As far as monetary value of the IP/source-code I agree, but I know one failed company whose founders are now extremely regretful of the whole situation, because they produced a bunch of interesting stuff that's now locked up. They'd like to open-source it so what they spent 7 years of their life on can at least influence other software, get them some recognition, and maybe be used by someone, but the investors aren't willing to, and aren't willing to sell it for anything close to the market value (which is pretty close to zero). There isn't really any prospect of a buyer, but from the investors' point of view, if they keep ownership of the IP of failed companies, they might get lucky with one of the IP bundles turning out to be valuable. Even if the chance is very small, it's at least a better chance than the guaranteed zero chance they'd have if they open-sourced it or sold it for peanuts today.
At Sonicity, we built a content distribution network in a box, optimized for RTP-style streaming delivery and FEC file transfers with FLID/DL-style congestion control (we did this before there was a FLID/DL, so I'm kind of proud of it). I owned the "router" component, which was tens of thousands of lines of Quantify-optimized asynchronous C++ and ran multicast routing over an IS-IS-style link state routing protocol (the product started out as "IRC without netsplits" and, uh, metastasized).
In the last 6 months at the company, we managed to refactor the design so that the hairiest routing and group management code could get hoisted out of C++, and most CDN participant hosts could just run a tiny stub forwarder that was simple enough to be kernel-resident (a design I awesomely rejected when one of my cofounders proposed it at the start of the company).
The whole source tree was many hundreds of thousands of lines; it was cross-platform Win32 and Linux with graphical clients on each platform and IE integration. There was a lot of good stuff there. My friends Danny, Kneel, Andy, and Tim all worked on it at various times and they are, unlike me, ridiculously good.
Poof. All gone.
This is what happens when you take VC and your company fails. What's better is that by taking VC, you drastically increase your chances of failing; VCs need you to shoot the moon. Our VCs decided what we should do with all that code was to go head to head against Akamai.
You may get lucky with a very, very cool investor who gifts your IP back to you. But you're not entitled to that and you probably won't get it.
This is a good argument (from the hacker-founder point of view) for open-sourcing all the non-`secret sauce' parts of the code from the beginning (i.e. from before you have any investors with the power to stop you from doing this).
