27% of that 3% is the Steam Deck / Lenovo Legion Go S. So most Linux players are in fact not on the Steam Deck.
If that is true then one of those other two claims has to be false:
1. Using the latest months recorded share (Oct-2025 - 3.05%): 4,026,000 estimated "monthly active users" for Linux+Steam.
2. Market research firm International Data Corporation estimated that between 3.7 and 4 million Steam Decks had been sold by the third anniversary of the device in February 2025.
27% of 4M gives us 1M Steam Deck + Legion users. Yet 4M were sold. That begs the question: How could it be? Do 75% of Steam Deck users run Windows? Have 75% of Steam Decks ended up in the landfill? Are the sale figures estimates wildly off-base?
- International Data Corporation is overestimating the amount of shipped Steam Decks.
- Modified Steam Decks (i.e. running Bazzite) don't report themselves to be Steam Decks
- Most likely: most Steam Deck users opt out of participating in the Steam Hardware survey/analytics.
Last year, I didn't participate in the Steam Hardware survey on my Deck, only on my PC. This year, I participated on my Deck and my desktop, but not my laptop. I still have three devices running Steam. To any survey, it'll look like the amount of Steam devices doubled even though I'm only reporting 67% of my devices to analytics.
- may depend on the period being measured.
I haven't looked at the article or their methodology, but if they were measuring over a certain period of time, a few hours, or even 24 hours, it will still likely only pick up a proportion of Steam owners.
I also have steam on multiple devices including a steam deck. On desktop I'm pretty much always logged in and I play games frequently, but most months I'm not selected for a survey. I use my steam deck less frequently and have maybe only gotten the survey prompt on it once or twice.
The survey does have other statistical noises/biases/errors though - but none nearly as large as this "gap". E.g. internet cafe reuse of the same machines by different accounts means if you're hoping for a "hardware popularity survey" instead of a "hardware users use popularity survey" then the number won't always make sense.