Earlier quoted context omitted.
> Economies of scale in commodities is the thing that makes it so we don’t have to do sustenance farming and everyone gets to have sewn clothes on :) There are really two different kinds of economies of scale. One is the one you ordinarily think of, where you're big enough to amortize certain fixed costs over a lot of units so the fixed costs become a small part of the price. But the scale you need for this isn't all…
I agree in general, but note some industries are exceptions from that rule. For example, for Boeing and Airbus these two scales are pretty close because modern airliners are very expensive per unit, and fixed R&D and tooling costs are sky high. Another example is TSMC. However, none of that applies to internet search, social networks, web services, and most other goods and services.
Not as many as you might think.
> For example, for Boeing and Airbus these two scales are pretty close because modern airliners are very expensive per unit, and fixed R&D and tooling costs are sky high.
Things that are very expensive per unit are where you don't need economies of scale, because the unit cost can absorb a lot of bespoke custom work.
What's going on in the airline industry is regulatory capture. Instead of training pilots how to fly "planes" and requiring planes to have standardized controls, the law requires them to be trained on a specific plane from a specific manufacturer. So the law requires the pilots to have "economies of scale" in order to make new design viable, even though a given pilot is generally going to be flying the same type of plane all the time or only one of a very small number and individual pilot training is extremely expensive (also as a result of regulatory rules).
The incumbents like this because it makes it hard for anyone new to enter the market because customers don't want to recertify their pilots on a new type of plane. The rationale for this is nominally safety, but it's not accidentally structured in a way that creates a regulatory barrier to entry, and it's the thing that led to the 737 MAX debacle, which is not an instance of safety occurring.
You also have design vs. manufacturing. Boeing doesn't have to be a vertically integrated conglomerate. One company could design a plane and then license the design to arbitrarily many others to manufacture them. The manufacturing companies themselves wouldn't have to be vertically integrated, you could have separate companies each making screws and fuselages and doing final assembly.
The designs could be protected by patents which then expire, so that anyone could manufacture the already-certified design of a plane once the patents expire.
This is all happening the way it is because the laws are purposely structured to make it happen, not because of any intrinsic characteristic of the underlying economics.
> Another example is TSMC.
This is plausibly the best real example, but it's also kind of not. TSMC currently has the best process, but it's only marginally better than Intel or Samsung. There are instances where that matters, but even more instances where it doesn't, and it wouldn't be two years of TSMC resting on their laurels before they were overtaken. And there are a lot more companies in semiconductor manufacturing than even those three: Global Foundaries, Nanya, SK Hynix, Micron, SMIC, Kioxia, etc.
What's really happening there is that it's a highly competitive industry where spending more on R&D yields an advantage. That isn't exactly economies of scale, it's more like the other way around. You get lots of scale if you have a competitive process that customers want.
Conversely, you get a competitive process by spending a lot of money, which is fungible and doesn't inherently need to come from having any scale at all in the existing industry. People speculate about Apple designing their own semiconductor process even though they don't currently even have their own fabs, and they could viably do it, not because they have extensive existing internal fabrication volume but only because they have money.
And Apple gets their money from having a lot of scale in another industry, but you don't even need that. All it would take is a charismatic CEO capable of getting the investment spigot open and then the money comes from telling a story to investors rather than any existing business of the company.
Compare this to, say, auto manufacturing. There are typically only one or two fabs on the world's best process node at any given time and if you built one you'd be the new TSMC as fast as you could tell anyone about it. Whereas Tesla has car designs that people are willing to buy, but it will probably be a decade from the release of the Model 3 before they can manufacture as many cars as Toyota does, and as a result they currently have to amortize their R&D over fewer units.