10 YR Treasury Yield is Going Vertical, Fast Tracking 2008 Relapse

Long-duration US Treasury yields are at multi-decade highs because war-linked energy costs and inflation are colliding with a large debt roll-over, so the practical risk is a confidence/solvency crisis if private and government debt loses buyer demand [08:01]. The video frames this as a possible 2008-style private-credit/Eurodollar freeze tied to AI data-center debt, not a confirmed market break [27:54].

Key points

• Long-end yields rose to multi-decade levels: the 10-year closed at 5.22%, the 30-year hit 5.50%, and Wednesday's move was the largest single-day 10-year jump since April 2025 [08:10]. The speaker ties this to war-linked oil above $103, record diesel, August CPI at 3.4%, and rising supplier prices at the fastest pace in four years [09:45]. The Fed is still expected to hike, with a 75% implied October hike, even though the inflation shock is supply-driven [12:11]. Debt compounds the risk: about $10 trillion rolls over in 12 months, and interest payments are expected to roughly double [14:01].

• Higher rates are already freezing household leverage: mortgage rates are over 7%, while half of existing mortgages are below 3%, so many owners will not sell into a 7.25%-7.5% market [25:49]. That cuts housing velocity, one of the main components of normal economic activity, even though last-quarter GDP growth was only 1.6% [25:00]. The transcript argues inflation expectations matter more than the inflation print itself, because fear of further price rises makes consumers and firms pull back [24:42]. That pullback is the pathway to a credit or spending slowdown, not merely a mild rate cycle [26:20].

• Oracle’s 5%-6% one-day drop followed a reported at-least-six-month delay at a New Mexico data center, which raised debt-service timing doubts and hit lenders such as Blue Owl [20:08]. The transcript treats this as stress in a larger AI capital program: data-center debt is packaged as investment-grade, but 401(k), insurance, and private buyers may carry exposure [21:50]. The constraint is energy rather than chips, with Musk citing a 2027 power gap versus compute demand [22:45]. The 2008 analogy is that opaque private credit can trigger Eurodollar confidence to seize [30:19].

• The counterpoint is that AI may generate the productivity gains used to justify the debt: Anthropic’s Claude surveyed DNA databases, with roughly 950 agents working for 21 hours to identify a previously unknown CRISPR-like molecular machine [40:10]. Its function is unproven; speakers stress the discovery is early and may not become a treatment [44:10]. The method matters more: falling token costs, including a claimed 1,000x reduction, could make large-scale biomedical search much cheaper [46:37]. This possible payoff is the reason the debt-funded AI buildout is taken seriously, not evidence that the energy or debt crisis is solved [18:24].