Milk Man speaks with AI researcher Vincent about why compute remains scarce even as newer chips enter the market. Demand from AI agents is helping older GPUs retain value, while neoclouds and SpaceX can command premium pricing by bringing powered data centers online faster than traditional hyperscalers.
Vincent argues that the market could begin shifting from scarcity and speed toward cost efficiency around 2028. That transition may favor operators with lower financing costs, specialized infrastructure, custom silicon, or vertical integration, while challenging neoclouds to preserve margins as pricing becomes more transparent and standardized.
The discussion compares the strengths and risks of neoclouds, hyperscalers, and SpaceX, including access to capital, dependence on Nvidia, operational complexity, and long-term compute costs. Vincent also explains why he avoids chasing volatile compute stocks without deep conviction and prefers a concentrated portfolio focused on AI infrastructure picks and shovels.