The number of programmers is pretty elastic. People can and will learn to program if it starts to seem like that's a valuable skill. E.g. there is a founder in this current YC cycle who came to the US to go to business school, but then realized that if he wanted to start a startup he'd be better off spending the time learning to hack instead.
I'd say the upper bound on the number of startups is the number of people who are determined and smart enough. That's probably between 1 in 100 and 1 in 1000.
People can and will learn to program if it starts to seem like that's a valuable skill.
But they won't tho', not really. In the dotcom boom there were plenty of such people who learnt HTML and were (well) paid for it for a couple of years. But they didn't or couldn't make the jump to possessing a skillset that kept them in the industry after the meltdown. The guys with C++ or Sybase or whatever under their belts, just kept right on truckin'.
Hacking is starting to be a new literacy. What's more, the most important part of startups isn't really programming, which is getting easier and easier, but understanding customer's problems. Here a wide range of backgrounds are useful.
"People can and will learn to program if it starts to seem like that's a valuable skill."
That's basically just a restatement of what I said (more people will learn to program if there are greater rewards for doing so). The thing is, there aren't greater rewards for doing so. Startups aren't any more profitable now than they were ten years ago, and programmers at big companies don't make more now than they did ten years ago (inflation-adjusted).
"I'd say the upper bound on the number of startups is the number of people who are determined and smart enough. That's probably between 1 in 100 and 1 in 1000."
Few could argue that, say, investment bankers are quite smart and extremely determined (and very good at sales besides), but there's extremely little overlap between investment bankers and technical startup founders, and that doesn't look like it's going to change. Same for corporate law, management consulting, etc. Plenty of i-bankers and consultants and corporate lawyers go into other fields, including non-technical startup positions (eg. Jessica Livingston). Indeed, one of the primary reasons people go into these industries is for the exit opportunities. But very few seem to learn how to code.
Maybe I should clarify what I mean by "upper bound". "Upper bound" means "how high could the number go if parameters X, Y, and Z are allowed to vary", but neither of us specified what X, Y, and Z were. The hypotheticals I was considering were "suppose the startup community all acted in unison towards increasing the number of startups". For instance, if the startup community exerted enough collective pressure, we could plausibly get Sarbanes-Oxley loosened, or force VCs to become more technically competent, or make it easier for founders to immigrate. But we can't just decide to increase how profitable it is to learn to program or start a startup; these things are driven by consumer demand.
I think this number also depends on the bound on markets that can be monetized, when you see the scale that some startups need to achieve to become profitable their is only really room for a finite number of successful startups in any one market space.
I'd say the upper bound on the number of startups is the number of people who are determined and smart enough. That's probably between 1 in 100 and 1 in 1000.