> or what is the price of a cutting edge fab these days?

Intel says they're going for two factories of 20 bln indeed [1], Samsung for 17 bln in Taylor, Texas [2], and Micron claims to go for 100 bln (over time) in Clay, New York [3].

> Why invest 20 billion dollars [...] to create supply without demand?

Why do we have to pay about 5 dollars per month for a VPS with only 1 vCPU, 1 GB RAM, and 10 GB storage? I'm certainly hoping these specs to all increase 10-fold over the next 10 years for the same price.

[1]: https://www.intel.com/content/www/us/en/newsroom/news/intel-...

[2]: https://news.samsung.com/global/samsung-electronics-announce...

[3]: https://investors.micron.com/news-releases/news-release-deta...

Re your last point: cloud specs are already far beyond that. Cloud companies just pocket the difference. Cloud lets hosting companies benefit from Moore’s law, not you.

Look at the machine you can build for one months’ typical AWS cost for a medium size SaaS company.

Cloud also charges insanely high rates for bandwidth.

I doubt that. I was not talking about AWS. The price I mentioned is from Hetzner (by heart though, so I might be a bit off) which is pretty cheap. I have also tested multiple budget VPS providers and they all don’t dare to go below aforementioned price even though there is a lot of competition in the VPS market. Sometimes the more budget providers provide more vCPUs but in my tests those usually turn out to be extremely slow.

Well, it's 2GB of RAM, and 23GB of disk.

It's quite low, but the real costs are on datacenter space and connectivity anyway, I have no idea what their cost structure looks like.

I would expect any real user to switch into renting servers as soon as small VPSs aren't enough. (But yes, the fact that there is a market of large VPSs tells people don't to that. I don't think I will ever understand this, as I don't understand most people usage of AWS.)

Most buyers are simply not savvy enough to pick the most effective hosting. I worked for a small e-commerce operation years ago, they had essentially no technical expertise in-house, but they knew their products and their market. Odds are very low that their VPS arrangement was optimal but how would they know? As long as the site stayed up and the orders came in.

> Or do we seriously expect that US companies will generate significant demand

Yes I think so, presuming that there is not an over-supply of capacity.

US companies would much rather rely on an EU country than one that is being threatened with invasion over a small gap of sea. US local supply will never be enough.

> US companies would much rather rely on an EU country

Agreed

> US local supply will never be enough.

But you won't just have US supply. You'll have US and Taiwanese supply, and I don't believe that Taiwan will happily let TSMC (the only cutting edge foundry left in the world, when you take Samsungs abysmal yields into account) build foundry redundancy in the western world.

But we'll see, you could definitely end up being right. I just hope we'll invest at least an equal amount of money into chip design.

Why not invest in both?

Besides, the cost of designs pales in comparison to the cost of producing locally.

ARM came out of europe. With the right industrial policy, chip manufacturing could be onsourced. The dutch already make most of the equipment that makes the fabs/chips.

It's not necessarily about supply and demand. These past few years have shown what shortages of chips can do to the supply chain. It's a strategic vulnerability if Europe does not at least think about this.

I guess it depends what fabs you build but we have seen very large numbers of manufacturers desperate for components. A fab anywhere in Europe could easily supply any factories in Europe so there should be demand.

On the other hand, if people are trying to build the cutting-edge, there might not be as much local demand since it is probably only needed for the latest IT equipment, most of which is built in the Far East.