Valve would like to disagree with this sentiment ;)

Sure valve isn't making GPUs or CPUs, but they did make the steam deck, in addition to the largest online game store, and several high quality AAA games, and the valve index. They apparently have around 250 people according to a quick Google search, but that still goes against the mindset that you need 10s of thousands of employees to make a successful large scale tech product (whether that's software or hardware).

A 250 person organization assembling semi-off-the-shelf components and gluing mostly-preexisting software together is a very different proposition than 20-50 people building a modern CPU from scratch. That's an order of magnitude more complexity than the steam deck, and an order of magnitude fewer people.

Right. It means there's a lot of things that don't get created--or companies cooperate in some sort of fairly tightly coupled network that doesn't look that different from one big company. (But probably less efficient if you follow Coase.)

How many people designed the CPUs we use? I read a few days ago that due to deadline pressure, a single man did most of the work on (I believe it was) the 8080. But that's obviously a "trivial" design compared to the monsters we have today.

One specific example you agree isn’t even a good comparison does not make a case.

Well you'd have to imagine a lot more collaboration between companies, so say company A just markets CPUs in the US nothing else, while company B works on just shipping logistics (to stores which handle the actual delivery to end customers). I think the real problem is that so many boundaries between disciplines make novel collaboration strategies between the parts much harder. It's difficult for say a marketing person to suggest a change to the box design

I don't see how those small companies would make money

Some of those businesses arent really viable, meanwhile other are really good

I don't see it working with them not being together, working as teams under one budget, so just as normal company.

So, Intel would be split into ~2400 50-person companies. Coordination of those, through contracts, threats and lawsuits would be nightmare. A basic economic theory ("theory of the firm") says that large companies are created, because they decrease the friction that exists when many small companies (or even just individuals) cooperate to achieve a large and complex thing.

I think I agree it would be worse, though not without advantages. It’s really hard in the normal economy to work out how much a subunit of a company is contributing to the overall success and it can be hard to contain the urge to exaggerate. I think this is why big companies grow inefficiencies and at least with many small companies the individual parts could compete and occasionally fail.

Presumably you would, only what are different departments within a single company now would be separate companies. Probably similar kinds of blessings and curses as microservices.

The once CEO of Sears tried splitting the corporation into thirty to forty companies: https://www.versobooks.com/blogs/4385-failing-to-plan-how-ay... (I recall reading a more entertaining and detailed write-up, but this is what I could find on a quick Google Search)

In short, turns out aligning thirty to forty competing profit motives is hard.

You can work together with other organizations (but that quickly becomes a sliding scale towards megacorparations)

Exactly... just as 50 people can not design, build and test a 747 jumbo jet.

Good point. Maybe companies working together? Like teams in a larger corporation.

Wouldn’t companies collaborate with shared interest and risk?

For me, it feels like corruption would be more the risk from a cap.