There are generally useful software projects that require initial investment of time and expertise that exceed what can be realistically achieved as a hobby for a first version. Most of the gaps in open source are software systems of this type.

If you need a type of software system that requires 100k LoC of high-expertise code to get to a Version 1.0, you will need to spend millions of dollars. You can neither recruit non-experts to do a credible job, severely limiting the pool of people that could contribute in practice, and most people with a day job will be reluctant to spend the additional time required to write tens of kLoC of difficult, production-quality code -- man-years essentially.

And this ignores that most "high-expertise" software designers are prohibited from working on open source versions of what they do separate from their paying job. This is the reason, for example, that the myriad exotic high-performance database engines don't have open source equivalents -- only a handful of people know how to design those systems and their closed source work precludes participation in open source. But even if they could, the level of individual effort required will exceed what they are likely to want to do for free.