If you click through to the link, you'll see that the results add up to 150%. That should be a sign that something odd is going on with this survey result and we need to interpret it with caution.

The question was "What is the primary operating system in which you work." (Emphasis mine.) The question itself implies that we should only be able to pick one, and yet the results clearly indicate that more than one was allowed. Most likely, people were listing the OSes they use a lot. A great many of us developers use a commercial OS locally, but also interact with Linux servers or containers on a regular basis.

That said, there's no simple interpretation of that "40% use Linux" number that passes the smell test. It's implausibly high for the number of people who use it as their primary workstation OS, and it's implausibly low for the number of developers who regularly work in Linux. More likely, what this number represents is that different respondents interpreted the question, which appears to have been ambiguously framed, in different and incompatible ways, and so the result is statistically meaningless.