Week ending 2026-09-04

The view

The market leaned toward stagflation fear without committing. Driven by crude, core CPI, front-end breakevens more than the back end. That shape is a near-term repricing, not a change in the long-run anchor. Nominal yields did not keep up with breakevens, so real yields slipped. Cyclicals underperformed defensives (-0.6sd). CCC spreads widened while broad high yield sat still: the weakest credit is cracking first, which is a warning rather than a break. The dollar firmed.

The baskets that rose were hard-asset and commodity-linked and AI capex; the ones that fell were domestic cyclical and defence. Datacentre baskets rising while software fell says AI capex is intact and the software sell-off is about rates and the application layer, not tech. Rails down with dry bulk up means seaborne commodity demand firm, domestic freight soft.

The most unusual move is Agricultural inputs at +2.6sd with 100% of members up. The discovery layer found that group on its own from raw co-movement, which makes it a real driver rather than noise. One unnamed cluster of 6 names moved together (+1.3sd, cohesion 0.41). Look at the ticker list before believing it; low cohesion clusters are usually noise. What decides next week: whether the 10-year clears the deadband, which would decide the quadrant; whether cyclicals against defensives clears the deadband; whether CCC widening spreads into broad high yield.