Large-scale investment in artificial intelligence could increase inflation before the technology’s productivity benefits are fully realised, according to economic research challenging the assumption that AI will automatically reduce price pressure.
The analysis, co-authored by incoming IMF chief economist Silvana Tenreyro, Bank of England economist Jenny Chan, and researcher Ludovica Ambrosino, examines the timing of investment, demand, and eventual productivity improvements.
Its central argument is that companies and households can increase spending in anticipation of future productivity gains before the additional productive capacity actually arrives.
Businesses may invest heavily in data centres, chips, software, electricity infrastructure, and organisational change because they expect AI to reduce future costs or increase output. Workers and investors anticipating higher future incomes may also increase consumption.
If demand rises faster than the economy’s ability to supply goods and services, prices can increase during the transition even where AI ultimately makes production more efficient.
Higher productivity is normally associated with an economy producing more output from the same amount of labour and capital. Over time, that can lower unit costs and reduce inflationary pressure.
The path towards that outcome can be less straightforward. AI requires unusually capital-intensive supporting infrastructure, and a rapid investment cycle can place immediate pressure on resources that are already constrained.
Computer components provide one example. Strong demand for memory, networking equipment, and graphics processors can lift hardware prices, affecting servers and other products competing for similar manufacturing capacity.
Electricity creates another constraint. Data centres require large and relatively continuous power supplies, while new grid connections, generation, and transmission infrastructure can take years to build.
Rapid AI investment can therefore increase demand for power and equipment before additional supply becomes available. Construction capacity and specialist technical labour can face similar pressure.
The inflation effect may also depend on where productivity improves. If AI makes domestically consumed services more efficient, extra supply can place downward pressure on prices. If gains are concentrated in export industries, stronger earnings and wages can increase spending elsewhere, including services where capacity remains constrained.
That distinction matters to central banks because monetary policy responds to economy-wide inflation rather than the cost structure of a single technology sector.
The Bank of England would have to distinguish temporary relative-price changes — such as more expensive hardware — from broader demand pressure capable of sustaining wage and services inflation.
Current UK conditions make the issue more immediate. Consumer-price inflation increased to 2.9% in July, while renewed volatility in energy markets has raised the prospect of another external cost shock.
If AI investment accelerates at the same time, policymakers could face stronger demand for power, construction, chips, specialised labour, and financing before the expected productivity benefit becomes visible in economy-wide output.
None of that removes the potential long-term economic case for AI. Successful adoption could materially increase productivity once technologies are integrated across businesses.
The uncertainty lies in the timing and scale of those gains. Companies can spend heavily on new technology without receiving an immediate return where processes, skills, management systems, and data do not change alongside the software.
Corporate execution therefore affects the macroeconomic outcome. Investment that improves output quickly expands supply sooner. Spending concentrated in infrastructure or experimentation can add demand for several years before delivering a comparable increase in production.
The research was published through the Bank Underground platform and represents the authors’ analysis rather than an official policy position of the Bank of England.
It separates two questions that are frequently merged in discussions about AI: whether the technology can improve productivity, and how prices behave while businesses make the investments required to achieve that improvement.
Companies will continue to assess individual projects through productivity, revenue, and competitive returns. Across the economy, the combined spending required to deliver those gains can influence inflation before the efficiency benefits are large enough to offset it.
You must be logged in to post a comment.