Key Takeaways:
Goldman Sachs estimates AI-driven demand for memory chips, software and electricity will add half a percentage point to US core inflation by year-end, challenging the Fed's rate-cut timeline.
The AI boom is reshaping US inflation from an unexpected direction. Goldman Sachs economist Megan Peters estimates surging memory-chip prices, software subscription hikes and rising electricity costs have already lifted core PCE by 0.2 percentage point, with the contribution set to reach 0.5 point by December.
"The magnitude of the AI-related price impulse is becoming too large for inflation forecasters to ignore," said Megan Peters, economist at Goldman Sachs. "These are supply-side cost shocks that monetary policy cannot easily reverse."
Memory prices have surged more than tenfold since early 2025 as data-center demand overwhelms supply. Goldman expects the software and accessories price index — closely tied to memory costs — to peak at roughly 30 percent year-over-year in November, contributing about 36 basis points to core PCE. Apple Inc., Microsoft Corp. and Dell Technologies Inc. have announced price increases of as much as 25 percent on consumer electronics. Separately, data-center electricity demand has added roughly 8 basis points through the utility channel, with effects far larger in PJM-grid states where server farms are concentrated.
The findings put the Federal Reserve in a bind. If AI-related inflation reaches 0.5 percentage point by year-end, it would further compress the already narrowing window for rate cuts in 2026. The latest FOMC minutes, released in July, flagged for the first time that "most participants" saw a scenario where "strong AI-related demand" keeps inflation persistently elevated — a risk that could force the central bank to hold rates higher for longer.
The debate has split Fed leadership. New York Fed President John Williams said in a recent speech that if AI demand "causes persistent shocks to supply-demand dynamics and pushes up inflation, I believe this should not be overlooked." His comments contrast directly with Fed Chairman Kevin Warsh, who wrote in November that AI represents an "important deflationary force" that would boost productivity and enhance US competitiveness.
The divergence matters for markets. Overnight index swaps currently price roughly two quarter-point cuts by December 2026, according to Bloomberg data. A 0.5-point inflation contribution from AI alone would consume the entire projected easing — and then some — unless other inflation components cool enough to offset it.
Data centers strain power grids
The electricity channel carries the longest tail risk. Goldman projects US data-center electricity consumption will nearly double from about 6 percent of national demand today to 11 percent by 2030. In the European Union, the share is expected to rise more modestly from 3-to-4 percent to about 6 percent. Ireland offers a cautionary precedent: data centers already consume 23 percent of the country's electricity, and their construction added roughly 360 euros to average household energy bills between 2015 and 2023, according to a report cited by Goldman.
For non-US developed markets, the direct memory-price channel is smaller — peak contributions range from 1 to 9 basis points, averaging about 5 basis points — because memory-related categories carry lower weight in their inflation baskets. But the electricity and software channels are global, meaning the AI inflation phenomenon is not confined to America.
Policy dilemma without a monetary tool
The Fed's predicament is that its primary tool — the federal funds rate, currently at 4.25-to-4.50 percent after being held steady since the 25-basis-point cut in September — has limited ability to address supply-driven price increases in memory chips, cloud software or electricity tariffs. Unlike demand-pull inflation, which rate hikes can cool by slowing consumption, these AI-related cost pressures stem from structural capacity constraints in semiconductor fabrication, power generation and data-center construction.
The capital expenditure cycle shows no signs of easing. Hyperscalers including Microsoft, Amazon.com Inc. and Alphabet Inc. have committed tens of billions of dollars to AI infrastructure through 2027, ensuring demand for memory chips and electricity will remain elevated. That suggests the inflation contribution Goldman has identified is not a one-off blip but a multiyear structural shift.
For investors, the implication is clear: the "AI deflation" narrative that has supported rate-cut expectations may need rethinking. If Goldman's forecast proves accurate, the Fed's ability to ease in 2026 will depend not on whether inflation falls, but on whether the non-AI components of the economy cool enough to offset the machine-driven price pressures building beneath the surface.
This article is for informational purposes only and does not constitute investment advice.
