Citadel Securities believes that the AI market is showing signs of increasing demand elasticity. Since the end of June, the usage-weighted token price has dropped by approximately 40%, but the H100 leasing price has rebounded from its June low. Among the top 1% of companies with the highest AI spending in July, per capita spending increased by about 49% month-over-month. In the second quarter, revenue for hyperscale cloud providers reached $106.3 billion, a year-over-year increase of approximately 43%, indicating that new computing power can still quickly find paying demand. The standards for measuring AI economics are also changing. Compared to the price of a single token, the cost of completing an effective task is more valuable as a reference. Open-source models with sufficient capabilities allow customers to reduce costs while maintaining similar output quality. However, this expands AI usage, supporting demand for GPUs and cloud services, but also weakens the pricing power of proprietary cutting-edge models. Citadel also cautions that the compute power pricing market is still relatively young and opaque, and the current high utilization rate does not guarantee a supply-demand balance in the coming years. If model architectures significantly improve efficiency, compute power demand may fall short of the expectations implied by existing capital expenditures.