That's right. But while Amazon and Apple probably are not going to grow 5 times either, they have a proven business model and they are unlikely to lose more than 50% in value. The same cannot be said about Facebook that depends for revenue heavily on a few partners. Therefore the risk/reward ratio should be much better for Facebook (to get me interested).

I don't consider a factor of 5 or 6 extraordinary given the risk. Take a look at Intel or Amazon for their growth since they IPOed.

At a valuation of $75B to $100B, how much can we expect Facebook's share price to grow? Apple's valuation is now at $400B to $500B, so maybe 5 to 10 times at most while there is a considerable risk for not much growth at all. The share price of Amazon and Apple grew over 100 times since their IPO but this can't happen with Facebook because their IPO is coming fairly late. So, while the IPO is a great day for early investors, it's not worth it for average investors.

The problem with most modern code generators and VMs is that they are tied to a specific way of doing garbage collection and exceptions that might suit some but not all languages. If you decide to provide none of these runtime services you must provide an API such that language designers can implement their own. However, designing this API is also quite difficult. I looked at the TuboVM pages but found no design goals or principles that would explain this but also did not want to dig trough the code. Would you mind to explain them here?