TechFlow News, June 14: Jake Brukhman, founder of CoinFund, recently stated that AI models are inherently highly centralized, making them more susceptible to government regulation and policy control. He believes Anthropic’s latest implementation of export-control compliance measures further confirms this trend.
Brukhman noted that decentralized networks could serve as a critical counterbalance to the current AI landscape, and building an open, public, sovereign decentralized AI ecosystem faces its greatest technical challenge in organizing and utilizing compute resources.
He pointed out that the market generally assumes only large tech companies with trillion-dollar market capitalizations possess the capability to train state-of-the-art AI models. In reality, however, vast amounts of general-purpose GPU compute resources already exist globally; what is truly needed is more efficient distributed training algorithms.
Brukhman mentioned that multiple teams—including Gensyn, Prime Intellect, Bagel, Pluralis, Nous Research, Macrocosmos AI, and Covenant AI—are exploring distributed AI training solutions. Although this direction faced widespread skepticism early on, practical implementation has demonstrated that these technologies are not only feasible but, in certain scenarios, can achieve efficiency comparable to traditional centralized training—at lower cost.
Beyond technical hurdles, he emphasized that decentralized AI must also address economic sustainability. Brukhman observed that while open-source models have accelerated industry innovation, they lack mature business models over the long term. Some projects are now experimenting with model weight sharing and tokenization mechanisms to explore new value-distribution frameworks.
Brukhman believes the AI industry stands at a pivotal development juncture: whether it will continue trending toward heightened centralization and stricter regulation—or evolve into a public AI ecosystem built upon open networks—will profoundly shape the industry’s future trajectory.
