September CPI Hit 3.7%. Gasoline Explains the Headline; the Core Is the Harder Question

The September Consumer Price Index, published by the Bureau of Labor Statistics on October 14, 2026, showed headline inflation of 3.7% over twelve months, up from 3.4% in August. The monthly rise was 0.6% on a seasonally adjusted basis. Most of the coverage has treated gasoline as the story, and for the headline it is. But the number the Federal Reserve will argue over at its October 27-28 meeting is core inflation, which rose to 2.6% from 2.4%, and one popular explanation for that rise, record prices for computer software, turns out to be too small to carry it.

What the print said

The figures below combine the BLS release itself with consensus numbers reported by IndexBox and Briefs. The BLS page we read gives actuals only, so the expectations column is as those outlets reported it.

• Headline CPI: +0.6% month on month (seasonally adjusted), +3.7% over twelve months. Reported consensus was 0.5% and 3.6%. August was 3.4% year on year.
• Core CPI (excluding food and energy): +0.3% month on month, +2.6% year on year. Reported consensus was 0.2% and 2.5%. August core was 2.4%, the lowest reading since March 2021 according to Trading Economics.
• Energy: +5.2% on the month and +20.4% year on year.
• Gasoline: +9.1% on the month (seasonally adjusted) and +34.2% year on year. Unadjusted gasoline prices rose 6.5% in the month, per the BLS release.
• Shelter: +0.4% on the month, +3.2% year on year.
• Commodities excluding food and energy (core goods): +0.3% on the month, +0.7% year on year. Used cars rose 0.7% on the month but are still down 1.5% year on year.

On the market side, IndexBox reported the probability of a quarter-point hike at the October meeting at 43.7% after the release, against roughly 20% before it, while Briefs put it at about 45%. IndexBox also reported markets pricing better-than-even odds of a December hike and the 10-year Treasury yield above 5.3%. Briefs reported that the Fed raised rates in September, its first increase since 2023, and that officials' median projection still shows one more rise this year. We have not independently confirmed the futures levels, which move by the hour.

Why gasoline is the easy half

Gasoline rose 9.1% in a single month and is up 34.2% over twelve months. IndexBox reported that gasoline accounted for more than half of the overall monthly increase. Briefs described it as the largest monthly rise since the Iran war began in March. That is a supply shock, and policy makers generally treat supply shocks differently from demand: a one-off jump in the price level raises the headline for twelve months and then drops out of the year-on-year comparison, unless it feeds into wages and the prices of other things.

One detail in the BLS release deserves more attention than it has received. The seasonally adjusted gasoline rise was 9.1%, but the unadjusted rise was 6.5%. The gap of 2.6 percentage points means the seasonal factors expect gasoline to be softer in September than the raw data showed, so the adjustment amplified the move. That does not make the price rise less real for a household at the pump, and the twelve-month headline of 3.7% is calculated on unadjusted data, but it is a reminder that the 0.6% monthly figure is partly a product of the adjustment. A reversal in crude oil would show up quickly in both numbers.

The risk that makes energy more than a one-off is pass-through. Ben Ayers, senior economist at Nationwide, said the firm core readings raise concern that fuel costs are passing into other goods and services. That is a plausible channel, since transport and delivery costs sit inside the price of almost everything, but the September core goods data do not yet show a broad version of it: core goods rose just 0.7% over twelve months, and used car prices are still falling year on year.

The core question: what moved 2.4% to 2.6%

Core CPI rose 0.3% in the month against an expected 0.2%. It is worth putting that in perspective. A 0.3% monthly gain, if repeated for twelve months, compounds to roughly 3.7% annualised, well above the 2.6% year-on-year rate. A single month rounds to the nearest tenth, so the true figure could sit anywhere from about 0.25% to 0.35%, which is an annualised range of roughly 3.0% to 4.3%. The sensible reading is that recent momentum is running above the twelve-month trend, not that the trend has been broken. August core was also 0.3% on the month, so this is the second such reading in a row, which is the kind of persistence the Fed pays attention to.

Within core, shelter rose 0.4% in the month. Briefs reported that owners' equivalent rent, the largest single component of shelter, rose 0.3%. Anna Wong and Troy Durie of Bloomberg Economics, as quoted by Briefs, argued that the composition of the core rise muddies the case for a hike and suggested the hot rent reading could prove an aberration. Briefs also reported that core goods posted their strongest increase since January 2025, helped by used cars, apparel and a record jump in toys, though the article gave no figure for the toy increase.

