AI's first act is to push up U.S. core PCE inflation

Overview

This report discusses the short-term impact of the AI investment boom on U.S. inflation. Its core view is that while AI may boost productivity and create disinflationary forces over the long term, its first phase is more likely to be inflationary: data center construction is pushing up prices for electricity, memory chips, computer software and accessories, and hardware, putting recent pressure on core PCE.

Core views

The report argues that AI-related inflation is not merely a supply shock, but a demand shock created by nearly $1 trillion of AI investment chasing constrained resources. Memory prices should remain supported in the near term, with gains potentially shifting from rapid spikes to a more moderate and differentiated trend. Although BEA revisions and quality adjustments may reduce reported inflation, as long as the rest of core PCE remains elevated, the Fed will find it difficult to fully look through AI-related inflation.

Analysis framework

The report focuses on narrowly defined AI-related categories that can directly enter the CPI/PCE basket, mainly including computer software and accessories, computers and peripherals, smart home assistant devices, and some game consoles; it also combines PPI, import prices, Korean producer prices, memory chip supply and demand, hardware vendor price hikes, and PCE/CPI weight differences to construct base-case and high-inflation scenarios and assess the implications for the Fed, inflation markets, and investment-grade tech credit.

Methodology notes

• Inflation component attributionPCE/CPI weight and contribution decomposition

  Compare the weight differences of AI-related categories in CPI and PCE, and estimate their contributions to headline PCE and core PCE.

  "Computer software and accessories" has a very low weight in CPI but a significantly higher weight in PCE, so the same price shock produces a more visible effect in PCE.

• Scenario analysisBase case and high-inflation case

  Use different assumptions for monthly price paths to assess how AI-related categories may support future core PCE.

  The base case assumes that price increases for software and accessories slow to half the recent pace in the second half of 2026 and to one-quarter in 2027; the high case assumes monthly increases of 5–6% continue this year and only fall to half that pace in 2027, while inflation in computers and peripherals accelerates.

• Price measurementQuality adjustment and BEA composite deflator

  Assess whether the current CPI deflator overstates inflation in software and accessories because of a lack of quality adjustment.

  The report acknowledges that the software category may require quality adjustment, but argues that higher memory-related accessory prices more likely reflect a real AI-demand squeeze on inventories; the BEA annual revision may reduce part of the reported inflation.

• Industry supply and demandMemory cycle and hardware cost pass-through

  Analyze how DRAM, HBM, and NAND supply constraints and AI data center demand affect hardware prices.

  Major memory makers are prioritizing high-margin HBM and server DRAM, limiting supply of standard DRAM and consumer NAND, which could push costs through to PCs, peripherals, and game consoles.

Asset mapping & comparison

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

• Core PCE

  AI-related categories are an important recent contributor to higher core PCE.

  Strengths

  PCE better captures spending weights for software, accessories, and goods-flow methodology, making it more sensitive to this shock.

  Weaknesses

  There are measurement and quality-adjustment disputes around the related categories, and the BEA annual revision may revise historical readings lower.

  Comparison

  Compared with CPI, Computer software and accessories carries a much higher weight in PCE.

  Risks

  If quality adjustment or the new deflator significantly lowers inflation in this category, the current contribution may be reversed.

• CPI inflation

  AI-related software and accessory categories carry extremely low weights in CPI, so the direct impact is limited.

  Strengths

  The category's weight in CPI is only about 0.03%, causing little disturbance to the overall index.

  Weaknesses

  CPI may not fully reflect the amplified effect of this category in PCE.

  Comparison

  The report emphasizes that this is more a PCE issue than a CPI market issue.

  Risks

  If price hikes spread to hardware, PCs, and game consoles, the CPI impact could widen.

• U.S. rate path

  Whether the Fed incorporates AI-related PCE upside will affect interest-rate market pricing.

  Strengths

  If inflation remains persistently elevated, the policy response may become more hawkish.

  Weaknesses

  If the Fed views the shock as concentrated and affected by measurement issues, it may choose to partially look through it.

  Comparison

  Compared with an oil-price shock, the report sees AI inflation as more demand-driven rather than a pure supply disturbance.

  Risks

  Stronger-than-expected inflation could raise the probability of rate hikes; lower revised inflation could depress hawkish pricing.

• U.S. inflation-linked market

  The first-order impact is limited, while the second-order impact depends on how the Fed interprets the PCE data.

  Strengths

  PCE pressure may alter policy expectations and indirectly affect inflation-risk pricing.

  Weaknesses

  AI-related categories carry small weights in CPI-linked markets, so direct pass-through is weak.

  Comparison

  If the Fed fully incorporates the upside pressure, it could imply a higher rate path and lower inflation risk; if it looks through it, the opposite would follow.

