Key Notes
- Electric Capital co-founder predicts a catalyst event for decentralized AI similar to Bitcoin's 2008 financial crisis moment.
- Dragonfly's managing partner argues that data volume is worthless for advanced AI models, emphasizing quality over quantity instead.
- Decentralized AI systems will always be inferior to centralized ones but must offer unique value like censorship resistance.
Franklin Bi, Haseeb Qureshi, and Avichal Garg—three well-known crypto venture capitalists—attended the 3rd Annual Summit on Decentralization and AI 2025 hosted by the Berkeley Center of Responsible Decentralized Intelligence as panelists, bringing valuable insights to the industry’s present and future.
In particular, Avichal, co-founder of Electric Capital, drew a parallel to the AI industry and the 2008 financial crisis, as a major catalyst to the creation and popularity growth of Bitcoin BTC $116 927 24h volatility: 0.2% Market cap: $2.33 T Vol. 24h: $34.85 B , saying decentralized AI is likely to experience a similar catalyst event.
The VC brought up the question of “What happens in a world where all governments in the world want to control this stuff [AI]?” urging the audience to think about what is “actually meaningful in that world” to make sure “we don’t end up in some dystopian place.” In that context, he made the “2008 moment” comparison.
“I don’t think we’ve had the 2008 moment yet. 2008 was sort of this crystallizing moment for the world’s financial system, and Satoshi wrote the [Bitcoin] whitepaper, but I don’t think we are far off from it. We are like in 2006. You can kind of see some things going on; you are not sure how it cracks, but it’s going to crack. (…) I think we are roughly in 2006, right before Bitcoin. Something is going to crack in the next couple of years, and then we will understand what decentralization is useful for,” Avichal Garg, Electric Capital co-founder, explained.
Avichal explained the importance of “having the technical depth and understanding of what’s coming.” According to him, people who manage to acquire this level of understanding have a “really tremendous opportunity over the next 10 years.” This understanding, in his opinion, would come from thinking about “this stuff really deeply” and thinking about “this stuff from a meaning perspective.”
Other Insights on Decentralized AI by the Panel
Before all that, Haseeb, managing partner at Dragonfly, dropped a bomb on many AI developers and enthusiasts: “Your data is worthless.” According to him, advanced LLMs are not interested in data volume but quality—a myth disproven by ChatGPT 4.5, trained with a large chunk of data but barely overcoming the previous version. The best models will likely pay an expensive price for carefully selected information by top minds on the planet, not by airdropping crypto tokens to mass tweets.
Moreover, Haseeb first introduced the idea that governments will try to control and censor AI , despite it not having started yet. This, as per what he said, will make decentralized AI “matter in the same way that the rise of financial surveillance made Bitcoin matter.”
Hot take from my panel on decentralized AI at UC Berkeley: Your data is worthless.
The idea that you'll get rich selling your data to train AI is a fantasy. The frontier is high-quality synthetic data, not your dumb tweets.
Second hot take: the government censorship of AI has… pic.twitter.com/t1uwnIgt7e
— Haseeb ＞|＜ (@hosseeb) August 8, 2025
Another interesting take from Dragonfly’s managing partner is that decentralized systems will inevitably be inferior to centralized systems in areas like performance or efficiency.
“Decentralized AI will never be better than centralized AI because a centralized AI provider could just grab a decentralized model, run it by an API, and say ‘Great, this is GPT 6.’ So, you are always paying a decentralization tax,” Haseeb Qureshi, Dragonfly managing partner, argued.
On that point, he raises the question of what people are getting “in exchange for paying this tax.” Again, in a parallel to Bitcoin, he believes that the cryptocurrency provides “censorship resistance” in exchange for paying that tax. Now, the decentralized AI community needs to find the answer for the same question, which he doesn’t believe has been discovered yet.
next
