If you're just treating it as a micropayment system, a hypothetical far more efficient blockchain service could potentially be cheaper than building a pure-custom system but it's unlikely to be easier than Steam/Apple/Google/Stripe/etc. and there's a limited margin for how much cheaper it could be (those players have plenty of room to lower their fees before becoming unprofitable). The problems with micro-transactions aren't really the overhead of processing charges.
The deeper question about this concept is about the idea of interoperability. Payments or authentication are comparatively simple problems which are well-understood. If I use Autho0, I'm cutting out the most expensive part of that problem and my app change are minimal — before I verified that you could receive email at you@example.org myself, now I get that as an attribute from the SSO service, but it's always been my unique identifier.
In contrast, integrating assets into games is a much harder problem without a turnkey solution. Putting an asset into a game requires the artwork, sound, etc. to be adjusted for the game engine (this is far more complicated than just resizing a JPEG since they need to adjust things like 3D meshes, lighting, etc.), custom behaviors implemented (does anything light up or respond to conditions?), if does anything which affects gameplay that has to be implemented and balanced against the rest of the game (you're going to regret paying extra if your favorite weapon is weak; nobody else will want to play if it's over-powered), and while those are substantial amounts of work they're not even the hardest part: the game developer needs to have a license to do any of that. You're not going to see Mario in Call of Duty because Nintendo doesn't want their billion-dollar franchise used that way. Gamers can't even resist advanced purchases for games of unknown quality, there's no chance that they're going to stop buying games until the publishers grant general access to assets or do so at less than usurious rates.
Now, look at it from the perspective of the implementer: if you want to take an asset from game A, you already spent the money buying it in that game. I can spend a lot of time and money licensing & developing support for that asset in my game in the hopes that it'll make you more likely to buy my game but that's a big upfront cost which limits my returns to the person or, hopefully, people who hold that particular NFT — or I can spend that time and money on things which I can sell to everyone, and don't require ongoing licensing fees to the developers of game A.