12% does seem a bit slim I agree. OTOH 30% is greed. It's not just about store features though. I've asked before here if another store showed up tomorrow with the same features as Steam, would people with 10+ years loyalty, payment info already stored, library of hundreds of games, social network, would they switch stores? No, they wouldn't. That's a huge moat.
Wrt how big a cut you think they need to develop features, I would argue that's not significant relative to the revenue. Way back in 2011 Valve were more profitable per employee than the big tech companies. Last year MAU grew by 26%. Sales revenue via Steam is probably, what, $10BN per year? Of which they keep up to 30%. I doubt they spent much on Steam Input last year. Whatever features they finished in 2018 and 2019 those people likely worked on new features for a similar cost, meanwhile MAU keeps growing as the market grows. I bet most of the staff could have taken the year off and MAU and revenue would still soar.
You need to provide a better service when you enter a new market against someone with a natural adavantage due to market tendencies. Nobody has really tried to do that, instead relying on providing a worse service and expecting to be rewarded for it.
The only real challenger that has actually tried to enter the market in what is normally a "natural" fashion, by providing something new has been MS with Gamepass. GOG tried, but their main selling point (no DRM) has a natural flaw due to not being that interesting for bigger devs (who end up driving a ton of the MAU).
MAU comparison on 2020 are funny when you consider that their main "big competitor" on the scene (EGS) saw their revenue stagnate in comparison (with MAU increase tho, but also very much on a spotty up and down MAU compared to Steam more constant increase). That is probably related to constant tweaks of the formula and avoiding stagnation (such as the big reworks of how sales are presented in 2020 and the new Steam Points shop to encourage more people to buy shit).
I would also say that yes, the main cost of new tech is very frontloaded but it still creates parasitic losses that makes it so that "the lowest you can go to" is higher.
Regarding the "profit per employee" statistic, it is really a bit meaningless on this kind of talk about "what cut is enough". Valve owns and operaties 2 of the bigger GAAS games righht now (CSGO and Dota 2) that could easily make them already extremely profitable. They also have a really small amount of employees (mostly relaying on subcontractors and consultants which arent counted on that statistic). That means that even if their profit margin is small (due to investments, natural expenses, cost of subcontractors, etc.) they only really need to have enough volume to have a ludicruous "per employee" statistic. They could have a real profit margin (profit/revenue) of 5-10% and still outprofit per employee Facebook and Epic whose profit margins are much bigger.
(Mind you, I dont think their profit margin is that small, but I dont think anything below 20% Steam cut would probably be viable with the current Steam regional payment structure).