
Jake Brukhman, founder and CEO of CoinFund, predicts we'll train a 400 to 500 billion parameter AI model on a decentralized network within the next six to nine months, with no data center required.
In this episode of The Index Podcast, Jake breaks down why decentralized AI training is becoming the real counterweight to centralized AI labs, and why token incentives might be the missing coordination layer for the next era of intelligence.
This is what happens when a crypto investor who called decentralized training two years ago before anyone believed it comes back to explain why it's actually happening now.
๐ In this episode, we cover:
๐ Jake's bold prediction: a 400 to 500 billion parameter model trained on a decentralized network within 6 to 9 months
๐ฐ Why open source AI models don't have a business model, and why decentralized AI training does
๐ Why cutting access to Fable 5 was a wake-up call for enterprises relying on centralized AI labs
๐ฅ๏ธ How training on a swarm of consumer hardware could actually be cheaper than a data center
๐ธ Why a trillion dollar inference market has almost no room for open models
๐ How token incentives could crowdsource $100 million worth of compute in a single weekend
๐ข Inside CoinFund's bets on Pluralis, Bagel, Gensyn, and Prime Intellect
๐ Follow & learn more:
๐๏ธ The Index Podcast: https://www.youtube.com/@theindexpodcast
๐งโ๐ป Jake Brukhman (X): https://x.com/jbrukh
๐ CoinFund (Website): https://www.coinfund.io
๐ฆ CoinFund (X): https://x.com/coinfund
More episodes from "The Index Podcast"



Don't miss an episode of โThe Index Podcastโ and subscribe to it in the GetPodcast app.








