Central banks are converging on a two-phase theory: AI spending drives prices up now, but productivity gains should push them down later
Research from the Federal Reserve, the Bank of Korea, the European Central Bank, and the London School of Economics is painting a consistent picture: AI will be inflationary first, then disinflationary.
The short-term squeeze #
A St. Louis Fed report from March 2026 laid out the mechanism clearly. AI optimism is fueling massive capital expenditure in data centers, semiconductors, and energy infrastructure. That spending surge pushes prices higher in the near term, before any of the promised productivity gains actually materialize.
AI-driven data center buildout is expected to cost over $700 billion, acting as a significant inflation driver across memory chips, processors, and electricity markets.
The Bank of Korea’s analysis adds another dimension. Governor Rhee Chang-yong has emphasized South Korea’s position as a leader in AI chip production, tying semiconductor demand directly to inflation monitoring and interest rate decisions. The Bank of Korea projects that AI-related investments could account for 39% of US growth in the second half of 2026.
The disinflation thesis #
The LSE published an analysis on October 8, 2025, concluding that unanticipated advances in AI create initial inflationary disruptions, which then transition to disinflationary effects as productivity improvements ripple through the economy.
The ECB’s October 2024 modeling tells a similar story with a slight twist. When AI-driven productivity gains are anticipated by markets, inflation actually rises first because demand adjusts faster than supply. But when AI adoption is unanticipated, the initial effect is disinflationary, eventually normalizing as supply and demand find equilibrium.
The St. Louis Fed’s framework reinforces this. Realized productivity growth eventually increases potential output, meaning the economy can produce more without overheating, and more supply relative to demand means prices moderate.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our