Pantera Capital junior partner Jay Yu said in an essay titled The Rise of Compute Markets that compute and data center spending has become a trillion-dollar category, while GPU procurement still largely happens through group chats, over-the-counter brokers, and bilateral agreements. According to ChainCatcher, he said compute financialization is still early and constrained by SKU, time, and location, but it could become a commodity asset class similar to electricity or oil over the next 5 to 10 years.
Yu said the compute market may develop a structure similar to the electricity grid-operator-node model, with hardware-supplier-cluster layers. He described new cloud providers as structural short sellers of GPUs, while on-demand platforms and the application layer are long compute.
He said that for every $100 spent on inference at the application layer, about $45 flows to the on-demand layer, about $50 to the new cloud or GPU layer, and about $5 to routing layers such as OpenRouter. Yu also said NVIDIA can be viewed as the central bank of compute and has announced residual value support of up to 25%.
He added that physical delivery has a more durable moat, with index products, cash-settled exchanges, and financing tools expected to follow. He said participants include SF Compute, Vast AI, Runpod, and Compute Exchange, and noted that the compute market does not require mandatory price transparency, which could increase basis risk.
Weatherly