Exactly. I've written about this at length because people who don't have much experiences in the hard science/engineering fields have a vast disconnect between what's possible in software vs. what's possible in the physical world. Startups are great for pure software and very simple and constrained science/engineering experiments. These, by definition, won't be groundbreaking.

Big science/engineering is very expensive, very long term and very risky. No one wants to foot the bill. Not corporations, not investors, and recently not even governments (NIH/NSF funding cuts, etc.).

I discuss this further here: https://news.ycombinator.com/item?id=8337837