The Fed Got The Inflation Print It Wanted And Bonds Sold Off Anyway

August core PCE rose 3.0% year over year against 3.3% expected, below the entire forecast range of 3.1% to 3.5%. The monthly figure was 0.247%. The surprise came from revisions concentrated in 2026 rather than from August itself — July was restated from 3.34% to 2.98%. On the revised basis, three-month annualized core inflation is 2.05% and six-month is 2.74%. Headline PCE came in at 3.4% against 3.7%. Q2 GDP was revised up to 2.2% from 1.5%. Inflation-adjusted consumer spending rose 0.6% in August, the strongest month since March 2025. Markets moved immediately. Odds of an October rate hike fell from roughly 70% to about 31%. The 10-year Treasury yield then rose to 5.304% — through its 2007 peak and the highest since May 2002. The 30-year closed near 5.64%. The spread between 30-year and 5-year yields widened to 55.5 basis points from 48.4 two sessions earlier. The curve steepened. Short rates fell on the Fed. Long rates rose anyway.

The week before this print, eight Federal Reserve officials made the same argument in public. Goolsbee said the expectation of future AI productivity gains creates a high danger of overheating now, and that the Fed needs to revisit the logic of looking through supply shocks. Musalem put the demand share of current inflation at about half and said the Fed cannot ease based on productivity that has not arrived. Cook said that if AI causes a rise in unemployment, the Fed would have limited tools because cutting would fuel inflation. Hammack said policy is not restraining activity outside housing. Barr said there is no path back to 2% in a timely way without further adjustment, and that AI's biggest short-term effect right now is driving up costs. Then the data contradicted all of it, and the bond market ignored the data. That is the finding. If the long end were pricing inflation expectations, the softest core print in a year would have produced a rally. It produced a 24-year high instead. Whatever is setting long-term borrowing costs is not the inflation outlook and not the policy path. The vice chair said as much the day before, arguing that rising yields do not signal shifting long-run inflation views. He is almost certainly right — and that is the problem rather than the reassurance. What remains is supply: a deficit that has to be funded, an issuance calendar that keeps growing, and an AI buildout whose financing needs compete for the same pool of capital. None of those respond to a 25 basis point decision in October. There is a second explanation that reaches the same destination from the opposite direction. Barclays argues the 30-year could reach 6% precisely if AI delivers sustained productivity gains, because faster growth raises the long-run expected policy rate. Higher yields if the buildout works. Higher yields if it does not. The Treasury has already tried to do something about it. Last Thursday it offered to buy up to $6 billion of 20 to 30-year debt and could only source $4.078 billion. It is trying again on Wednesday, this time in shorter maturities.