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

1 mikewarot 4 8/4/2025, 4:51:44 PM
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.

Comments (4)

Someone · 1h ago
I would think patenting the idea is the way to go, rather than verifying your design on an ASIC. Chances are that’s much more expensive, though, especially for a worldwide patent.
mikewarot · 47m ago
Patenting it would have been a good idea a decade ago before I first started talking about it in public
bigyabai · 2h ago
> How does that lead to actually realized profit?

It depends. How many billion dollars are you willing to pay TSMC for fab access? If it's not at least $5 bln then you might as well pack it up and go home, Nvidia and Apple can take things from here.

mikewarot · 2h ago
Small scale fab access is in my budget, enough to prove the point. (Less than $500)