# Nuclear can't arrive in time — distributed training is the answer

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2025-01-02
- **Theme:** Decentralized AI — https://www.gregosuri.com/predictions/decentralized-ai/
- **Source:** "AI Sovereignty: Building the Path to Individual AI with Greg Osuri" (The Outpost Podcast)
- **Watch at:** 00:31:27 — https://podcasts.apple.com/us/podcast/ai-sovereignty-building-the-path-to-individual-ai/id1779395525?i=1000682416476
- **Canonical:** https://www.gregosuri.com/predictions/decentralized-ai/2025-01-02-nuclear-can-t-arrive-in-time-distributed-training-76yh/
- **Cite as:** Greg Osuri, "Nuclear can't arrive in time — distributed training is the answer," "AI Sovereignty: Building the Path to Individual AI with Greg Osuri" (The Outpost Podcast), 2 Jan 2025, 00:31:27. https://podcasts.apple.com/us/podcast/ai-sovereignty-building-the-path-to-individual-ai/id1779395525?i=1000682416476

## Quote — verbatim

> So we have to, you know, build nuclear reactors in the next two years, which is not going to be reality. So we're going to have to explore very seriously about distributed training. Right? So, distributed training also means distributed grid. That means you can actually go, instead of concentrating too much compute in a single data center, you distribute that compute all over.

## Context — site annotation, not spoken

He cites Prime Intellect's INTELLECT-1 (10B parameters trained over the internet via DiLoCo) as the two-month-old proof point ([00:32:08]).