# By 2030 only nuclear-scale power can train a state-of-the-art model

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2025-06-24
- **Theme:** Energy &amp; AI — https://www.gregosuri.com/predictions/energy-ai/
- **Source:** Akash Accelerate 2025: Official Livestream (Akash Network)
- **Watch at:** 01:39:34 — https://www.youtube.com/watch?v=wiVrlo7XJA0&t=5974s
- **Canonical:** https://www.gregosuri.com/predictions/energy-ai/2025-06-24-by-2030-only-nuclear-scale-power-can-train-a-state-yped/
- **Cite as:** Greg Osuri, "By 2030 only nuclear-scale power can train a state-of-the-art model," Akash Accelerate 2025: Official Livestream (Akash Network), 24 Jun 2025, 01:39:34. https://www.youtube.com/watch?v=wiVrlo7XJA0&t=5974s

## Quote — verbatim

> Considering the energy need, particularly concentrated energy need for training, is doubling every two years, the expectation by 2030 the only way to train the state-of-the-art model is with a gigawatt supply. So nuclear will be the only way to do so. Unfortunately, the last nuclear reactor we built in America took about 14 years.

## Context — site annotation, not spoken

Greg framing a question to SoftBank's Javier Villamizar during the Stargate fireside; a recurring quantified forecast of his.
