According to CoinMeta, author OKX indicates that the infrastructure of AI is entering a capital-intensive expansion phase, and GPU as well as data centers are becoming increasingly dependent on external financing. While Take-or-pay can lock in some future cash flows to some extent, it cannot eliminate risks such as GPU rent, residual value, and refinancing. Therefore, there is a growing demand for computing power indices, forwards, futures, and OTC hedging tools. The risks that are truly likely to enter the derivatives market are those that are still exposed to floating prices in the public market and cannot be absorbed by the existing balance sheets. Since GPU-hour cannot be stored, and due to the high degree of non-standardization in terms of model, region, network, cluster, and SLA, the early market is more likely to rely on OTC and dealer perpetual contracts for near-term price discovery, while fixed-term futures are more suitable for corporate hedging, term pricing, and physical fulfillment. In the short term, those that are most likely to capture commercial value first will be broker / dealer that control the real computing power order flow, as well as midstream platforms that manage inference order flow and scheduling. As the market becomes more standardized, the importance of price indices and exchanges will gradually increase.