US CPI July 2026: Headline 3.4%, Core 2.5% — In Line With Forecast, but Core Goods Turned Positive for the First Time in Three Months

The July Consumer Price Index landed on Wednesday 12 August 2026 at 8:30 a.m. Eastern, and on the four numbers everyone watches it did exactly what it was supposed to. Headline rose 0.1% on the month and 3.4% over twelve months, down from June's 3.5%. Core rose 0.2% and 2.5%, down from 2.6%. That is the Dow Jones consensus on all four lines, and it is the Cleveland Fed's nowcast to within six basis points — from a model that missed the same report by 36 basis points a month earlier. A print with no aggregate surprise sounds like a non-event. It is not, because the one line that did move is the line the Federal Reserve cannot describe as an energy shock.

Core goods rose 0.20% in July after two straight negative months. That is a small number attached to a large question, and it is where this report earns its attention.

What actually happened

Four numbers carry the release. Here they are against what was expected, and against the month before:

Every figure here is computed from the Bureau of Labor Statistics published index values rather than rounded off a headline, which is why they carry two decimals. Headline annual inflation fell about seventeen-hundredths of a point; core fell about eleven.

The component detail is where the report has texture:

The nowcast that missed by 36 basis points landed within six

Our preview of this release spent a section on the Cleveland Fed's nowcast and a longer one on why not to trust it. On 9 July, the day before the June data, the model had June headline at −0.06% month-on-month and core at +0.23%. The actual prints were −0.42% and −0.02% — a 36-basis-point miss on headline and 25 on core, on a one-day horizon.

This month the same model had July headline at +0.09% and core at +0.21%. The actuals were +0.07% and +0.22%.

The pump arithmetic worked exactly as advertised

This was the part of the preview most likely to read as wrong to anyone who drives, so it is worth closing the loop. Gasoline on EIA's weekly all-grades retail series:

Within July the price rose 8.1%, from $3.911 to $4.228. Across the months the July average sat 2.9% below June's. CPI reports the second fact, and the seasonally adjusted gasoline index duly fell 2.86%.

The raw month-average change and the published seasonally adjusted figure landed almost on top of each other — −2.9% against −2.86%. That is a coincidence of this particular month's seasonal factor, not a rule, and a reader should not expect the two to agree next time.

Core goods turned, and that is the line that matters

Core goods measures commodities less food and energy commodities: physical, tradeable things. It had been subtracting. In May it fell 0.11%, in June 0.09%. In July it rose 0.20%, and no month since September 2025 has printed higher.

Why a 0.2% move in one category outranks the aggregates: the Committee's July statement described inflation as elevated "in part reflecting supply shocks that have driven price increases in certain sectors, including energy." That framing is what allows a central bank to look through a price rise. It works for gasoline. It does not work for goods broadly, because a broad goods increase is not one sector and not one shock.

Two cautions keep this honest. One month is not a trend, and this category is noisy — it printed above 0.2% repeatedly through mid-2025 without signalling anything durable. And the July collection period ran largely before the 24 July tariff schedule took effect, so this is not yet a reading on tariff pass-through. Our note on the July PPI release works through that timing problem in the producer-price data, where the same collection-window issue applies.

The wedge narrowed, but check why

Subtract core from headline and you get what food and energy contribute on top. In June that was 0.94 points. In July it is 0.88.

That narrowing is not underlying disinflation. It is energy's annual rate easing from 15.7% to 14.7% while gasoline still runs 24.6% above a year ago. The two-economies problem the preview described is intact: one where a supply shock has pushed a narrow set of prices up by a quarter, and one where the underlying trend sits a few tenths above the 2% goal. July did not merge them. It moved the gap by six-hundredths of a point.

The dovish and hawkish readings of this same report therefore both survive, which is the practical definition of a print that settles nothing:

• Read headline and core goods. Inflation is 3.4%, has been above target for years, goods have stopped subtracting, and three Committee members have already voted to hike.
• Read core and shelter. Underlying inflation is 2.5% and falling, shelter is decelerating at 3.18% annual, and the labour market just shed 23,000 jobs with 103,000 cut from prior revisions — covered in our breakdown of the print that broke the dollar through 100.

What it leaves for September, and for the dollar

Going into this release, CME's FedWatch tool showed a 60% probability of a hold in September, having been 45% the day before the July payrolls report, and the ICE dollar index closed at 99.498 on 7 August, through the 100 level it had been defending. An in-line print gives that positioning no reason to move much on the data itself.

The transmission is the ordinary one. The interest-rate factor moves first on the core number, and core did not move it. The growth factor pulls the other way if tightening odds rise into a labour market already shedding jobs. And the commodity factor sits upstream of both, because the same barrel setting the energy line is setting the rate expectation. Where each currently scores the dollar against the other seven majors is on the USD factor page, and the eight-currency, five-factor approach is set out on our about page.

The Committee sees exactly one more CPI before it decides: the August print on 11 September, five days before the 15–16 September meeting and inside the pre-meeting blackout, when officials cannot publicly reframe it. That meeting also carries a Summary of Economic Projections, so the dot plot gets rewritten with it in hand.

Base effects still do the work into year-end

The August energy contribution is already being set at the pump. Gasoline printed $4.211 on 3 August and $4.141 on 10 August, so the month-to-date average of $4.176 sits 2.8% above the July average — the sign flips, and energy adds to headline instead of subtracting.

Then the base takes over. Freeze the price at the 10 August level of $4.141 — no forecast, no geopolitics — and the annual comparison does this:

Nothing has to happen for the energy contribution to headline inflation to grow into year-end; the 2025 floor simply falls away. That is the cleanest explanation for a market pricing a hold in September while still carrying meaningful hike probability for October and December. Those are not contradictory prices. They are a market reading the base-effect table.

The symmetry matters too. If crude falls materially from here, the same arithmetic runs in reverse and the annual rate collapses faster than a price chart suggests. The base is a lever with two ends, and July's report did not touch it.

Educational macro context only — not investment advice.