I'm not blaming the author, because determining causation is hard, and checking there's correlation between two variables can still be interesting, but the article falls short of answering the question.
> We can see a correlation between GPU prices and the value of crypto currencies, but correlation does not mean causation. A causation is plausible in this case.
And then they go on to repeat a very reasonable argument, which I happen to believe to be true, but the kind of argument that everyone was already making before looking at any data.
The fact that the correlations do not hold as much for CPUs is good, but still not enough to stablish causation. As you said, there might be some confounding factors that affect crypto and GPUs but not CPUs.
if those non-crypto factors were confounding (say also happened to correlate with crypto-prices maybe due to some third unknown factor) then we would see CPU and GPU prices rise in response to crypto prices. we do not see that.
This. Those who can't afford anymore powerful video cards now fall back to CPUs with decent integrated graphics, so that the higher than normal demand is therefore affecting their prices too.