GOOD MORNING, Treasury tried to calm the bond market Wednesday. By Thursday, yields were climbing again. The 30-year Treasury yield returned to roughly 5.25%, oil rose on renewed Iran threats and Wall Street posted its worst session in three weeks. The bigger story is why the pressure keeps coming back: Washington is borrowing heavily at the same time AI companies are competing for enormous pools of long-term capital. Alibaba offered the corporate version of that trade Thursday — AI cloud revenue is surging, but so is the bill required to build it. MARKETS | TLDR
BOND MARKETTreasury Bought More Bonds. Yields Came Back Anyway.The Treasury Department doubled the size of buybacks for securities with maturities between 10 and 30 years to $4 billion per operation, an attempt to improve liquidity and ease pressure in the long end of the market. The announcement initially worked: long yields fell Wednesday. By Thursday, much of that relief had reversed, with the 30-year Treasury yield back near 5.25% and the 10-year near 4.70%. The failure of the rally to hold matters more than the size of the buyback itself. U.S. federal debt has crossed $40 trillion, deficits remain large and AI infrastructure builders are simultaneously tapping capital markets for hundreds of billions of dollars. In other words, the Treasury and the AI economy are increasingly competing for the same long-duration capital. Higher yields then feed back into mortgages, corporate debt and equity valuations. Treasury Secretary Scott Bessent said the buybacks could be expanded further and argued that current yields do not reflect underlying fundamentals. But the program does not remove the structural causes investors are pricing: fiscal supply, inflation uncertainty and an unusually capital-intensive technology investment cycle. The government can improve market plumbing. It cannot make the demand for capital disappear. AI ECONOMICSAlibaba's AI Revenue Is Growing. Its Profit Bill Is Growing Faster.Alibaba reported a 9% increase in quarterly revenue as demand for cloud computing and AI services accelerated, but net profit fell 75%. Revenue from cloud and AI compute services rose 45% to 48.44 billion yuan, while the company said its model-as-a-service business has already reached more than 16 billion yuan in annual recurring revenue. The growth is real. So is the cost of creating it. Alibaba has already spent roughly half of the 380 billion yuan ($56.4 billion) it planned to invest in AI infrastructure between 2026 and 2029. The company is buying chips, building compute capacity and developing proprietary silicon because management believes it must own enough infrastructure to capture future enterprise AI demand. CEO Eddie Wu said Alibaba expects those AI investments to break even within three years at current average gross margins. That three-year target is the number worth watching. AI investors have spent much of the cycle rewarding capex because cloud and compute demand keep accelerating. Alibaba turns the abstract ROI debate into a measurable promise: spend heavily now, then prove the infrastructure can earn back its cost. If more hyperscalers start giving investors explicit payback timelines, the AI trade may increasingly be judged on capital efficiency rather than capex growth alone. HEADLINES
UPCOMING
DEEP INSIGTHSTreasury's Upsized Buybacks and the FedRead this for the policy conflict underneath the bond move. Treasury is trying to contain long-term borrowing costs while the Fed is still fighting inflation, raising a broader question about what happens when fiscal and monetary objectives begin pulling different parts of the yield curve in opposite directions. Alibaba's AI Capex BetAlibaba provides a useful non-U.S. case study of AI economics. Cloud demand is accelerating, model revenue is becoming material and management is willing to spend tens of billions of dollars ahead of that demand — but it has also put a three-year break-even target around the investment, giving investors a concrete benchmark for whether the capex ultimately earns its cost. |