I think he is speaking of the opportunity cost. it is the conservative way to generate funds, but not the most efficient in the end.
It probably also depends on what you are trying to build. His initial example of building chips would probably be nearly impossible without a pretty big up front investment.
If you are building something with software, it is a lot more likely that you could self fund.
This was in 2001 where just a single server colo would cost ~$300/mo. Other costs have come down as well, while technologies have developed to make a full stack web app development possible for a single person to do. The average dev is easily 10x more capable today than 12 years ago.