
People say that, but money talks and the people don't buy those.
Honestly, that's not a CPU problem.
A laptop running below 5W needs to turn down its storage, RAM, GPU, LAN, WiFi, USB controllers, etc. The OS needs to act more like a smartphone, put everything to sleep, and use consolidated polling and interrupts for async behaviors.
I think this is the real advantage Apple has with their vertical integration. They can play a lot of complementary games between hardware and OS.
The steam deck's chip runs at 15w maximum and has I believe a 4, 6, and 8w mode. Maybe not 5w but pretty close. The linux desktop on it is a perfectly unimpressive but usable daily driver kind of thing. The problem is consumers dont want this. The failure of the netbook era proved it. People dont want to deal with slow speeds so this kind of thing wont be a mainstream mass produced item.
ARM faking x86 is probably the way to go for this case.
You should look into chinese laptops using the intel n150/n250 cpus. Those have a 6W TDP.
The issue with those cpus is that they're pretty much never used in laptops meant to do anything serious. They usually are employed into cheap plastic laptops using ridiculous batteries.
I would really like the form factor of 2011-2013 thinkpads, with the chonky 90WHr battery on the back and the 12" form factor.
With a modern cpu (low tdp and low power draw) that would be unbeatable everyday carry.
Isn’t the Apple M series basically the best of both worlds? Extremely high performance, and extremely low power use when not making use of it.
Have you checked RAM prices lately ...
I bought 32 GB of DDR5 RAM two weeks ago and it is $100 more already.
Thanks, TSMC.
Absolutely nobody stops you from forming a competing company and charging less for equally good or better results
Absolutely everything in the modern economy stops him from forming a competing company, this naive stupid market is the solution idea needs to stop being regurgitated, it's toxic to society.
I think most everything shows that it's just really hard (see Intel) and they're really good at it, so have a moat that they can take advantage of.
[flagged]
Chip making has very has barrier of entry. Between fab cost and access to talent you are really not looking at an enticing business opportunity.
Free market is not as free as the name would imply.
Freedom refers to a potentiality. You are conflating a potential (freedom) for a hoped-for benefit (highly competitive). A free market is a means towards a competitive market because it lowers the artificial barriers to entry into the market. Those barriers imposed by the universe simply are and while we can invest into R&D and education, market forces cannot change the laws of physics.
> Freedom refers to a potentiality.
And tariffs and sanctions refer to what?
> A free market is a means towards a competitive market because it lowers the artificial barriers to entry into the market. Those barriers imposed by the universe simply are...
Are you talking about tariffs or about banning the sale of semiconductor equipment to China, so they can't compete in the market for high performance chips? Is that "a means towards a competitive market" or is it "barriers imposed by the universe"?
That's an interesting universe you live in.
High barriers to entry have nothing to do with free markets.
>In economics, a free market is an economic system in which the prices of goods and services are determined by supply and demand expressed by sellers and buyers.
https://en.wikipedia.org/wiki/Free_market
As long as a buyer and seller can negotiate whatever number they want, or walk away, it is a free market.
From just a bit further down in your link:
https://en.wikipedia.org/wiki/Free_market#Low_barriers_to_en...
I am not seeing the discrepancy:
> A free market does not directly require the existence of competition; however, it does require a framework that freely allows new market entrants. Hence, competition in a free market is a consequence of the conditions of a free market, including that market participants not be obstructed from following their profit motive.
The barriers to entry being referred to in the Wikipedia article are presumably government related. No government is stopping Intel or Samsung or any of the other chip makers from selling what TSMC sells, hence the “framework that freely allows new market entrants” is there.
And the subsequent paragraph seems to be exactly what is happening in the market of high end chips:
> Hence, competition in a free market is a consequence of the conditions of a free market, including that market participants not be obstructed from following their profit motive.
You don't see a discrepancy between your claim that free markets have nothing to do with high barriers to entry and your own link claiming free markets require free entrance to the market?
And no, it is not specifically referring to government related barriers, although those certainly count. Government is very often the thing that generates barriers to entry, but it is by no means the only barrier to entry. Any barrier to entry is, from a strict theoretical sense, a departure from a pure free market, although that's more in the "spherical cow" sense of the idea.
You are using "free market" as a political term. In that sense, it refers to little more than to the absence of unnecessary regulations.
Economics is more concerned about whether "free market" is a useful description of the system. If there is nothing in principle to prevent the emergence of new competition, but barriers to entry (e.g. availability of capital / talent / machinery / raw materials, long-term contracts, or the time required to set up a new business) make it difficult in practice, the system may not behave like a market. Then you need to focus more on politics, both between and within governments and companies, to understand how the system is actually working.
Yeah I'll get a loan off Dad for my first EUV machine and we'll snowball from there
Judging by the response to the parent come back, evidently TSMC deserves the high prices it will be charging. Why?
It's unrealistic to expect just anyone to start up a new fab. But if not one person among billions will start up a new fab it points to intrinsic difficulty or unpleasantness or lack of prestige in the task. A correctively high price has all kinds of advantages for the society.
I haven't seen any argument that difficult tasks shouldn't be priced high. Only name calling.
In this case its more its a 100+ billion investment to possibly produce a competing fab and its encumbered with many thousands if not more patents the likes of which will produce a minefield for any entrant. So far the silicon product market has just shrunk over the years with more and more failing to keep up with the latest process technology as the price of entrance and the patent warheads has risen.
Entering a multi billion risky market is no easy task even if you have investira and teams with deep knowledge. That's the reason monopolise should be avoided.
TSMC isn’t a monopoly, they are just better than all their competitors. Nothing forces you to buy TSMC; you can buy Intel, Samsung, SMIC, GlobalFoundries and various others.
This is getting faded out, but this is absolutely right. For very few use cases do you truly need the bleeding edge. So many things do not have such strict requirements, and will meet all necessary requirements on an older node. An ATTiny85 is still an incredibly useful microcontroller even today.
This monopoly is unavoidable due to the limited number of people in the world with the know-how to make these products, with the second constraint being that you need to risk $10B+ and 10+ years with a risk that you fail.
Edit: per link below, seems like you need $100B+ and 10+ years
For reference, TSMC's yearly capital investment budget is about $50B/year (https://www.taipeitimes.com/News/biz/archives/2025/03/07/200...)
This business is expensive, making and running fabs is way way more expensive than most anybody thinks. There are very few companies able to do it profitably.
> due to the limited number of people in the world with the know-how to make these products
Absolutely nothing to do with this lol, and everything to do with it being ridiculously capital-intensive.
I'm rooting for Rapidus. A lot of people will look foolish if they succeed.
Great idea. Can you lend me $10 billion?
Piddling sum in this domain, that amount won't even buy you enough for 1 production line, let alone where to put in, staff it and get all the necessary collateral to allow people to design to it.
And no amount of money will get you one of ASML EUV machines any time soon as the entire production line for the next handful of years has already been spoken for.
I think that is a on the cheap side. Unless you are content with 1990s node sizes
Except for me. I am single handedly holding off all of these people from building a fab.