Why gamble? When you need it, and you have money making projects in queue, you buy it.
Yeah I have to say from a business standpoint there's no real great increase in speed to "gamble" over, but at the same time there's no reason to suggest that a faster computer really means more revenue.
I've generally found that we've not been constrained by the machines are work. If we rendered stuff faster we wouldn't be getting much more work. I can see situations where increasing your ability to churn through volume can pay dividends but I don't think that's a universal bottleneck tech solves.