Inflection Meets Liquidity

Liquidity is entering an inflection point as AI becomes revenue-generating, Treasury and banking policies ease financial conditions through the financial plumbing, and bond-market anomalies suggest a changing regime for long-term rates.

• Jensen Huang’s statement that “compute is revenue” signals a shift from speculative AI capex toward productive assets, supported by revenue growth at OpenAI and Anthropic, hyperscaler order books, and a roughly 50% month-over-month rise in July AI spending per employee among the top 1% of enterprises in the Ramp AI Index.
• The author argues that cloud revenue growth could exceed capex growth next year, while consensus expects AI cash flow to recover toward previous peaks by 2028 and double again by the end of the decade.
• The bond market rallied after Stanley Druckenmiller warned about unsustainable U.S. deficits and opposed attempts to control bond prices, an apparent anomaly that may indicate a liquidity or policy inflection.
• From 1942 to 1951, when debt-to-GDP exceeded 100%, Federal Reserve yield-curve control capped long-term yields and helped debt-to-GDP fall as nominal growth supported a strong economic expansion; the author sees current Treasury buybacks, bill issuance, foreign-exchange intervention and liquidity measures as potentially softer forms of yield management.
• Treasury bill issuance, possible use of the Treasury General Account, banking deregulation and higher stablecoin demand could increase market liquidity, while M2 money supply has recently accelerated to new highs; the Federal Reserve’s statutory objective of moderate long-term interest rates may also become more important.

Takeaway: Investors should treat AI infrastructure as an increasingly self-funded productive asset and monitor Treasury operations, bank balance-sheet capacity, stablecoins and long-term rates for evidence that liquidity is easing, while recognizing that the current policy interpretation remains uncertain and that the historical precedent is not directly comparable.