The argument against this logic is that most companies who employ developers are not in the business of making an arbitrarily large number of software products. Many companies have just one product (or service), which is often tied to a non-software product or service, e.g. banks, schools, hairdressers, etc. So, their development needs are very finite.

Yes, society as a whole has a much larger appetite for software, but society-as-a-whole doesn't hire developers: specific firms hire them, and they hire them to work on specific projects that theoretically have "finish lines" which could potentially be reached in the near term if the development they can afford were more productive -- which is what AI threatens to do.

And these firms only form around activities that they think can support a profit stream. But many of society's software needs might not lend themselves to harvesting a profit, for the obvious reason that many areas of human life are not mediated by financial transactions. And yes, AI devbots will probably make it easier to fill unprofitable gaps, but the reason people are concerned about AI coming for dev jobs is that people need jobs to pay for shelter and food. So, society's bottomless appetite for software is not going to salvage this situation.