Steam machines didn't take off because there were no games for it [1], the price point for the pre-builds was too high [2], nobody wanted to buy another (inferior) PC just for gaming [3] and the Steam Controller that was billed as part of the package was an experimental concept rather than being something people could pick up and play [4].
All of these kind of made sense at the time, but kind of felt like they had the signature Valve over-engineering. They have learned from this and have done the following:
For [1]: Proton
In 2018 in on fell swoop, Valve brought thousands of Windows games to Linux. Basically everything other than Multiplayer games with EAC or Battleeye work now (and EAC has a Linux version for native games).
From this point on I'd say that even though Linux still isn't 'the future of gaming', it's definitely a contemporary option. There are limitations, but it's viable.
For [2]: Scrap steam machines
It was a good effort, but I don't think Steam Machines will happen unless they can go toe to toe with the consoles. They don't need to, though, because people who want to run Linux are happy installing it on kit they already have.
For [3]: Steam Link
They released Steam link at the time, but now there isn't the competition of Steam Machine vs. Steam link - you can just run your Linux desktop and stream to your TV or to a raspberry pi.
For 4: Controller support with Steam Play
Steam play now has exceptional APIs which make it easy to play with any controller. I do think they need to release a Steam branded 'boring' xbox style controller, but I think they'd wait until another Steam OS version before then.
They've also invested massively in surrounding tech such as KWin, DXVK and a bunch of others. Linux may not be the 'future' of gaming, but it's definitely going to be a valuable safety net if Windows or the Consoles go rogue.