Artificial intelligence is often presented as a powerful deflationary force. By automating tasks, accelerating production and helping companies operate more efficiently, AI could eventually reduce costs across many industries.

The economic transition, however, may be far less straightforward. In the short and medium term, the rapid expansion of AI infrastructure could increase demand for scarce resources, redirect investment and create new inflationary pressures.

AI investment is reshaping demand

Petra Tschudin, a member of the Swiss National Bank’s governing board, has warned that artificial intelligence could affect prices in both directions.

The global race to build AI capabilities requires substantial investment in data centres, semiconductors, energy infrastructure, cloud capacity and specialised talent. As capital and resources move towards these areas, other sectors may face higher costs or more limited access to essential components.

Semiconductor shortages are one potential example. Growing demand for advanced chips could push prices higher, with the effects spreading across industries that depend on the same supply chains.

AI adoption may therefore contribute to inflation before its promised efficiency gains become visible across the wider economy.

Higher productivity does not automatically mean lower inflation

Over time, AI could help companies produce more with fewer resources. Better productivity could reduce operating costs and make certain products and services more affordable.

But lower production costs do not necessarily create lasting deflation.

Inflation measures how prices change from one year to the next. For AI to generate a sustained deflationary effect, reductions would need to continue repeatedly rather than represent a one-time adjustment.

Productivity improvements have supported economic growth throughout history, but they have not generally pushed entire economies into structural deflation. AI may follow a similar pattern: generating considerable value without continuously reducing the overall price level.

Research from Bank of England staff has raised a related concern. Even if artificial intelligence produces significant productivity gains, the technology may not automatically lower inflation. The final outcome will depend on how quickly companies adopt it, where the benefits appear and whether savings are passed on to customers.

Central banks are watching an unfamiliar transition

The Swiss National Bank currently expects annual inflation to remain within its target range of 0% to 2% through the first quarter of 2029.

That forecast does not mean interest rates will necessarily remain unchanged. The SNB’s projections are conditional, showing how inflation may evolve if the policy rate stays at its current level. If new information changes the inflation outlook, the bank can adjust monetary policy accordingly.

This distinction matters because AI introduces several variables that are difficult to model. Its economic effects will depend on the speed of investment, infrastructure constraints, labour-market changes and the distribution of productivity gains between companies and consumers.

What this means for technology leaders

For CTOs and business leaders, the discussion is a reminder that AI adoption cannot be evaluated only through the cost of a model or the number of tasks it can automate.

The broader cost structure also matters:

access to computing capacity and advanced hardware
cloud and energy consumption
integration with existing systems
competition for specialised talent
governance, security and compliance requirements
the time needed before productivity gains become measurable

An AI initiative may promise long-term efficiency while increasing costs during implementation. Organisations need realistic business cases that account for this transition instead of assuming that automation will immediately produce savings.

The real impact will depend on execution

Artificial intelligence could eventually improve productivity and make some goods and services cheaper. Before that happens, the investment required to build and deploy the technology may place additional pressure on prices.

The result is unlikely to be uniformly inflationary or deflationary. Different industries, regions and companies will experience the transition at different speeds.

For businesses, the priority should be to identify where AI solves a genuine operational problem, measure its total cost and introduce it within systems that can be governed and improved over time. Productivity gains are possible, but they depend on disciplined implementation rather than technology alone.

Source