``` AI intensifies U.S. inflation? Goldman Sachs: Rising prices of memory, electricity, and software may push core PCE up by 0.5 percentage points by the end of the year. ```

The AI boom is reshaping the U.S. inflation landscape in surprising ways.

Goldman Sachs economist Megan Peters calculates in her latest report that the triple effects of AI-driven memory price spikes, software price increases, and rising electricity costs have already lifted the U.S. core PCE year-on-year inflation rate by more than 0.2 percentage points, with the estimate rising to 0.5 percentage points by the end of the year. This estimate does not fully reflect various spillover effects, so the actual impact may be even greater.

This warning has resonated within the Federal Reserve. For the first time, the latest Federal Open Market Committee (FOMC) minutes specifically noted that most officials are concerned about "inflation remaining high due to strong AI-related demand."

The Fed’s semiannual monetary policy report also listed "increased demand for high-tech products supporting AI applications" as one factor driving prices up.

Fed Faces Internal Dissent

The issue of AI-driven inflation has sparked rare public disagreement among Fed executives. New York Fed President John Williams stated in a speech that if AI demand "continues to shock supply and demand and pushes up inflation, I think this situation should not be ignored."

This stands in clear opposition to Fed Chairman Powell's stance. Powell wrote last November, characterizing AI as an "important deflationary force," believing it would enhance U.S. competitiveness through productivity gains.

While this assessment may have merit in the long term, the surge in memory prices is currently causing undeniable cost increases in consumer electronics.

Three Main Channels Raising Core Inflation

Goldman Sachs quantified AI's inflation transmission routes from three perspectives.

Memory prices are the most direct channel. Since early 2025, strong data center demand has driven some memory prices up more than tenfold.

Goldman Sachs forecasts that software and accessory price indices, highly correlated with memory prices, will peak at a year-on-year increase of about 30% in November, contributing roughly 36 basis points to core PCE. Recent price hikes of up to 25% by major consumer electronics companies like Apple, Microsoft, and Dell directly reflect this trend.

Memory price increases are a global phenomenon, but the impact varies by country. In the Eurozone, UK, Canada and other developed markets outside the US, categories directly affected by memory have peak inflation contributions of about 1 to 9 basis points, averaging about 5, significantly lower than in the US—mainly because the relevant weights in the US PCE basket are much higher than in other countries.

Software price hikes are the second channel. In January 2025, Microsoft raised the price of consumer Microsoft 365 subscriptions for the first time since launching in 2013, citing the introduction of the AI Copilot feature.

UK statistical data shows that in March 2025, software prices posted a record-high quarterly increase of 20%. As software has very low weight in the inflation baskets of most developed countries, this effect is relatively limited outside the US—in the UK, a 20% overall software price hike contributed less than 4 basis points to core inflation, and even less elsewhere.

Rising electricity prices form the third transmission chain. Data centers’ electricity demand has begun to push up residential electricity bills in the US, with the effect especially notable where data centers are concentrated.

Goldman Sachs estimates that current data center power demand has contributed about 8 basis points to core PCE inflation through the electricity price channel, with some PJM grid states seeing far higher impact than the national average.

Data Center Power Demand Continues to Rise

The medium- and long-term trends of electricity price pressure are also concerning.

According to Goldman Sachs research forecasts, the proportion of U.S. power consumption by data centers will rise from about 6% currently to 11% by 2030, nearly doubling. By contrast, EU data centers’ share of power use has stayed at 3%-4% and is expected to rise only slightly to about 6% in 2030, with far less pressure.

A Goldman-cited report shows that in Ireland, where data centers account for as much as 23% of power use, data center construction has pushed up average household energy bills by about 360 euros from 2015 to 2023.

This case may serve as a future reference for the US. For mid-sized developed and emerging markets with dense data center construction, the scope for further electricity price pressure remains large.

Policy Dilemma: The Fed Can’t "Reverse" Memory Prices

The above analysis puts the Fed in a tricky position.

When core PCE inflation faces extra upward pressure from technology changes on the supply side, monetary policy tools have no direct influence over memory prices, electricity, or software subscription fees. Meanwhile, the AI capital spending boom shows no sign of cooling in the short term, meaning inflation pressure will persist for the foreseeable future.

For market participants, if Goldman’s estimates bear out and core PCE inflation contribution hits 0.5 percentage points by year-end, it will further restrict the Fed’s window for rate cuts and may force policymakers to reconsider their reliance on a "long-term deflationary" narrative for AI.

Risk Warning and DisclaimerThe market bears risks, investment needs caution. This article does not constitute personal investment advice nor takes into account the specific investment goals, financial situation, or requirements of individual users. Users should consider whether any views, opinions, or conclusions in this article fit their particular circumstances. Investment based on this, is at your own risk.