AI inflation is mainly a U.S. story, with limited impact in developed markets outside the U.S.

Overview

This Goldman Sachs global economic commentary discusses the inflation impact of AI investment and deployment. The report argues that AI-related demand is affecting inflation through three channels: storage price increases, price increases from adding AI capabilities to software packages, and elevated electricity prices from higher data center power demand. Because of differences in weights and measurement methodology, U.S. core PCE is more affected, while related items have relatively lower weights in inflation baskets of other developed markets, so overall impact there is materially smaller.

Core views

The key view is that AI-driven inflation pressure exists but is concentrated mainly in the U.S., especially U.S. core PCE. Goldman Sachs estimates that the three channels currently contribute more than 0.2 percentage points to U.S. core PCE year-on-year inflation, potentially rising to 0.5 percentage points by year-end. By contrast, peak core inflation support in major developed markets outside the U.S. is about 8-11 basis points, averaging around 10 basis points. Storage price shocks are global, but storage-related weight is lower in non-U.S. inflation baskets; software price increases are included in inflation, but software weights are very small outside the U.S.; and the power channel currently mainly shows electricity-price pressure in concentrated U.S. data center build-out areas.

Analysis framework

The report uses a channel decomposition approach, splitting AI inflation pressure into storage hardware, software services, and power costs, and combines inflation basket weights by economy, announced electronics price increases, software price changes, data center electricity demand, and power-forward curves to estimate marginal contributions to core inflation.

Methodology notes

• Macroeconomic inflation decompositionAI inflation three-channel framework

  Storage prices, software prices, power prices

  Decomposes AI-related price shocks into three components—hardware storage cost increases, software price increases from adding AI features, and electricity price increases from higher data center electricity demand—and estimates each part’s contribution to core inflation.

• Inflation basket weight analysisCPI/PCE weight reassessment

  Price shock multiplied by basket weight

  The same price increase yields different inflation effects across economies depending on the weight of the relevant categories in CPI or PCE; the report emphasizes that software and accessory weights in U.S. core PCE may be overstated.

• Cross-market comparisonDeveloped-market inflation contribution comparison

  U.S. versus non-U.S. developed market comparison

  By comparing category weights and price paths in markets such as the U.S., U.K., Canada, Japan, Australia, and the euro area, the report assesses whether AI inflation is globally transmissible.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• U.S. core PCE

  Primary gauge carrying AI price shocks

  Strengths

  Sensitive to storage, software, and power channels, it can reflect the impact of the U.S. AI investment boom on the price index.

  Weaknesses

  Some weights may contain measurement bias, especially software and accessory-related items that can amplify AI inflation contribution.

  Comparison

  Compared with U.S. core CPI and non-U.S. developed-market CPI, AI inflation contribution is higher in core PCE.

  Risks

  If markets misinterpret measurement effects as broad demand-driven inflation, U.S. inflation persistence could be overestimated.

• CPI in developed markets outside the U.S.

  Relatively limited spillover from AI inflation

  Strengths

  Weights of relevant hardware and software are lower, and power futures suggest limited near-term pressure.

  Weaknesses

  Still affected by global storage prices and electronics price increases.

  Comparison

  Peak impact is around 8-11 basis points, far below the roughly 50 basis points for U.S. core PCE.

  Risks

  If local data center build-out accelerates or power supply becomes constrained, smaller economies could see stronger electricity-price pressure.

• Storage chips and consumer electronics

  AI demand passes through storage costs into end-consumer electronics prices

  Strengths

  Data center demand has pushed storage prices sharply higher, and smartphone, PC, and gaming-console makers have shown pricing responses.

  Weaknesses

  Inflation impact depends on electronics-category weights in each economy, and the U.S. team believes some storage price increases are nearing their peak.

  Comparison

  Japan and Australia may have larger electronics spillover effects than some other developed markets due to higher smartphone weights.

  Risks

  If storage prices continue to rise more than expected, consumer-electronics inflation spillover could widen.

• Software services

  AI-bundled feature pricing can feed directly into the price index

  Strengths

  Most software prices are not quality-adjusted, so price increases after AI feature integration are more likely to appear in inflation.

