Why the clock is ticking on the Fed-fuelled AI bubble (www.scmp.com)

The piece argues the current AI boom is being artificially prolonged by financial plumbing rather than pure technological progress: profitable firms recycle cash into loss-making AI ventures, and are likely to sell stock to fund them. That buffering may buy quarters, not years, and could escalate if the US government starts injecting borrowed money — a move that risks destabilizing the Treasury market. The author likens this dynamic to the dotcom era, where capital flows (and sometimes fraud) temporarily sustained irrational valuations; here, strategic reinvestment and balance-sheet engineering are the lifelines keeping marginal AI plays afloat. Concrete examples underline the systemic mechanics and technical implications: Nvidia’s reported $100 billion commitment to OpenAI would convert cash into equity while potentially driving more than $100 billion of incremental chip demand via leasing arrangements, a win-win for chip suppliers and investors. Despite GPT-5’s reportedly modest gains over prior models, plentiful capital still incentivizes a push toward AGI, inflating valuations (OpenAI could be valued at hundreds of billions) and rewarding hardware vendors with large capital gains. For the AI/ML community this means abundant funding and hardware but also distorted incentives — resources may flow to grand AGI narratives over incremental, verifiable research, and broader market fragility could swiftly curtail that support.