In my opinion, 1) Barrier to entry is way too high compared to other software niches. Good luck designing a cycle-accurate CPU simulator, or creating a reliabl

In my opinion, 1) Barrier to entry is way too high compared to other software niches. Good luck designing a cycle-accurate CPU simulator, or creating a reliable and robust analog circuit simulator that competes with offerings from EDA giants like Cadence and Synopsys. 2) The adoption process. I learn tool X in school, I start a company that uses tool X, I teach my hires tool X, etc. Once tool X becomes a central part of the design process, it's almost impossible to get rid of it.