  Risks

  Uncertainty around the policy reaction function is the main trading risk.

• Memory chips and investment-grade tech credit

  AI data center demand and supply discipline support DRAM, HBM, and NAND prices.

  Strengths

  Major manufacturers such as Micron, SK Hynix, and Samsung are being relatively restrained in adding capacity, with higher priority given to high-margin advanced memory.

  Weaknesses

  The memory industry itself is cyclical, and additional supply over the next 1–2 years may gradually ease tightness.

  Comparison

  High-performance memory may continue to outperform more standardized memory products.

  Risks

  Capacity recovery, slower demand, or inventory rebuilding could reduce the slope of price increases.

• PCs, peripherals, and game consoles

  Higher memory costs may pass through to end-hardware prices.

  Strengths

  Price hike cases by Apple and Microsoft show early signs that cost pass-through has already begun.

  Weaknesses

  The category weights are small, and a price increase by a single brand may not represent the national sample.

  Comparison

  Computers and peripherals matter more than video games and consoles because the latter have a PCE weight of about 0.05%.

  Risks

  If more vendors follow with price hikes, CPI and PCE pressure in 2026–2027 may exceed the base case.

Key data

• Contribution of AI-related categories to PCE inflationAbout 25bp headline PCE, about 20bp core PCEThe report estimates this as the recent cumulative contribution to annual inflation, representing a first-order effect.
• Increase in Computer software and accessoriesUp 17% cumulatively since December 2025, with monthly gains of about 4–5%This category had long been disinflationary before recently becoming an important source of the PCE-CPI inflation gap.
• Weight of Computer software and accessoriesAbout 0.03% in CPI, about 1.1% in PCEThe weight difference makes this category much more influential for PCE than for CPI.
• Contribution of this category to monthly core PCEAn average of about 5–6bp per month since December 2025The effect is more pronounced for core goods PCE.
• Computers, peripherals and smart home assistant devicesAbout 0.3% CPI weight, about 0.5% PCE weight; up about 6% cumulatively since December 2025This category includes PCs, laptops, tablets, and peripherals, but excludes flash drives.
• Extreme Apple price-hike scenarioIf a 20% price increase applied to the entire computer category, it could lift core PCE by about 10bp as a one-offThe report believes the actual near-term impact may be limited unless more vendors follow.
• PCE weight of video games and consolesAbout 0.05%The weight is very small, so the impact on PCE is limited unless prices rise sharply.
• BEA annual revisionTo be implemented on September 30, 2026, covering revisions from Q1 2021 to Q1 2026The new composite deflator may lower monthly core PCE in Q1 2026 by an average of about 2–3bp.
• Fed path viewRemain on hold through the end of 2027; rate hike risk around 40%If future inflation is stronger than expected, the risk is tilted toward hikes.

Impact & implications

For markets, the first-order CPI impact of AI inflation is limited, but the PCE impact is more visible. If the Fed fully incorporates AI-related PCE upside into its reaction function, markets may price a higher rate path and a lower inflation risk premium; if the Fed places more emphasis on trimmed-mean measures or quality adjustments and chooses to look through the shock, the direction would be the opposite. For tech credit and the hardware supply chain, memory prices staying higher for longer would benefit memory manufacturers' revenues and profits, but would also increase cost pressure for PCs, peripherals, game consoles, and data center customers.

Risks

• The BEA annual revision on September 30, 2026 may lower inflation in Computer software and accessories through a new composite deflator.
• The software and accessories category lacks quality adjustment, so current inflation readings may be overstated.
• If memory supply improves, capacity utilization rises, or new capacity comes online faster, upside in memory prices may slow.
• If core PCE continues to cool after excluding AI-related categories, the Fed may be more inclined to look through concentrated price pressure.
• If more hardware vendors follow Apple and Microsoft in raising prices, AI cost pass-through may broaden to a wider range of consumer hardware.
• If inflation comes in stronger than expected over the coming months, the risk of Fed rate hikes may rise.

What to watch

• The BEA annual revision on September 30, 2026 and the new deflator weighting for Computer software and accessories.
• Monthly trends in Computer software and accessories, computers and peripherals, and core goods PCE in PCE and CPI.
• DRAM, HBM, NAND, and enterprise SSD prices, as well as U.S. PPI, import prices, and Korea semiconductor PPI.
• Whether Apple, Microsoft, and other PC, laptop, and game console vendors continue to raise prices.
• Changes in Fed speeches, SEP, and language around trimmed measures and AI-related inflation.
• Hyperscaler capex, memory procurement agreements, new wafer capacity, and the pace of cleanroom ramp-up.