  Weaknesses

  Software weights in non-U.S. inflation baskets are generally low, so contributions remain limited.

  Comparison

  Although UK software prices showed a notable three-month increase after Microsoft 365 price hikes, overall macro inflation weights remain small.

  Risks

  If more software vendors broadly roll out AI upsell bundles, software inflation could be temporarily higher than estimated.

• Power and data center-related energy demand

  AI compute expansion lifts data center electricity demand and can affect household power prices

  Strengths

  U.S. data center electricity share is expected to rise to about 11% by 2030, putting upward pressure on local electricity prices.

  Weaknesses

  Current impact in major developed markets outside the U.S. is limited, with two-year electricity prices broadly flat or down.

  Comparison

  The U.S. power channel currently contributes about 8 basis points, while EU data center electricity share is rising more slowly.

  Risks

  Smaller economies with high data center concentration, including Ireland and some emerging markets, may face stronger electricity-price pressure.

Key data

• Current AI inflation contribution to U.S. core PCEMore than 0.2 percentage pointsContributed jointly by storage, software, and power channels.
• Potential peak U.S. core PCE contribution by year-endAround 0.5 percentage pointsThe report says the peak impact is reduced after methodological adjustments but remains materially above other regions.
• Impact on U.S. core CPILess than 0.1 percentage pointsMainly because software and accessory weights are lower than in core PCE.
• Direct storage impact in developed markets outside the U.S.Around 5 basis points on average, range about 1-9 basis pointsDirectly exposed items have lower basket weights than in U.S. core PCE.
• Spillover impact from electronics pricesAround 5 basis pointsJapan and Australia may be higher because categories like smartphones have larger CPI weights.
• Software price impact in developed markets outside the U.S.Below 4 basis points in the U.K., about 1 basis point or lower elsewhereAssumes overall software prices rise 20%, but software weights are generally very low outside the U.S.
• Current U.S. power-channel contributionAround 8 basis pointsHigher U.S. data center electricity demand raises household electricity prices and passes through to core PCE.
• Data center power demand share outlookU.S. about 11% by 2030, currently about 6%; EU about 6% by 2030, currently about 3-4%U.S. data center electricity growth is faster, while EU increases are more moderate.
• Overall AI inflation impact in major developed markets outside the U.S.Peak about 8-11 basis points, average about 10 basis pointsMaterially below the roughly 50 basis points peak estimate for U.S. core PCE.
• Ireland energy bill impact exampleCumulative about €360 from 2015-2023The report cites an estimate that data centers accounted for 23% of Ireland’s electricity demand in 2025, which may already have pushed up household energy bills.

Impact & implications

In investment terms, AI-driven price pressure should not be treated as globally synchronized inflation acceleration. U.S. inflation data—especially core PCE—may be more disrupted due to measurement weights and concentration in data center construction, while developed markets outside the U.S. can still be affected through storage and electronics price increases but to a much smaller extent, with more limited marginal impact on inflation trajectories and interest-rate policy decisions. It is important to distinguish measurement factors in U.S. PCE from genuinely broad inflation pressure and avoid mechanically extrapolating the U.S. AI inflation experience to other economies.

Risks

• If AI-related storage prices rise sharply again, electronics price spillovers could increase.
• If software vendors more widely bundle AI features with price increases, unadjusted price indices may continue to rise.
• If U.S. data center construction exceeds expectations, electricity price pressure could become more persistent.
• Small developed economies or emerging markets with limited grid capacity and high data center usage could experience energy-bill impacts above the average in major developed markets.
• If U.S. core PCE measurement weights are misread by markets, judgments on inflation persistence and policy-rate paths may be biased.

What to watch

• Monthly changes in software, accessory, and power components in U.S. core PCE.
• Storage chip spot and contract prices, and end-device makers’ pricing decisions for smartphones, PCs, and gaming consoles.
• Whether Microsoft 365 and other software subscription services continue to raise prices due to added AI features.
• Changes in U.S. and European data center electricity share of total power demand.
• Two-year power price trends, especially in U.S. data-center-concentrated regions and high-exposure economies such as Ireland.
• Changes in statistical agencies’ methods for quality-adjusting software and hardware.