BlockBeats news, on June 14, Jake Brukhman, founder of CoinFund, stated that AI models inherently possess centralized characteristics and are a primary focus for government regulation and control, with Anthropic’s latest export control compliance actions further confirming this trend.
He pointed out that decentralized networks can serve as a crucial counterbalance to the current landscape, and the core challenge in building sovereign, open, and public decentralized AI lies in computational power. Although it is widely believed that only trillion-dollar tech companies can afford training cutting-edge models, there is actually sufficient general-purpose GPU compute power available globally—the key is developing new distributed training algorithms.
Brukhman said that teams such as Gensyn, Prime Intellect, Bagel, Pluralis, Nous Research, Macrocosmos AI, and Covenant AI have been exploring this direction; although initially widely considered infeasible, it has proven that distributed training is not only achievable but also lower in cost and nearly as efficient as traditional approaches.
In addition, he believes another major challenge for decentralized AI is economic sustainability. While open-source models are important, they lack mature business models; Pluralis explores a commercial path for tokenized AI models by distributing model weights among participants.
Brukhman said that we are at a critical juncture: whether AI moves toward full centralization, censorship, and unilateral government control, or is built on open, decentralized networks as public AI, will determine the future direction of the industry.