Goldman Sachs: AI Demand's Impact on U.S. Inflation Is Meaningful but Moderate, Seen Nearing Peak and Expected to Ease Next Year

Goldman Sachs released a research report stating that artificial intelligence (AI) demand is currently pushing up U.S. core personal consumption expenditures (PCE) inflation by about 45 basis points, of which roughly half is a real effect and the remainder stems from statistical measurement bias. The bank expects this impact to peak at 50 basis points by the end of 2026, then fall back to 10 basis points by the end of 2027.

The bank estimates the impact on headline PCE inflation will be an additional 5 to 10 basis points over the same period.

The bank analyzed four transmission channels.

First, AI is driving up inflation in software and accessories, but most of the roughly 20 basis point contribution from this category to core PCE inflation is estimated to be measurement bias, which exaggerates the impact of sharp increases in accessory prices; the failure to adjust software prices for quality improvements also accounts for part of the contribution.

Second, cost pressures on electronic components are pushing up prices of computers and other consumer electronics, estimated to contribute a combined 15 basis points by the end of next year.

Third, electricity demand from data centers is driving up power prices, estimated to contribute 5 to 10 basis points to core inflation through corporate energy costs and an additional 5 to 10 basis points to headline inflation through household electricity bills.

Fourth, rising prices of industrial metals used in data center construction are estimated to push up core inflation by about 5 basis points, though no significant spillover effect on construction worker wages has been observed.

In summary, the report said the impact of AI demand on inflation is meaningful but moderate, largely still stemming from measurement bias, and is already nearing its peak; it is expected to ease next year, mainly because sector analysts anticipate a significant slowdown in the growth of memory and computer prices.

Even if all AI effects are counted at face value and assumed to persist next year, the bank still expects core PCE inflation to fall significantly to near the 2% target next year, as tariffs and the Iran energy shock are estimated to account for 0.9 percentage points of the remaining 1 percentage point overshoot.