The latest estimates from Goldman Sachs (GS) economist Megan Peters show that three factors driven by AI—soaring memory prices, price hikes in software services, and rising electricity demand from data centers—have collectively pushed up the year-over-year core PCE (Personal Consumption Expenditures Price Index) inflation rate in the United States by more than 0.2 percentage points. Goldman Sachs projects that by the end of 2026, this contribution could rise to 0.5 percentage points, and this estimate does not fully account for various spillover effects, meaning the actual impact could be even greater.
This development has already drawn attention within the Federal Reserve. The latest Federal Open Market Committee (FOMC) meeting minutes explicitly noted for the first time that a majority of officials are concerned about the potential scenario of “persistently elevated inflation driven by strong AI-related demand.” The Federal Reserve’s Semi-Annual Monetary Policy Report also listed “increased demand for high-tech products that support AI applications” as one of the factors pushing up prices. However, at the policy response level, a clear divergence of opinion has emerged among senior Fed officials. New York Fed President John Williams publicly emphasized that if AI demand continues to impact supply-demand dynamics and drive inflation, this effect “should not be ignored.” This stance contrasts with the earlier view expressed by Fed Chair Warsh, who wrote in November 2025 that AI is a “significant deflationary force” that would boost productivity in the long run. Although the long-term logic holds, the actual pass-through of surging memory prices to consumer electronics costs is already an indisputable reality.
Goldman Sachs quantified the transmission channels of AI’s impact on prices across three dimensions. The foremost is memory prices. Since early 2025, robust demand from data centers has pushed up prices for certain types of memory by more than tenfold. Goldman Sachs estimates that the price index for software and accessories, which is highly correlated with memory prices, will peak at a year-over-year increase of approximately 30% in November 2026, contributing about 36 basis points to core PCE. Major consumer electronics manufacturers such as Apple, Microsoft, and Dell have recently announced price increases of up to 25%, which directly reflect this trend. While memory price hikes are a global phenomenon, their impact is particularly pronounced in the United States, given that the relevant weight in the U.S. PCE basket is far higher than in other developed markets such as the euro area and the United Kingdom, where the peak contribution averages only about 5 basis points.
The second channel is software subscription price increases. In January 2026, Microsoft raised the subscription price for its consumer version of Microsoft 365 for the first time since the subscription model was introduced in 2013, citing the addition of AI Copilot features. U.K. statistics show that software prices surged 20% quarter-over-quarter in March 2025, marking the highest quarterly increase on record. Since the weight of software in inflation baskets is extremely low in most developed economies, this effect is relatively limited outside the United States. In the U.K., for example, a 20% overall software price increase contributes less than 4 basis points to core inflation.
The third transmission chain comes from electricity price increases. Electricity demand from data centers has begun to push up residential electricity bills in the United States, particularly in regions with dense data center concentrations. Goldman Sachs estimates that the electricity price channel currently contributes about 8 basis points to year-over-year core PCE inflation, with some PJM grid states experiencing impacts far above the national average. Medium- to long-term pressures are also not to be underestimated. Goldman Sachs projects that the share of U.S. data center electricity consumption in total national electricity usage will rise from the current level of about 6% to 11% by 2030, nearly doubling. As a reference, in Ireland, where data centers account for 23% of electricity consumption, related construction has cumulatively added about 360 euros to household energy bills between 2015 and 2023.
If Goldman Sachs’s projections hold true, the additional 0.5 percentage point contribution by year-end would significantly compress the Federal Reserve’s room for rate cuts within the year and could force policymakers to rethink their prior reliance on the “long-term deflationary” effect of AI.
