The Economy — Digest
TL;DR
The week before Wednesday's September CPI has been a one-way repricing: October hike odds fell from 55% to 22%, and the debate is now whether September was the only hike of the cycle. Meanwhile, the freshly released FOMC minutes settle an open question from last time — the AI buildout is no longer one official's hobbyhorse; it's a named, committee-level inflation channel, with Fed staff not projecting a return to 2% until 2029.
The October hike is dying quietly, and the argument is now about whether there's a second one at all
The dovish slide has been fast and complete: October hike odds went from roughly 70% before the Sept 30 BEA revision, to 37% after it cut core PCE to 3.0%, to — as of Monday — 22% (CPI Eve finding). Per Morningstar: "As of Monday, investors saw a 22% chance of an October hike, down from 55% immediately following the Fed's September rate call." The print itself is expected to be a split personality — headline at +0.5–0.6% m/m / 3.6–3.7% y/y on an energy surge, core cooling to +0.2% m/m / 2.4% y/y — and markets will trade the core, not the headline.
"September's CPI is set to look alarming on the surface and calm underneath, and the Fed has to decide which one to believe." — investinglive CPI preview
The most aggressive call on the board is Natixis, forecasting core at just +0.19% and arguing "it is quite possible that September was the lone hike of the cycle" — with an October hold conveniently avoiding "the political fallout from hiking just before mid-term elections in early November." Against that, Musalem (Oct 8) is still explicit that "more monetary policy firming will be required," and the minutes split the difference: another increase "would likely be appropriate by year end," but with an open mind at each meeting. The calendar is compressed — CPI Wednesday, PPI and retail sales Thursday, Warsh in Bangkok early Friday before the blackout — so by the end of this week the December question is essentially the only question left.
What to watch: Wednesday 8:30 a.m. core CPI — 0.2% locks the hold and makes Natixis's "lone hike" thesis live; 0.3% or hotter revives October.
The AI buildout just became a standing member of the Fed's inflation-risk set
The open thread from last cycle — is the AI buildout actually in the Fed's inflation framework, or just Waller's framing? — resolved this week, and the answer is yes, at the committee level. The Sept 15–16 minutes, released Oct 7, contain three separate AI-inflation passages, including: "Some participants commented that the AI buildout could cause aggregate demand to outpace aggregate supply over the medium term, putting upward pressure on inflation" (AI buildout finding). The policy implication is spelled out too — a couple of participants argued a higher policy rate "would help prevent sector-specific price increases stemming from energy market disruptions and AI-related demand from broadening out."
"inflation is projected 'to reach 2 percent in 2029,' with the path contingent on 'the effects of tariffs, geopolitical developments, and the AI buildout' waning." — AI buildout finding
That 2029 staff baseline is the number that matters: the Fed's own work sees roughly five more years of above-target inflation unless the AI buildout's price effects fade. The minutes also tie the buildout directly to the long-end selloff — Treasury yields rose partly on "competition for capital from heavy private debt issuance" for AI infrastructure — and describe a two-tier labor market where AI-adjacent skilled workers are pulling strong wage gains while aggregate wage growth stays moderate. Practically, this gives hawks a demand-side argument that doesn't depend on oil coming back down, which raises the ceiling on the hiking cycle independent of the war. BofA's pre-CPI framing shows the market watching the same channel: second-round effects from "higher energy prices and AI-related demand should remain relatively contained" — and Wednesday's soft-core consensus (+0.2%) is exactly the test of "contained."
What to watch: whether a hot core print gets attributed to AI-related demand, not just energy — that's what would make the channel rate-relevant rather than narrative.
Household pressure is at records even as the rate debate cools
The consumer backdrop into the print is ugly in ways the core number won't show. The NY Fed's September survey put one-year inflation expectations at 3.9%, up from 3.6% — the highest since May 2023 — and AAA's national regular-gasoline average sat at ~$4.37 in early October, "the first year the national average has been above $4 per gallon in October," up 39.4% y/y, with LendingTree pegging diesel up 72% (CPI Eve finding). The Sept 16 minutes put the distributional point plainly: "low- and moderate-income households faced strains, with higher energy prices weighing disproportionately on their real disposable income." This is the same tension from the U-Mich data last cycle, now sharper — expectations rising at the household level while market-priced October odds collapse. And the energy channel stays live: Brent is still above $100 after the Houthi attack on Riyadh's airport killed at least 12, which keeps Goldman's oil-toward-$85 condition for an October hike nowhere in sight.
What to watch: whether the December decision gets framed as a choice between the energy shock and the election — Natixis has already made that trade explicit.
What surprised us
- The "lone hike" thesis has a name on it. Last cycle's digest called December the knife-fight; Natixis now argues September may be the only hike of the entire cycle — a call that would have seemed fringe at 70% October odds a few weeks ago.
- The Fed's staff sees 2% inflation in 2029. Not a hawk talking point — the staff baseline. Five years above target, contingent on the AI buildout's price effects fading, is a much longer horizon than the rate debate implies.
- The AI buildout resolved from question to fact in one release. The minutes contain three distinct AI-inflation passages plus a Treasury-market transmission mechanism — this went from "one official's speech" to committee record between digests.
- Expectations are rising while rate odds fall. Households at 3.9% one-year expectations (highest since May 2023); markets at 22% October odds. Someone is wrong, and it's usually the households who get believed by the Fed.
Open threads worth a vote
- Brent crosses $110 on Saudi infrastructure damage — still the trigger that would resurrect the October hike and blow up the "lone hike" thesis in one print.