The simple version

The August consumer price report landed this morning. Prices rose 0.4% for the month and 3.4% over the year, and core prices, which exclude food and energy, rose 0.3% and 2.4%.

Inside the detailed tables, a category called computer software and accessories is up 25.4% over twelve months. That is a consumer category, meaning it measures what households pay rather than what companies invest, and this story has not previously appeared on that side of the economy.

The numbers

• All items rose 0.4% in August and 3.4% over twelve months; all items less food and energy rose 0.3% and 2.4% (Bureau of Labor Statistics, Consumer Price Index, released September 11, 2026)
• The category BLS names computer software and accessories rose 25.4% over twelve months, unadjusted (BLS, CPI Table 2, detailed expenditure categories, August 2026)
• The same category fell 2.2% in August itself, seasonally adjusted, so the twelve-month figure and the latest month point in opposite directions (BLS, CPI Table 2)
• Its relative importance in the index is 0.031, meaning roughly three hundredths of one percent of the basket (BLS, CPI Table 2)
• The Washington Post reported the twelve-month rise as the category's largest yearly increase on record. The BLS news release does not use the word record and does not mention the category (Washington Post, as reported)
• The Federal Reserve notes that the category covers more than software: its definition encompasses not only traditional software but also media storage devices, such as flash drives (Federal Reserve, FEDS Notes)
• The same Fed note ties storage prices to artificial intelligence directly, observing that the massive data storage requirements for AI training are largely met by solid state drives, which contain flash memory (Federal Reserve, FEDS Notes)
• The consumer price index measures prices paid by households, which is a different question from what businesses spend on capital equipment (BLS)

Two different halves of the same story

Everything written about the economics of artificial intelligence over the past year has lived on the corporate side. Companies spending hundreds of billions on data centers. Ratings agencies examining the debt behind it. Convertible bonds financing it. All of that measures what firms are doing.

The consumer price index asks a different question: what are households paying. A category inside it moving this much is a statement about prices at the household level, which is not where this story has lived.

The connection between the two halves is not our inference, which matters because the statistical agency publishes prices and does not attribute them to causes. The Federal Reserve published a note in May examining this specific category, and it draws the line explicitly: the storage requirements for training artificial intelligence models are met largely by solid state drives, and those drives contain flash memory.

So the mechanism runs through hardware rather than through software licenses. Demand for a component rose because of what is being built, and a consumer category containing that component rose with it.

The category is not what its name suggests

This is the first correction the data forces, and it changes how the number should be read.

The Federal Reserve note states that the BLS definition encompasses not only traditional software but also media storage devices, such as flash drives. So a category most people would read as meaning applications and subscriptions also contains physical storage hardware.

That distinction is the difference between two very different claims. Software subscriptions becoming more expensive is one story. Flash drives becoming more expensive because data center construction is consuming the world's memory chips is a different one, and the Fed's note points at the second.

The same note is candid about measurement. It observes that standard BLS practice relies on a matched-model approach, which tracks monthly price changes for identical items over time, and that the explicit absence of hedonic quality adjustments for software within the CPI means improvements may go unmeasured in price indices. A reader should hold the figure with that in mind.

Why a record increase can barely move the total

The second correction is scale, and it is the one most likely to be lost in coverage of a large percentage.

The index is a weighted average, and each category counts in proportion to its share of typical household spending. This category's relative importance is 0.031, roughly three hundredths of one percent of the basket. It is among the smallest lines BLS publishes.

A category that size can post an extraordinary percentage move and contribute almost nothing to the headline rate. Both statements are true at once: the increase is real and notable as a fact about that category, and it is not a driver of the inflation rate. Coverage tends to hold only the first.

There is a third fact that complicates the picture further. In August itself, the month this report covers, the category fell 2.2%. The twelve-month figure describes a year. The monthly figure describes the report it appeared in, and the two point in opposite directions.

The Real Cost lens on a category most people underestimate

Household spending on software and its accessories has changed shape in a way the category name obscures.

• Software used to be a purchase: a box, once, that kept working for years. Much of it is now a subscription that recurs monthly or annually
• A recurring charge behaves differently in a budget from a one-time purchase, because it renews without any new decision being made
• That also makes increases easier to miss, since a subscription that rises at renewal does not require anyone to walk into a store and see a higher number
• None of that is advice about any purchase or subscription, and a national category average does not describe what any household actually pays

The useful move for a reader is not to react to the index but to look at what recurring charges actually appear on their own statements, which is a question no national average can answer.

What this means

When a category posts a striking number inside an inflation report, three questions make it usable: what share of spending the category represents, what the latest month did as opposed to the year, and whether the explanation on offer comes from the agency or from somewhere else. All three answers were available this morning and none of them appears in a headline.

The broader point is that a large investment cycle eventually touches households somewhere, and finding where means reading the components rather than the summary. The components are published monthly, free, and mostly unread.

What this is NOT

This is not a prediction of inflation, prices, Federal Reserve decisions, or markets. This is not a characterization of inflation as controlled or uncontrolled. The statistical agency publishes prices and does not attribute categories to causes: the connection to artificial intelligence drawn here is the Federal Reserve's published analysis of this category, and the description of the twelve-month rise as a record is attributed to the Washington Post rather than to BLS. This is not advice about any purchase, subscription, security, or fund, and no company is named or evaluated. National category averages do not describe any individual household's costs, index figures are subject to revision, and the Federal Reserve's own note describes measurement limitations in this category. This is not investment or financial advice of any kind.

Sources

• U.S. Bureau of Labor Statistics, Consumer Price Index news release, August 2026: https://www.bls.gov/news.release/cpi.nr0.htm
• U.S. Bureau of Labor Statistics, CPI Table 2, detailed expenditure categories: https://www.bls.gov/news.release/cpi.t02.htm
• Board of Governors of the Federal Reserve System, Measurement of Computer Software and Accessories Inflation, FEDS Notes: https://www.federalreserve.gov/econres/notes/feds-notes/measurement-of-computer-software-and-accessories-inflation-20260522.html
• U.S. Bureau of Labor Statistics, CPI relative importance tables: https://www.bls.gov/cpi/tables/relative-importance/