AI Boom Is Starting to Push Technology Prices Higher, Adding to Fed’s Inflation Problem

The U.S. artificial-intelligence buildout is beginning to reverse a long-running source of disinflation in technology, with prices for electronic components, software and data-center inputs climbing as the Federal Reserve is still trying to return inflation to its 2% goal.

The shift does not mean AI is the main reason inflation remains above target. Energy, housing, tariffs and other forces still matter. But the scale of AI investment is now large enough that economists and Federal Reserve researchers are treating its near-term price effects as a measurable part of the inflation picture rather than a theoretical risk.

That distinction matters because technology has historically tended to become cheaper, or at least deliver more computing power for the same money. Now demand for the hardware and infrastructure needed to train and run AI models is straining some of the same supply chains that once helped pull goods inflation lower.

Technology prices are moving in the wrong direction for the Fed

The clearest evidence is appearing first in producer prices. The Bureau of Labor Statistics’ July producer-price report showed prices for electronic components and accessories were 28% higher than a year earlier, even after a 0.7% decline in July itself. Switchgear, switchboards and industrial-control equipment were up 9.3% from a year earlier, while air-conditioning and refrigeration equipment rose 4.6%. Those are all categories with exposure to the data-center construction boom.

The same report showed materials and components for construction up 4.8% over the year. Final-demand construction prices were 5.2% higher, while prices for construction used for private capital investment also rose 5.2%. Not all of those increases can be assigned to AI, but the figures show the broader cost environment into which hyperscalers and data-center developers are pouring capital.

A June analysis from the Federal Reserve Bank of Richmond found that the price increases in several AI-linked categories were at or near multi-decade highs. Semiconductor and other electronic-component manufacturing prices had risen 26% from a year earlier in April, a record in data going back to 1984. Computer and peripheral-equipment manufacturing prices had posted a 7.8% annual increase in February, also a record for that series.

The Richmond Fed also found unusually strong inflation in software and power-related equipment. Application-software publishing prices were up 8.5% in May from a year earlier, and engineering services tied to utility and power projects were up 7.3%. The point is not that every price increase in those categories was caused by AI, but that the timing and concentration of the moves are consistent with a capital-spending boom that is competing aggressively for computing, electrical and construction capacity.

Consumer inflation is showing a version of the same story. The Richmond Fed noted that the personal-consumption-expenditures price index for computer software and accessories was up 14.5% in May from a year earlier, the fastest increase in the history of that series. That is striking because software prices had generally fallen for decades.

Some of the measured software inflation may be overstated

There is an important measurement warning attached to the software numbers. A Federal Reserve staff note published in May found that the PCE category for computer software and accessories had made an unusually large contribution to core inflation since late 2025, but it also identified reasons the official measure may be overstating the true increase in prices.

One issue is that the consumer-price index used to help construct the PCE software measure includes products such as portable memory and blank media, while the PCE category itself is much more heavily weighted toward software publishers and data-processing services. The Fed researchers also noted that standard price indexes may struggle to separate a higher sticker price from a genuine quality improvement when software is rapidly adding AI features.

That caveat cuts against the most dramatic interpretation of the data. It would be too strong to say the full rise in measured software prices represents pure inflation caused by AI. The same Fed research estimated that correcting for category mismatch and some quality effects could reduce the apparent contribution substantially.

Even with that adjustment, the physical side of the AI buildout is harder to dismiss. Memory chips, circuit boards, storage devices, electrical equipment and cooling systems are real inputs whose prices have been rising sharply. The Federal Reserve staff note specifically linked the surge in flash-memory prices to the data-storage requirements of AI training and the broader data-center investment boom.

CIBC Capital Markets economists Helen Lao and Avery Shenfeld tried to put a number on the combined effect in a July research note. Using higher-than-normal inflation in information-processing equipment and electricity, plus an estimate of how AI-driven investment and stock-market wealth are affecting aggregate demand, they calculated that AI-related forces could add roughly 0.4 percentage point to U.S. inflation in 2026.

The CIBC estimate is not an official measure and depends on several assumptions. The economists described their direct-price approach as a rough guide, and part of the calculation uses forecasts for 2026 growth. Still, it gives a sense of why a few tenths of a percentage point can matter when inflation is already above the Fed’s objective.

AI creates a timing problem for monetary policy

The policy problem is that AI may be inflationary before it becomes disinflationary. Over time, better software, automation and faster decision-making could raise productivity and lower the cost of producing goods and services. In the buildout phase, however, companies are spending heavily on scarce chips, power equipment, data centers, engineering services and network capacity.

That sequence leaves the Federal Reserve dealing with the costs first. The Fed’s preferred PCE price index was 3.7% higher in June than a year earlier, while core PCE inflation was 3.3%. The July consumer-price index offered some relief, with headline CPI easing to 3.4% and core CPI to 2.5%, but those readings are still not the same as a return to the Fed’s 2% PCE inflation target.

The Federal Open Market Committee held the federal-funds target range at 3.5% to 3.75% on July 29. Three officials dissented in favor of a quarter-point rate increase, underscoring that the debate has already shifted toward whether policy is restrictive enough to finish the inflation fight.

AI does not automatically imply another rate increase. The Fed has to judge whether higher technology and infrastructure costs are persistent, whether they spread into broader prices and wages, and whether productivity gains begin to offset them. It also has to separate genuine inflation from statistical noise in fast-changing products such as software.

There is another complication: the AI boom is supporting economic growth at the same time it is pushing up some input prices. Federal Reserve staff research published in July found that software, data centers, power facilities and computer equipment made a meaningful contribution to GDP growth from 2025 through the first quarter of 2026. Strong investment can keep demand firm even when higher interest rates are restraining housing and other rate-sensitive sectors.

That makes the near-term tradeoff unusually awkward. If the Fed raises rates to lean against an investment-driven inflation impulse, it could slow the very capital spending that may eventually deliver productivity gains and lower costs. If it waits too long and AI-related price pressure proves persistent, inflation could remain above target for longer.

The next major checkpoint will be the July PCE report on Aug. 26. That release will give the Fed a more current reading of its preferred inflation gauge and help show whether the recent technology-price surge is becoming a broader monetary-policy problem or remains concentrated in a handful of AI-sensitive categories.