that's asking the cart before the horse; is there any data on what compute actually results in real GDP improvements?