The rapid buildout of artificial intelligence capacity is creating upward pressure on prices across multiple parts of the economy, according to analysis from CIBC Capital Markets. The bank's chief economist, Avery Shenfeld, identifies higher trucking fees, rising costs for construction materials, and memory chip price inflation feeding through to personal computers as visible channels for AI-driven price effects. Alongside those supply-side frictions, a tight labor market is adding to inflationary momentum - at least so far - which in aggregate overwhelms the limited labor-cost savings AI tools have produced to date.
Shenfeld also highlights uneven results within companies that have put AI tools into employee workflows. Some firms have seen measurable time savings, but others have found those gains offset by rising token costs associated with running models. That mixed evidence complicates the inflation outlook because realized productivity gains are not yet uniformly strong enough to counterbalance the accompanying spending on data center capacity, chips and related infrastructure.
Those cost pressures have already altered the policy conversation, CIBC notes. The analysis points to a recent shift in Federal Reserve considerations - an evolution in which Fed Chair Kevin Warsh, who had cited AI-driven productivity gains as a rationale for potential rate cuts as recently as last year, is now confronted with the possibility that higher interest rates may be needed if AI-related spending keeps pushing core inflation up.
CIBC lays out two alternative paths that would both, under the bank's assumptions, lead to cooling price pressures by late 2027 - but via different mechanisms.
In the scenario where AI delivers on its promise, one more year of implementation should allow firms to steer usage toward business functions where productivity gains outstrip token costs. That reallocation would generate labor-cost savings through reduced hiring or job displacement in roles made redundant by AI, which in turn would ease wage pressure. At the same time, growth in spending on data centers and power generation could slow, taking some of the upward pressure off chip prices. A move toward importing equipment rather than relying on domestic construction could also relax capacity constraints in the U.S. economy.
The alternative - and equally disinflationary - scenario is based on a pullback in enthusiasm. If companies pare back investment in underperforming AI tools, token expenditures would fall. More importantly, a reassessment of the investment case for AI by capital markets could shrink project budgets, reduce overall demand and relieve supply chain strains. A sharp enough equity market correction tied to disappointment in AI returns could further cool consumption and create room for the Fed to ease financial conditions.
CIBC underscores that the timing of either outcome is highly uncertain. Financial markets could continue to bankroll an acceleration of AI projects for another year or two if return-on-investment expectations remain elevated, sustaining hyperscaler capacity expansion and ongoing supply-chain stress. That persistence would leave policy-makers with limited room for inaction if core inflation remains too hot in the meantime.
The bank's baseline forecast is for no Fed rate hikes this year, but CIBC emphasizes that view is conditional on upcoming growth and inflation data. It also allows that any AI-driven hikes could later be reversed if the disinflationary forces described by either path materialize.