AI Predicts Higher Inflation, Challenging 'Disinflationary' Consensus

Got story updates? Submit your updates here. ›

A new research note from Deutsche Bank reveals that leading AI models predict a higher probability of inflationary pressure than price declines in the near term, challenging the prevailing 'AI disinflation' consensus. The bank's economists found that AI systems like dbLumina, ChatGPT, and Claude cited massive capital expenditure on data centers and energy constraints as immediate demand-pull inflation drivers, rather than the expected productivity gains and cost savings. This divergence signals a critical mispricing in bond markets, where the yield curve fails to account for persistent structural inflation driven by the AI build-out.

Why it matters

The market has spent two years pricing in a secular decline in inflation, convinced that artificial intelligence will act as a deflationary shock to the global economy. But Deutsche Bank's internal AI tools are pushing back on this thesis, predicting that the capital-intensive AI build-out will actually drive up costs and inflation in the short term before any productivity gains materialize. This has major implications for bond yields, corporate margins, and the broader economic outlook.

The details

Deutsche Bank's economists queried three distinct AI systems - their own proprietary dbLumina, OpenAI's ChatGPT-5.2, and Anthropic's Claude Opus 4.6 - asking for probability assignments on four inflation outcomes over one-year and five-year horizons. The results showed a stark deviation from the street consensus, with every model rating the probability of AI raising inflation higher than the probability of it meaningfully reducing prices. The culprit identified by the models is the AI investment boom itself, as data centers multiply, semiconductor demand surges, and electricity consumption from AI workloads strains grid capacity - all of which creates demand-pull inflationary pressure rather than the expected deflationary effects.

• In March 2026, Anthropic released a study highlighting a critical adoption gap, where AI tools are theoretically capable of automating 94% of computer and math work and 90% of office administrative roles, but actual deployment is a fraction of that potential.
• In February 2028, Citrini Research rattled markets with a scenario of a 'white-collar recession', arguing that AI won't just ease prices - it will destroy the consumer base that sustains them.

What they’re saying

What’s next

As the fiscal year progresses, the divergence between AI hype and AI reality will widen. Smart capital will move away from speculative growth and toward defensive positioning, securing partners who can navigate the operational complexities of this new inflationary regime.