>Microchips were an incremental improvement where the individual increments yielded utility far greater than the investment.

You should look up how modern EUV lithography was commercialised. This was essentially a big plasma physics puzzle. If ASML hadn't taken on a ridiculous gamble (financially on the same order of magnitude as a new collider, esp. for a single colpany) with the research, Moore's law would have died long ago and the entire tech industry would be affected. And there was zero proof that this was going to work beforehand.