The software and computer prices: real, but sized wrong

The detail that spread fastest was Briefs' report that CPI computer software and accessories rose a record 28.4% year on year, and that computers, peripherals and smart home assistants rose a record 9.7%. Briefs attributed both to pressures tied to the data-center buildout. That attribution is the outlet's framing; the article cites no statistician or economist for it, and the BLS release page we read does not break out these categories at all, so we are reporting the percentages as Briefs' and have not confirmed them against the BLS detailed tables.

Assume they are right. The question is how much they can move the core index, and a Federal Reserve staff note answers it. In a FEDS Note dated May 22, 2026, economists Alessandro Barbarino, Michael Diercks and Marc Miran reported that computer software and accessories made up just 0.035% of the core CPI basket as of March 2026. Their note also found the category had already risen at a 73.1% annualised rate between November 2025 and March 2026.

Multiply the two: a 28.4% rise on a 0.035% weight contributes about 0.01 percentage points to core CPI over twelve months. The weight changes over time and the March figure will have drifted, but even a weight ten times larger would contribute only about 0.1 points. Core rose 0.2 points between August and September on the year-on-year measure. Software prices cannot be what moved it. We could not find a published weight for the computers and peripherals category in the sources we read, so that line item is unsized here, and we do not claim it is also negligible.

The same Fed note raises a second reason to be cautious about the headline percentage. The CPI category is roughly 80% software, 9% apps and 11% portable memory by weight, according to the authors, and it includes flash drives and blank media. The authors found flash drive prices up about 69-70% since November 2025 and said the index has no explicit quality adjustment, which may not capture AI-driven software improvements or the shift to subscription pricing. They described their estimates as illustrative. In other words, a record year-on-year print in this category partly reflects what is inside it and how it is measured.

A mechanism worth watching: input costs versus shelf prices

If AI-infrastructure demand is raising the price of memory and compute, consumer prices would feel it through a specific chain: component makers raise prices, device makers pass some of the increase to retailers, and retailers pass some to shoppers. Each step has a lag and a weight. The categories most exposed, such as storage media, PCs and smart-home devices, are small in the CPI basket, which is why a large percentage rise registers as a rounding error in core. The larger consequence would be if the same pressure spread to bigger categories such as phones, appliances or vehicles with heavy electronics content. Nothing in the September data shows that yet: new vehicles rose 0.6% year on year and apparel 3.4%.

What it changes for October 27-28

These are scenarios, not predictions. In the case for a hike, core has now come in at 0.3% twice running, headline is 3.7% and still rising, and Kevin Warsh's Fed has already raised once. The cost of misjudging is asymmetric if energy pass-through takes hold, because inflation expectations tend to follow headline prices. In the case for waiting, the biggest contributor to the headline is a supply shock that policy rates do not directly fix, the core goods trend is only 0.7%, and the Bloomberg Economics analysts argue the rent print may not last. A reversal in gasoline would take much of the pressure off the headline within a month or two, so a patient committee could reasonably wait for the November data.

For households the practical effects run through rates and wages. A 10-year yield above 5.3%, if it holds, feeds mortgage pricing. Briefs also reported that inflation-adjusted average hourly earnings fell year on year for a sixth straight month, which is the squeeze that gasoline-led inflation produces, and that the Social Security Administration announced a 3.6% cost-of-living adjustment for 2027, the largest in about four years, based on third-quarter CPI. Small-business owners with floating-rate debt have the most direct exposure to a hike.

Readers who want the earlier context can see our October 12 preview, which looked at the gap between core CPI and core PCE before this print. This piece deals only with the September release, which has since moved the core rate to 2.6% and the hike odds sharply higher.

Bottom line

Gasoline explains the headline, and the seasonal-adjustment gap means even that is somewhat overstated month to month. The core story is a second 0.3% reading and a 0.2-point rise in the year-on-year rate, driven by shelter and a broad mix of goods and services rather than by the headline-grabbing software line. The data-center explanation for computer prices is interesting and may prove important later, but at a 0.035% weight it did not move the Fed's decision this month. What would change that is evidence of pass-through into larger categories, which is the thing to watch in the October report.