Test case for your claim: someone privately intensively develops some open-source product that uses Gurobi as optimization backend. I guess on-demand will become very expensive.
> It would be best to develop against open models and plug Gurobi in separately in this situation I expect.
That is basically what I did in the end.
> If you are working on a problem that NEEDS Gurobi's level of quality, someone will pay you to build it
Here I vehemently disagree: there exist quite a lot of problems that would insanely profit from a great MILP solver:
- Only a small subset of these problems have a (huge) commercial potential (i.e. are suitable for making some customer money).
- Even if there exists an insane commercial potential, building great applications/models, and being able to convince someone with huge pockets (or even get access to them) are completely unrelated skills.
