Daniil and David Liberman are serial entrepreneurs and investors discussed in a Forbes Russia interview about AI, internet monopolies, and decentralized compute. In the interview they frame AI not as a normal software product but as infrastructure comparable to electricity or the internet: whoever controls access to cheap compute can shape who participates in the AI economy.source: forbes-liberman-brothers-ai-infrastructure-2026.md
Their core economic lens is the "copyable product": expensive to create once, cheap to reproduce many times. They use this to explain why search engines, social networks, marketplaces, operating systems, and AI models tend toward monopoly-like structures. Once everyone benefits from using the same network or copied artifact, control over that common layer becomes the main political-economic problem.source: forbes-liberman-brothers-ai-infrastructure-2026.md
The brothers argue that open models alone are insufficient because the real bottleneck shifts to AI compute infrastructure: GPUs, data centers, energy, chip supply chains, and inference capacity. Their proposed answer is decentralized AI compute: an open protocol where participants provide hardware, earn protocol tokens, and make compute publicly accessible without a company or nonprofit board retaining central control.source: forbes-liberman-brothers-ai-infrastructure-2026.md
Their argument connects agent-loops to infrastructure economics. Agents can burn far more tokens than chat interfaces, so they create demand for cheap, routable inference and may automatically choose cheaper model/compute providers when asked to optimize cost.source: forbes-liberman-brothers-ai-infrastructure-2026.md
Related pages: ai-compute-infrastructure, decentralized-ai-compute, agent-loops, ai-labor-market, test-time-compute-evaluations.