To keep it simple let's invest $100 in each of those 40 companies. He defines success as companies that have grown at least 100x over. We know 5 companies have hit that mark and know basically nothing about the other 35 that we'll just consider failures (even though they might only be 10x or 50x[a success in other areas for sure]).
Failed$ = ($100x)35 x is the multiplier on that money across all 35 companies. Perhaps he lost it all and it's 0, maybe it's slightly above 1, who knows. Regardless, in this totally mad eup case the total potential loss (ignoring potential to be invested elsewhere) is $3,500 or 87.5% of initial outlay.
The remainder using the simple information and interpretation available to us looks like this:
Success$ = ($100* 100) * 5 = $50,000
That is 12.5x the initial outlay of $4,000. Making 12.5x your outlay in presumably only a handful of years is massive. Obviously growth doesn't mean exactly 100X, but I just wanted to show that a handful of hits (and I think 5/40 is bigger than a handful) can easily pay for everything else and more.
Reread the article, this is only 2-3.5 years of duration. That seems insanely quick for these kinds of results.