- cross-posted to:
- pcgaming@lemmy.ca
- cross-posted to:
- pcgaming@lemmy.ca
Steam on Linux use in March hit an all-time high of around 5.33% marketshare. Since then it pulled back to around 4% while for September 2026 the Steam Survey results are now in and it’s clocking in just above 3%.
The Steam Survey results for September that were published this evening indicate a rather steep decline of 0.85%, down to 3.05%. Meanwhile macOS is at 1.92% and Microsoft Windows at 95%.
As with most months where Steam Survey indicates an abnormally large decline for Linux use, the second stop is checking on the language use… Once again a case of a big spike in Simplified Chinese use. Steam’s Survey reports a 7.42% increase in Chinese use, up to 31.39% overall of the Steam population. Usually with those large month-over-month swings where Steam China use shoots up dramatically is a boost to Windows and a detriment to Linux. Some attribute the large increases in Simplified Chinese use to being a reporting/sampling issue.
When diving into the Steam Linux data exclusively, CachyOS has spiked by another 2.65% to now account for 18.12% of the sampled Linux gaming population. That even outperforms SteamOS Holo coming in at 17.37%.
Among Linux gamers on Steam, AMD CPUs still account for the high 60 percentage space thanks to the Steam Deck and Steam Machine using AMD APUs and their open-source/Linux support being popular with enthusiasts.
Those wanting to dive in to all the Steam Survey stats for September can find the fresh data on SteamPowered.com.



I still don’t get why I can’t opt in to participate in the survey every month, I’ve only been asked twice in the last three years on Arch
Because thats not how good stats works. They want random samples to draw actual stats from.
Volunteering ruins the point.
Yeah, you have a point, but still, twice in three years? I don’t know how representative it really is, ESPECIALLY when you see 50-100% ups and downs in the Linux user base. 5 to 3 percent and back cannot be the actual stat, that has to be unreliable selection of users.
steam has at least 130 million monthly active users
a margin of error of 1% (resulting in a range of 3–5%) is very, very good
They could just pull the stats from every system
privacy issue
censuses only have marginally better results than random sampling
Probably. But it’d 50x the cost of the tool. And for only marginally better results.
Would pushing a request through steam to participate in the survey realistically increase cost?