What I don't see mentioned too often is promoting chip design. Compared to building a fab, making a startup that simply designs chips should be straightforward but through my initial research into it, that's not the case. Tech cos like Google and Amazon have already begun endeavors into in-house fabless chip design through ex.: TPU and Graviton, but hardware startups seem to be few and far between (there are a few like SciFive and Cerebras, but it's not like they are exactly ubiquitous).
In part it has to due with the huge complexity of modern day chips and the fact that if any new startup wants to be competitive it requires hiring talent with deep and niche industry experience.
I think what's more of an issue is accessibility for newcomers into the space. From what I've read, the top 3 EDA companies control the vast majority of their market and license their tools for hundreds of thousands of dollars. These tools aren't accessible to the majority of students and most likely fledgling startups. I know these tools are extremely complex, but it always stuck me as odd how IDEs like VS Code and the ones made by JetBrains are free or are priced so that the vast majority of developers (whether as individuals or as companies)can afford it, while the top of the line EDA tools are only accessible by established players. I know there is an open source effort to change this, but there's a long way to catch up.
Maybe things will change. AMD started out reverse engineering Intel chips and now are beating Intel in most metrics. Hopefully hardware becomes more hyped over time just as machine learning and software startups have.