Payment processing is a big part of it, but it's not just that. The deployment framework, testing branches, the workshop/user content/mods support, etc. That stuff could certainly be replicated, but the exposure to the existing user base of active customers is the tough one.
Admittedly it is supposedly lower value now, with Steam Direct replacing Greenlight, but it shouldn't be overlooked -- it's the same reason that tons of products are sold on Amazon even though you can also purchase the same product directly via the brand's website...
FWIW, I'm lead dev of SimAirport[0] which is in Steam's Early Access program. The 30% has been tough to swallow, but without Steam (and given our lack of internal marketing/spend) we probably wouldn't have sold a quarter of what we have thus far. And while we wouldn't qualify for any of these tiers had they been in place previously, it's still a solid upside "just in case" your game blows expectations out of the water.
I'd love to see them add a tier around $2-5M at 27-29%, even if just as a "bone" to indie developers; it'd make developing a 'game as a service' (ie long & consistent feedback-driven dev cycle) more palatable and likely result in demonstrably better games.
Tangentially related at best, but sometimes I wonder what Steam _actually_ optimizes for -- if their algorithms always show the "best" games (ie ones that players spend hundreds++ hours playing) then I'd expect that to actually hurt their revenues short term. In theory they'd be better off marketing/optimizing to push games that satisfy users but that have low play-time/hours on average, so that they can sell the player a different game that much sooner. It's a strange set of incentives & it's hard to tell how aligned they really are (with either players or developers).
[0] - https://store.steampowered.com/app/598330/SimAirport/