My theory is that 500 2 person companies have a worse chance at making a real splash than if the 100 best ideas from that group had 10 good people working on them. That's part of what he's saying - it's harder now for good startups to grow when everyone wants to and is incentivized to make their own web startup.
The other part he's saying is that people can have a larger marginal impact on the world working at a place like Facebook than they could in their own startup. For the median case, this is true.