Inflation Cooled to 3.4% Because Fuel Got Cheaper. Everything Else Sped Up.
July CPI landed at 8:30 this morning. The headline softened, gasoline fell 2.9%, and core inflation went from flat to rising — three facts that only look contradictory if you read the top line.
The Bureau of Labor Statistics published July's Consumer Price Index at 8:30 this morning, and the number people will repeat all day is 3.4 percent — all items, over the last twelve months, down from 3.5 in June.
Cooling. Modestly, undramatically, cooling.
Now here are three more sentences from the same release, and they complicate that considerably.
"The gasoline index decreased 2.9 percent over the month." · "The index for all items less food and energy rose 0.2 percent after being unchanged in June." · "The index for shelter rose 0.1 percent in July, accounting for roughly two-thirds of the monthly all items increase."
Read together: the headline softened because fuel got cheaper, while the part of the basket that isn't fuel accelerated from flat to rising. Both things happened in July. Only one of them made the summary.
The two-thirds sentence
Of everything in this release, the line worth taping to a wall is this one:
“The index for shelter rose 0.1 percent in July, accounting for roughly two-thirds of the monthly all items increase.
”
Sit with the arithmetic implied there. Shelter went up by one tenth of one percent — a rounding error in the life of anybody paying rent — and that tiny move produced about two-thirds of the entire month's inflation.
That is not a statement about housing costs. It's a statement about weight. Shelter is roughly 35.6% of the index, and when something that large moves even slightly it swamps components that moved far more. Gasoline fell twenty-nine times further than shelter rose, and it still didn't run the number.
We published a piece this morning on exactly this — that the CPI basket is not your basket, that shelter dominates it, and that a quarter of the whole index is owners' equivalent rent, an imputed figure derived from asking homeowners what they think their home would rent for. I did not expect BLS to hand over a one-sentence proof of it on the same morning. It did.
What actually moved
The shape of that chart is the story. One large negative, several small positives, and a headline that lands almost exactly in the middle because the negative and the positives cancelled.
If you buy a lot of fuel, July was genuinely good. Delivery operations, anyone running vehicles, anyone whose freight is priced off diesel — your input costs fell, and by considerably more than the headline suggests.
If you don't buy much fuel, July was slightly worse than June. Core went from unchanged to +0.2%. Nothing in the headline tells you that, and the headline is what your landlord, your lender and your customers will have heard.
The trap in a cooling headline
A softening headline driven by energy is the least durable kind of good news, because energy is the component most likely to reverse. Fuel prices move on supply decisions and geopolitics on a timescale of weeks; shelter and services move on lease cycles and wage settlements over quarters.
So a month where energy does the cooling is a month where the cooling can be undone quickly, while the slower-moving components underneath carry on doing whatever they were already doing — which in July was accelerating slightly.
This is not a forecast. It's a statement about which numbers are volatile, and it's the reason core exists as a series at all.
The one to watch if you sell food
Food away from home rose 0.3% in the month against overall food at 0.1%.
That gap matters commercially. Prepared food carries labour, occupancy and energy inside its price in a way groceries don't, so it tends to track services inflation rather than commodity prices. A restaurant, a café, a caterer, anyone with a food line in an office — your costs are on the 0.3% track, not the 0.1% one, and definitely not the −2.9% one.
It also connects to something this desk has been following: transactions at small businesses have now fallen year over year for nine straight months while sales held up on ticket size alone. If prepared-food prices keep rising faster than food generally, that trade-down pressure gets stronger, not weaker.
What to do with this today
- 1
Work out your own exposure to the number that fell
Fuel is the whole reason the headline looks calm. Take last month's spend and ask what share is fuel or fuel-linked — freight, delivery, plant, anything with a surcharge. If it's meaningful, your July was better than 3.4% implies. If it's near zero, ignore the headline entirely; the relevant figure for you is core at 2.5%.
- 2
Do not reprice off a headline that energy is holding down
Setting prices to a cooling headline works right up until fuel turns, at which point you have locked in a number set by the most reversible component in the basket. Price against your own input mix — which is a different exercise from indexing to CPI, and the one that protects gross margin.
- 3
Read the monthly figure alongside the annual one, not instead of it
Annual: both headline and core cooled a tenth. Monthly: core went from 0.0% to 0.2%. Neither is wrong and they are answering different questions — where we've been over a year, versus what happened in the last four weeks. Anyone quoting only one at you has made a choice about which story to tell.
- 4
Wait for the revisions before treating any of this as settled
Seasonally adjusted CPI figures get revised, and the labour data this desk covered yesterday was restated by 103,000 jobs across two months without much comment. First prints are estimates. Act on the direction; hold the decimal place loosely.
Frequently asked questions
If shelter barely moved, why does it get blamed for everything?
Because contribution is weight times change, and shelter's weight is enormous — roughly 35.6% of the index. A 0.1% move on a third of the basket beats a 2.9% move on a component worth a fraction of that. It isn't blame, it's arithmetic, and it's exactly why "shelter drove inflation" headlines can appear in months when rents were nearly flat.
So is inflation getting better or worse?
Both, honestly, depending on which question you're asking — and I'd distrust anyone giving you a single word. Over twelve months it improved: 3.4% headline and 2.5% core, each a tenth below June. Within July itself, core went from unchanged to +0.2% while energy fell. The most defensible summary is that the annual trend is still easing and the monthly detail is less reassuring than the headline, which is a sentence with no headline in it.
Should I be pleased that gasoline fell 2.9%?
If you buy fuel, yes, genuinely — that's real money and it's the largest single move in the release. The caution isn't about enjoying it; it's about building on it. Energy is the most volatile component in the basket, so a plan that depends on July's fuel price holding is a plan with a specific, identifiable failure mode. Take the saving, don't capitalise it.
Why does this piece keep pointing at the other CPI article?
Fair, and I'd rather say it plainly: that piece explains the mechanism — the weights, owners' equivalent rent, why the basket isn't yours — and this one applies it to a specific morning. If you only read one, read the mechanism, because it works every month and this one is about July.
Two figures, one release
The thing I'd hold onto isn't 3.4%. It's that a release can contain gasoline down 2.9% and core turning up and be summarised, accurately, as inflation cooling.
Nobody lied. The headline is correctly computed and it genuinely is lower than June. It's just that a single number laid over a basket of thousands is a summary, and every summary is a decision about what to leave out — in this case, that the component that fell is the one most likely to come back, and the component that rose is the one that tends to stay.
Read the parts you actually buy. They're published, they're free, and they were sitting three paragraphs below the number everyone quoted.
It's just business — and the average was never yours.
Sources
- U.S. Bureau of Labor Statistics — Consumer Price Index Summary, July 2026 (released August 12, 2026)
- U.S. Bureau of Labor Statistics — CPI relative importance of components, December 2025
- U.S. Bureau of Labor Statistics — Measuring Price Change in the CPI: Rent and Rental Equivalence
- U.S. Bureau of Labor Statistics — CPI release schedule
Keep reading
This article is educational and satirical content from Business Dog. It is not financial, legal, or tax advice. It's just business.