Goldman Sachs: AI's Triple Price Surge Could Push Core PCE Up 0.5 Percentage Points by Year-End
Taylor Wilson
Goldman Sachs estimates AI-driven rises in memory, software, and electricity prices have already added over 0.2 percentage points to U.S. core PCE inflation, with that contribution set to reach 0.5pp by year-end — and a rare open split inside the Fed over whether AI is inflationary or deflationary is now on the record.
What does 0.5 percentage points actually mean?
Goldman economist Megan Peters calculates that AI-driven price increases have already lifted core PCE year-on-year by more than 0.2 percentage points, rising to an estimated 0.5pp by December.
This means → AI alone could consume a significant share of the Fed's room to cut rates this year — every extra 0.5pp of core PCE makes the case for easing thinner.
Goldman notes the estimate does not yet fully capture spillover effects; the real impact could be larger.
How much does each of the three channels contribute?
Memory prices are the biggest driver: data-center demand has pushed some memory prices up more than tenfold since early 2025. Goldman expects the related price index to peak at roughly 30% year-on-year in November, contributing about 36 basis points to core PCE.
Software price hikes follow: Microsoft raised consumer Microsoft 365 subscription prices in January — the product's first increase since its 2013 subscription launch — citing the addition of AI Copilot features.
Electricity costs form the third channel: data-center power demand is already pushing up residential utility bills. Goldman estimates this channel currently adds about 8 basis points, with data-center-heavy regions well above the national average.
Why is this problem especially acute in the U.S.?
Memory inflation is global, but the weight of affected categories in the U.S. PCE basket is far higher than in other countries — peak contributions in the eurozone, the U.K., and Canada range from just 1 to 9 basis points, averaging about 5.
In plain terms = the same chip price surge hits American consumers several times harder than Europeans, because Americans spend a larger share of their budget on these goods.
Software tells a similar story: U.K. software prices jumped 20% in a single quarter — a record — yet the category's tiny basket weight means it added fewer than 4 basis points to U.K. core inflation.
How long will the electricity-price pressure last?
Goldman forecasts U.S. data-center electricity consumption will rise from roughly 6% of total national power use today to 11% by 2030 — nearly doubling.
This reflects a structural, not one-off, expansion in AI compute — electricity-price pressure will only build over time.
A cautionary precedent: in Ireland, where data centers account for 23% of power use, their build-out added roughly €360 to the average household energy bill between 2015 and 2023.
Why has an open split emerged inside the Fed?
New York Fed President John Williams stated explicitly that if AI demand "causes a sustained shock to supply and demand and pushes up inflation," monetary policy should not look the other way.
This directly clashes with Fed Chair Kevin Warsh, who in November described AI as "an important deflationary force" that would boost competitiveness through higher productivity.
This means → two top Fed officials have reached opposite conclusions on whether AI raises or lowers prices — a divergence that will directly shape the pace of rate decisions this year.
What does this mean for the rate-cut window?
Monetary-policy tools have no direct lever over memory prices, electricity rates, or software subscription fees — and the AI capex boom shows no sign of cooling.
In plain terms = the Fed's rate lever cannot stop chips from getting more expensive, but the inflation those chips generate can tie the Fed's hands.
If Goldman's estimate proves correct and AI contributes 0.5 percentage points to core PCE by year-end, the window for rate cuts this year narrows further — and policymakers may be forced to revisit their reliance on the "AI is long-run deflationary" narrative.
Content is for reference only, not financial advice.