For two years the story of the AI boom has been told in capital expenditure: the hundred-billion-dollar data-centre commitments, the GPU order books, the venture funds raised to wire up the physical layer. New analysis from the Federal Reserve Bank of Minneapolis adds a line to that story most people have not been watching — the boom is now showing up in what Americans pay at the till.

Minneapolis Fed economists Neil Mehrotra and Michael Waugh, in research reported on 29 August, find that surging demand for the memory and computer hardware the AI build-out runs on has pushed up the price of video and information-processing equipment by 12.2% year over year through July 2026. That single category has added roughly 0.4 percentage points to core PCE inflation — the Fed's preferred gauge — which is at least as much as the entire contribution of the tariffs imposed early last year. For comparison, the same analysis puts the tariff contribution to core PCE at 0.2 to 0.4 percentage points as of July.

Put plainly: the AI investment wave is now doing to consumer prices roughly what a trade war did.

An extraordinary reversal

What makes the figure striking is not just its size but its direction. This is a category of goods — laptops, monitors, processing equipment — that for years got cheaper as manufacturing improved. From 2015 to 2019, prices here fell by about 6.5% a year. Consumers had come to expect that the same money bought more computer every cycle. A category that was reliably deflationary has swung hard the other way, and the analysis pins the reversal on AI hardware demand.

The mechanism is straightforward supply and demand, playing out at industrial scale. The largest technology companies are simultaneously bidding for the same finite pool of central processors, graphics processors and memory to feed their AI infrastructure. That concurrent scramble squeezes what is left for everyone else, and the pressure travels down the chain to the devices ordinary buyers pick up — a MacBook, a work laptop, a monitor. The chips that train frontier models and the chips inside a consumer machine are drawn from overlapping supply.

Why it matters beyond the spreadsheet

The finding reframes a debate that has mostly treated AI and inflation as separate subjects. Through 2026 the inflation conversation has been dominated by tariffs — how much of the trade measures would pass through to prices, and how long the Fed should look through them. This analysis argues there is a second, quieter driver sitting alongside the tariffs, and that it is not a policy choice anyone voted on. It is a side-effect of the buildout itself.

It also sharpens the question of who bears the cost of the AI race. The headline spending is borne by the hyperscalers and their investors. But if the demand those firms generate is lifting the price of a laptop for a household or a small business, some of the bill is being spread far more widely than the balance sheets suggest — and largely invisibly, because nobody labels it.

For central bankers, it complicates an already awkward call. Tariff-driven inflation is often argued to be a one-off level shift that policy can see past. Hardware inflation driven by a multi-year investment supercycle is harder to dismiss as temporary, because the demand behind it is not going away on any near horizon — the commitments already announced run for years.

None of this makes the AI boom a bad bet, and the analysis does not argue that it is. Productivity gains, if they arrive at the scale the investment implies, could more than offset a few tenths of a point on a price index. But it is a reminder that a build-out this large does not stay confined to the companies making it. It is now measurable in the general price level — and Minneapolis Fed economists have now placed AI hardware alongside the tariffs as a driver.