Heavy debt supply pushes long Treasury yields to their highest levels this century

August's inflation miss is the same size as the revision that came with it. Economists polled by Dow Jones expected the personal consumption expenditures price index to rise 3.7% from a year earlier and the core index 3.3% ^[5]. The Commerce Department reported 3.4% and 3.0% ^[4]. The same overhaul of the index, applied to five years of data, lowered July's rates from 3.7% to 3.4% and from 3.3% to 3.0% ^[6][7]. Both misses are 0.3 points, both revisions are 0.3 points ^[1][2], and August's annual rates equal July's revised ones exactly ^[3].

The monthly figures are cleaner. Headline PCE rose 0.3% from July, as forecast, while core rose 0.2% against an expected 0.3% ^[8]. I think most of the annual beat is a change in what the index counts, and the tenth of a point on monthly core is the part that shows prices actually cooling. One change stops booking higher portfolio management fees as inflation when stock prices rise ^[9]. Software now includes video games and web hosting ^[10], and legal services moved to a producer price series after the consumer one was discontinued ^[11].

The bond market did not treat the print as relief ^[2]. Seoul Economic Daily reports that Treasury yields climbed to their highest levels of this century, led by long-dated maturities ^[2]. The drivers it lists are federal borrowing to cover a large deficit, hyperscalers still selling bonds to fund AI infrastructure ^[3], crude held near $100 by Middle East instability and an economy stronger than expected ^[12]. The growth part shows up in the data. Second-quarter GDP was revised to a 2.2% annualized rate from a preliminary 1.5% ^[13], and final sales to private domestic purchasers rose 4.6%, 2.8 points faster than the first quarter's 1.8% ^[14][4].

The Fed's expected hold at its Oct. 27-28 meeting ^[1] concerns the policy rate. The pressure the report describes sits at the long end, in supply the policy rate does not set. The hyperscalers have not stopped issuing to fund AI infrastructure ^[3]. They are borrowing against a Treasury curve at its highest this century ^[2] instead of waiting for cheaper money.

The 6% figure comes from the same report. It says some in the market leave open a 30-year yield at that level and argue the rise need not be bad for the global economy, up to a point ^[15]. Their case is that yields are pricing long-term productivity gains from AI ^[16], and New York stocks have not come under heavy pressure ^[17]. The report does not give the current 30-year yield, so the distance to 6% cannot be measured from it.

The first path: supply keeps coming, growth holds, and 6% gets tested. In a second, AI returns reach earnings soon enough that higher yields become a cost of growth equities can carry, and the report names that timing as the deciding question ^[19]. The third runs through interest bills, with indebted Big Tech companies and households cutting investment and consumption ^[18] and long yields falling back with growth. I'd plan for the first, because both supply sources the report names are still running ^[3]. If core PCE keeps rising 0.2% a month and long yields fall with it, inflation was the driver after all and the supply explanation is wrong.