There is an anti common sense law called the Jevons paradox. In the 19th century, coal in Britain became increasingly cheap, and everyone thought that its usage would decrease. The result was a surge in usage, as affordability triggered more usage scenarios. History seems to be repeating itself in terms of AI computing power. In the past two years, the inference cost of GPT-4 API at the same level has decreased from $36/million tokens to less than $2, a decrease of 94%. You think companies will spend less money, but Amazon, Google Meta、 Microsoft's AI capital expenditures are all on the rise. Huang Renxun just announced that Vera Rubin will further reduce token costs by 90%, and the market's response is surprisingly to order more GPUs. The same logic is applied to the chain: the gas fees for Base and Solana are extremely low, and the cost of each AI agent call is close to zero. When the call is almost free, AI agents will initiate more transactions, the more, the more, until cheap becomes the reason for the increase in volume. Now Base+Solana accounts for 97% of AI agent to agent transaction volume across the entire network. This number is still rising because the lower the cost, the more active the agent. Low gas cost chains in the AI agent economy result in exponential growth in transaction volume, as cheap ones trigger more usage, while expensive chains are marginalized. Ethereum may need to consider this issue. After all, AI doesn't have as many E-guard beliefs, so AI is definitely looking for cheaper options.