The problem is code is a lot more context sensitive than most capital expenditures. If I pay for a machine that makes really good chalk, then when my company eventually folds that machine is still probably worth a fair amount to my competitors. In contrast, the code I wrote today which probably is going to save my company ~200 man/hours a year is almost certainly completely and utterly worthless to literally anyone else, because it automates a hyperspecific piece of a company-specific workflow.