Initially Valve's Steam Hardware & Software Survey for December 2025 showed Linux at 3.19%, but they appear to have amended it with a nice boost for Linux.
You just described Sigmoid curves, roughly speaking. The only issue is your incorrect use of “exponential”.
The idea is that it’s not exponential for two main reasons:
It caps at 100%. You can’t grow infinitely.
You also need to consider the reverse: going the other way, going from 99% to 98% is a ~1.01% decline. Going down from 2% to 1% is losing half your remaining users. That’s huge.
Exponential growth is used colloquially for any situation where there’s an upward curve to the trend; in calculus terms, the second derivative is positive. But there are a lot of functions with that property, and exponential functions are only 1 type. Sure, it’s a common one, but so is parabolic, cubic, and other polynomial functions; a variety of trigonometric functions (over certain domains, like sine from -1 to 0); rational functions (again, over certain domains), etc.
Sigmoid curves (colloquially known as S-curves) are very common in any situation where there’s both a contagion factor (like popularity, word of mouth, network effects, etc.) and a limit on growth or maximum carrying capacity. The later is always the case when your function maps to percentages of a population since it caps at 100%.
That doesn’t actually make much sense for Linux adoption stats though
As linux gets bigger, more people will hear about it and consider it, there will be more pressure for Linux support, generally more focus on Linux
It seems like it will indeed be roughly exponential, until it levels out sigmoidly near 100%.
You just described Sigmoid curves, roughly speaking. The only issue is your incorrect use of “exponential”.
The idea is that it’s not exponential for two main reasons:
Exponential growth is used colloquially for any situation where there’s an upward curve to the trend; in calculus terms, the second derivative is positive. But there are a lot of functions with that property, and exponential functions are only 1 type. Sure, it’s a common one, but so is parabolic, cubic, and other polynomial functions; a variety of trigonometric functions (over certain domains, like sine from -1 to 0); rational functions (again, over certain domains), etc.
Sigmoid curves (colloquially known as S-curves) are very common in any situation where there’s both a contagion factor (like popularity, word of mouth, network effects, etc.) and a limit on growth or maximum carrying capacity. The later is always the case when your function maps to percentages of a population since it caps at 100%.