Steam

Not at all sure if it's true. But based purely on observational data, spikes in simplified Chinese tend to be short-lived and tend to occur right when windows percentage spikes temporarily. Internet cafes using Windows certainly would be a valid explanation to the observations.

To be clear, it's probably also true that people in China likely use Windows more. But that only partially explains our observations. Namely it doesn't explain why these spikes are temporary and tend to only last 1 month. More likely, these spikes are caused by a singular event (or a small handful of events) that skew the data heavily when it happens. For instance, if a large number of people all log in to Steam on a computer that they don't typically use, that could lead to the data that we see. Internet cafes would be consistent with this model, but we can imagine that other hypotheses may also produce similar results