Is AI inflation transitory?

CIBC warns that the AI spending boom is showing up as inflationary pressure across supply chains (e.g., trucking fees, construction materials, memory chip costs) and a tight labor market, outweighing early labor-cost savings. The Fed may need to consider rate hikes if core inflation stays hot, even as an AI-driven disinflation/cooling toward late-2027 remains possible via either labor savings (optimists) or reduced token spend and AI budget pullbacks (sceptics). CIBC’s base case is no Fed rate hikes this year, but the timing and outcome are highly uncertain and depend on incoming growth/inflation data.

Analysis

The cleanest market read is not “AI is inflationary” but “policy volatility rises before productivity shows up.” That is bearish duration and high-multiple growth over the next 1-3 months if sticky service inflation, power demand, and capex-related input costs keep Fed cut odds moving out. In that window, the market tends to punish anything financed on distant cash flows more than it rewards eventual efficiency gains.

Second-order, the AI spend cycle is becoming a capital-allocation filter: suppliers with scarce capacity and pricing power should hold margins, while semicap tools, data-center REITs, and levered tech buyers will feel the squeeze if ROI scrutiny rises. If the AI skeptics are right, the reversal can be abrupt because project budgets are discretionary and the supply-chain inflation they created unwinds faster than labor costs do. KEP is only a partial beneficiary here: power demand growth helps load factors, but regulated pass-through means the equity still behaves like a rate-sensitive utility, not a pure AI winner.

Contrarianly, the consensus may be underestimating how long the inflation impulse can persist even if AI productivity eventually arrives. The real falsifier is a visible slowdown in data-center orders, memory pricing, or hiring tied to AI infrastructure; absent that, the base case is higher-for-longer policy noise, not an immediate disinflation trade. CM is a tactical hedge only if rates stay elevated without a credit turn; if loan-loss provisions rise, any NIM benefit gets erased quickly.

Key Decisions for Investors

• Short TLT on rallies over the next 1-3 months; target a 25-40 bps backup in the 10Y if CPI/core services stay sticky. Cover if the 10Y breaks below ~4.0% or two consecutive inflation prints soften.
• Fade high-multiple AI beneficiaries with a basket short in XLK/QQQ versus long XLF on any dovish rate rally; the setup is a multiple reset, not a fundamental collapse. Stop if capex guidance re-accelerates without margin compression.
• No outright trade in CM yet; use it as a watch item for higher-for-longer. Consider a small tactical long only if rate expectations rise while North American credit stays benign; abandon if provisions or mortgage stress start moving up.
• Keep KEP on the radar as a cheap AI-power proxy, but only buy on confirmation that tariff pass-through or regulated returns improve. If Korean utilities remain boxed in, the valuation may stay cheap for the wrong reason.