A thousand-fold surge in China's average daily token call volume has yielded only 3.07 billion yuan (approximately $455 million) in full-year 2025 public cloud MaaS revenue, highlighting a mismatch between large-model computing power and monetization. According to Northeast Securities analyst Song Xinzhu, AI profit capture consists of four mechanisms: scarcity premiums, generational premiums, integrated internal settlement benefits, and switching-cost premiums. Profits are entering financial statements sequentially from upstream to downstream, with the upstream computing infrastructure layer first realizing scarcity dividends, the midstream model layer facing deflation from commoditization, and the downstream application layer absorbing falling computing prices to build moats via switching costs. Due to differing payment endowments, incremental AI value in the US is captured within high-priced software subscription systems, while China's low-priced token dividends spill into the application layer. China's average daily call volume rose from roughly 100 billion in early 2024 to 100 trillion by the end of 2025, but most consumption occurs internally within major tech firms' ecosystems. As of the second quarter of 2026, the trailing twelve-month capex-to-operating cash flow ratio for the four major US cloud companies had risen to between 0.63 and 1.05. Alphabet reported negative free cash flow for a single quarter for the first time, while Meta's free cash flow plummeted 91% year-over-year.