That's because even with their cut the price consumers pay stays reasonable most of the time (and there is the second hand market). Thanks to sales and events you won't be paying 60 bucks for a 5 year old game. But thats how it works in the Mac App Store (I guess we should be thankful its not outright subscription yet). The reality is that developers follow Apple commercial practices on their platform (can't blame them) and it doesn't lead to an interesting competitive market. In fact I would argue that most peoples are getting priced out of the Apple market, they just didn't realise yet. It seems that their strategy is making and accommodating only luxury/premium priced products, so they are going to make a good amount of money for a bit but in the long term their technology will become irrelevant. And this is exactly what's happening with their M silicon despite what the general sentiment seems to be. Apart from lower power they are not actually that impressive in real world tasks and particularly not in gaming (optimised or not).

In fact after switching to their own silicon, Apple has mostly increased the price to access a particular performance level, and I would argue that previous Intel machines were better gaming machines at equivalent prices points. Without concerning yourself about prices one can see that in best case scenario (taking Metal benchmarks for reference and extrapolating when needed) : M2 can be about as good as an AMD Ryzen APU destined to notebook/laptops (6800U) M2 Pro seems to be in between a RTX 2060 and RTX 2070 and it would be a slightly slower RTX 3080 laptop. M2 Max can be slightly faster than a RTX 3060 but slower than a RTX 4060 laptop ; it seems to be equivalent to the Radeon RX 7700S (AMD describe it as a notebook part...). M2 Ultra is irrelevant because it doesn't provide real improvements in graphics performance.

Realistically the only way to get this low level of performance would be to buy a laptop (not too high end) or to go for pretty old or entry level desktop part.

Now if we talk about AAA games, M2 is pretty much useless, it is somewhat better than previous Intel iGPUs but that isn't saying much ; you still won't be able to run anything graphically intensive at 1080P in an enjoyable manner. M2 Pro could potentially run some of the less demanding games with a decent experience (around 60 fps stable) but it starts around 1.5K in a desktop form factor (naked, no K&M nor display) yet it perform worse than cheaper 3050Ti laptops (that have all those things !). In laptop form factor it starts at 2.4K but is generally much worse in gaming performance than a similarly priced/specced PC laptop. And it feels that talking about M2 Max in any form factor is a major waste of time because it starts at 2.4k for the naked desktop but you can buy a better performing (for games at least) desktop PC and still have money left over to buy a Macbook Air (not even the cheapest version). So unless you really need that power for a macOS only software that properly use the new arch it doesn't make any sense. Not only very few will buy but they won't even think about gaming on it. It is unsurprising, after all the M2 Max has barely more Gflops than the 550$ Xbox Series X...

Before their own silicon Apple was not particularly good at gaming (especially on the entry level) but at least if you were willing to pay you could access a decent level of performance. It wasn't very competitively priced but the premium was worth all the other benefits. For example an iMac 27 inch with a RX 5700 XT would cost you over 3K but it was worth it for the whole package (especially the display). Now even if you want to spend 3K on a mac, not only you won't get any package at all but the graphics performance will perform worse than this 3 year old iMac. And that is before talking about architectural optimisation problem and general support.

So I believe Valve sees that and decided (rightfully) that it wasn't worth their time. Previously it was just a matter of making a build that could run in the macOS GUI and it wasn't too much work because both GPU and CPU APIs could be almost identical. I'm sure there were some differences at compile time but it didn't ask for a large parallel port and a decent amount of customers could get a pretty good experience (high end Macs and eGPU users). But now they would need to maintain a parallel port that would diverge in significant ways (even if we just account for the GPU code) and on top of that even the most high-end customers would only get a performance that would be considered ok for mid-range/high-end laptop. It is just not worth it.

Considering that the 3nm A17 Pro isn't really the revolution it was touted to be ; this is probably the right move. Not only Apple hardware is becoming irrelevant to enthusiast but it seems it will not improve for a while (maybe ever) considering the A17 already has heat issues even though it isn't a whole lot faster it looks like M3 chips will run hotter and use more power without really closing the gap.

Reality is catching up to the Apple arrogance, you can buy monopoly power on a technical advantage once but eventually competition is going to come around. Now is the time and I think it is only the start. Most cross platform software that needs a significant rewrite won't get ported to Apple Silicon, it doesn't make sense considering the weak hardware. They'll come and say : but it doesnt use a lot of power. Yeah I would be pretty mad if my slow small car would use as much gas as Porsche...