
CoinFund founder Jake Brukhman argues idle GPUs can power decentralized AI training, backing the thesis with $158M in investments across Prime Intellect and Gensyn.
Jake Brukhman sees a concentration problem in artificial intelligence. A handful of companies own the biggest models, the largest GPU clusters, and the data pipelines that feed them. His answer is decentralized networks built on distributed GPU resources.
Brukhman's firm, CoinFund, has raised $158 million to back crypto and AI startups. Its portfolio includes Prime Intellect, Pluralis Research, and Gensyn – each building pieces of a decentralized compute stack.
At the Theta Capital Legends4Legends conference, Brukhman argued that consumer and data-center GPUs sitting idle around the world can be aggregated into networks powerful enough to train advanced AI models collaboratively. Technological advances have moved decentralized training from a speculative concept to an emerging reality, he said. A race in decentralized AI training is coming, and it will happen sooner than most expect, he predicted.
Prime Intellect, focused on an open-source decentralized AI stack, raised $5.5 million in seed funding co-led by CoinFund in April 2024. The company later added $15 million.
Pluralis Research secured $7.6 million in a seed round co-led by CoinFund and Union Square Ventures in 2025.
Gensyn runs a decentralized training network with an ERC-20 token called $AI. The token covers verification, staking, payments, and governance. Total supply is 10 billion.
For crypto investors, these projects introduce a new category of token utility. Rather than tokens driven mostly by speculation or DeFi mechanisms, AI-related tokens are tied to real compute workloads. A token used to pay for GPU time on a decentralized network sees demand from actual resource consumption, not just trading volume.
Brukhman founded CoinFund in 2015, originally as a Slack group for early decentralization advocates.
The metric to watch for investors tracking this space is not token prices. It is actual compute utilization on these decentralized networks. If projects like Prime Intellect and Gensyn can show distributed GPU clusters training models at scale, the investment thesis shifts from speculative to structural.
Prepared with AlphaScala editorial tooling from the source reporting linked above. Indexable analysis may include a cited Alpha Score value. Publishing checks screen each story before release. Educational coverage, not personalized advice.