Alessio Fanelli

Partner · Decibel

Recommendations and personal stack

Insufficient evidence — no named books, apps, or tools yet.

Published claims

AI progress can be increased by scaling up more, leading to more than 50% profit.

if AI progress is fast and you can increase the progress by scaling up more, you should just have more than 50% and not make profit.

13 Feb 2026 · 1:03:13

Log returns to scale can be significant, and investing in serving inference or hiring engineers might be more profitable than scaling up.

Remember the log returns to scale. If 70% would get you a very little bit of a smaller model through a factor of 1.4x... That extra $20 billion, each dollar there is worth much less to you because of the log-linear setup.

13 Feb 2026 · 1:03:26

Spending on training and inference is not 100% efficient, and there's an equilibrium where companies spend less than 100% on training and inference.

Why doesn't everyone spend 100% of their compute on training and not serve any customers? It's because if they didn't get any revenue, they couldn't raise money, they couldn't do compute deals, they couldn't buy more compute the next year.

13 Feb 2026 · 1:05:10

The underlying economics of the industry are profitable, but there's a challenge with demand prediction for the next year's compute.

So the underlying economics are profitable. The problem is you have this hellish demand prediction problem when you're buying the next year of compute and you might guess under and be very profitable but have no compute for research. Or you might guess over and you are not profitable and you have all the compute for research in the world.

13 Feb 2026 · 1:06:08