Answer by Dennis J Frailey:
Short answer: because large companies have to produce products that perform well for lots and lots of real customers.
True story: when I was a young software developer, a group of my colleagues quit my large company to form a startup. Their product was a computer aided design system and their technical advance was to do it all in LISP on a LISP machine because it could be done in such an elegant manner. The product had a lot of very attractive features. It was truly a better mousetrap in terms of features and capability. But the company went out of business in a few years. The problem: their product was 1/50 as fast as the competing products that were written in conventional languages. They couldn't sell any.
True story #2: my large company hired a group of AI experts to develop a capability that was especially important to a particular product line. Unfortunately, their solution was also written in LISP for the then-popular LISP machines and was both slow and incompatible with conventional computers and languages used for the rest of the product line. The application was on an aircraft, so you couldn't just use a bigger, faster computer (which was always the solution proposed by the AI experts when you pointed out that their applications were too slow). There was also a problem of finding people who could do maintenance on the software, although that was a solvable problem. I was involved with converting it to something that was suitably fast and maintainable. When we were done we had replaced a lot of LISP code with decision tables and other rather conventional mechanisms that ran on conventional computers and languages and was very fast.
Why do startups always use slow language such as Python and Ruby but large company uses fast language like C++, Java and Scala?