Since almost all of the points lie flat near y=0 for x≤1, then only one or two points drift upward—and one extreme outlier—there isn’t a single straight line that captures the bulk of the data. Most points would lie far from the regression line you’d draw through that big spike.
In other words, the scatter doesn’t hug a line very tightly, so the correlation coefficient r will be small in magnitude.