Ask HN: What's the Path for Innovation in Silicon?

If someone comes up with an idea, and pushes it through TinyTapeout or some other ASIC shuttle, and it actually meets design goals, such as lowering the cost of LLM computation by an order of magnitude. How does that lead to actually realized profit? Are there any VCs who specialize in this?

It seems to me that if Meta and others are going to spend a Trillion $USD on datacenter, saving even 50% out to be a good value add.