Artificial intelligence is often presented as a technology that could make companies more productive, reduce costs and eventually help bring prices down. But before those benefits arrive at scale, the enormous investment needed to build the AI economy may be doing something very different: adding new inflationary pressures.
That tension is making AI inflation an increasingly important issue for central banks.
Billions of dollars are flowing into data centres, advanced computer chips, electricity networks and other AI infrastructure. This spending is supporting economic growth and creating demand for construction, equipment, energy and skilled workers at a time when policymakers are still trying to keep inflation under control.
At the same time, AI could eventually allow businesses to produce more with fewer resources, which would be disinflationary.
The problem for central bankers is timing.
AI may be pushing demand and prices higher today while promising greater productivity and lower costs tomorrow. Deciding which effect will dominate — and when — makes setting interest rates much more complicated.
The Bank for International Settlements has warned that the AI boom is making it harder for central banks to read traditional economic signals because the technology is affecting both the demand and supply sides of the economy.
Why AI Inflation Is Becoming a Central Bank Problem
Central banks normally respond to inflation by trying to determine whether an economy has too much demand relative to its ability to supply goods and services.
If consumers and companies are spending too aggressively, higher interest rates can cool demand.
AI complicates that calculation.
Technology companies are investing extraordinary amounts of money in the infrastructure required to train and operate increasingly powerful AI systems.
The BIS estimates that the world’s five largest hyperscale technology companies are set to spend more than $1 trillion on AI-related capital expenditure across 2025 and 2026.
That investment supports construction companies, semiconductor manufacturers, electrical-equipment producers, utilities and numerous businesses along global supply chains.
In the short term, all that spending acts as additional demand in the economy.
If the supply of electricity, chips, engineers or grid equipment cannot expand at the same speed, prices can rise.
That is where the AI inflation problem begins.
The AI Investment Boom Is Stimulating Demand
Building artificial intelligence infrastructure requires far more than buying computers.
Companies need enormous data centres filled with specialised processors. Those facilities require land, cooling equipment, electrical substations, transmission lines, backup systems and reliable power supplies.
Construction workers must build the facilities. Engineers must design them. Manufacturers must produce the servers and electrical equipment.
The result is a major investment cycle.
According to the BIS, AI-related investment helped support global growth during 2025 despite higher tariffs and broader economic uncertainty. Optimism around AI also boosted equity markets and financial conditions.
Normally, strong investment is good news.
But when an economy is already operating close to capacity, a sudden investment boom can create additional price pressure.
Companies compete for workers, equipment, land and electricity. Suppliers facing unusually strong demand may raise prices.
Central banks then have to decide whether these pressures justify keeping interest rates higher.
AI’s Huge Electricity Appetite Adds Another Inflation Risk
Energy may be one of the most immediate channels through which AI affects inflation.
Data centres require enormous amounts of electricity to operate servers and cooling systems around the clock.
US electricity demand is now expected to reach record levels in both 2026 and 2027, with the Energy Information Administration identifying rapid expansion of AI and other data centres as one of the major drivers. Electricity consumption is projected to increase from 4,195 billion kilowatt-hours in 2025 to 4,268 billion kWh in 2026 and 4,391 billion kWh in 2027.
When electricity supply expands quickly enough, growing demand does not necessarily produce dramatic price increases.
But grids cannot always build new generation and transmission infrastructure as quickly as data centres are being constructed.
The BIS says AI development is already encountering bottlenecks involving electricity, advanced semiconductors and grid equipment, with rapidly increasing demand for computing capacity putting pressure on electricity prices and other input costs.
That pressure can spread beyond technology companies.
If households, factories and shops pay more for electricity, businesses may eventually pass some of those costs to consumers.
Data Centres Are Putting Local Power Markets Under Pressure
Virginia offers a particularly striking example.
The US state hosts one of the world’s largest concentrations of data centres. Rapidly increasing electricity demand has forced utility Dominion Energy to rely more heavily on wholesale power markets.
Dominion’s fuel expenses increased from $2.31 billion in 2021 and are projected to reach about $4.35 billion by mid-2027. Residential customers could also face higher electricity bills as the system absorbs growing demand.
That does not mean AI alone determines electricity prices. Grid investment, regulation, power-generation capacity, fuel costs and market design also matter.
But the example demonstrates how concentrated data-centre development can create pressure on infrastructure faster than utilities can comfortably expand supply.
This matters to central banks because electricity feeds into almost every part of an economy.
Higher power prices can raise the cost of manufacturing, transport, retail, food production and services.
AI Could Also Push Natural Gas Prices Higher
Another concern is what happens when additional electricity demand is met using fossil fuels.
The European Central Bank has examined a scenario in which electricity required by AI-driven data centres is supplied entirely by natural gas.
Under that extreme scenario, gas prices could rise by around 9% in Asia and Europe and 7% in the United States by 2026, with AI-related data-centre demand responsible for roughly two percentage points of the increase.
That is a scenario rather than a forecast of what must happen.
Nevertheless, it illustrates how AI infrastructure can interact with commodity markets.
If electricity systems cannot add enough renewable, nuclear or other generation, additional data-centre demand may increase competition for gas and other fuels.
Energy prices are particularly sensitive for central banks because increases can quickly influence headline inflation and potentially spread into broader prices.
Skilled AI Workers Could Add Wage Pressure
Electricity is not the only scarce resource.
The AI boom has created fierce competition for specialised workers, including machine-learning engineers, semiconductor specialists, data scientists and infrastructure experts.
Where demand for workers grows faster than the number of qualified professionals, companies may offer higher salaries to attract them.
Higher wages are not automatically inflationary. Workers becoming more productive can earn more without requiring companies to raise prices.
The difficulty arises when compensation grows faster than productivity.
A company paying much more for scarce expertise may attempt to recover those costs through higher prices.
These pressures may be concentrated in technology rather than spread evenly throughout the labour market, making them harder for central banks to interpret.
AI Inflation Is Not Just About Technology Prices
One common misunderstanding is that AI-related inflation would simply mean more expensive technology.
The effects can reach much further.
Imagine a new data-centre complex competing with manufacturers for electricity.
If limited grid capacity pushes electricity prices higher, the manufacturer may face higher operating expenses.
A construction company may also struggle to find enough electricians because data-centre developers are offering attractive salaries.
Demand for transformers, generators, cooling equipment and other infrastructure may increase at the same time.
Those higher costs can eventually appear in prices well outside the technology industry.
This is why central banks are increasingly interested in AI as a macroeconomic force rather than simply another technology-sector development.
But AI Could Eventually Reduce Inflation
There is another side to the story.
AI could become strongly disinflationary if it delivers the productivity improvements its supporters expect.
Productivity essentially measures how much output an economy can produce using available workers and resources.
If AI allows an employee to complete twice as much work in the same amount of time, a company can expand output without doubling its workforce.
Businesses might also use AI to optimise logistics, automate administrative tasks, improve manufacturing efficiency and reduce waste.
Over time, those improvements could lower the cost of producing goods and services.
The BIS has already found signs that US industries with greater exposure to AI have experienced stronger productivity gains, although those gains have also been associated with slower employment growth in some cases.
If AI significantly increases the economy’s productive capacity, it could allow stronger growth without creating the same inflation pressures normally associated with rapid expansion.
Productivity Is the Great Unknown
This is where central bankers face perhaps their biggest challenge.
Nobody knows exactly how large AI productivity gains will become or how quickly they will spread across the wider economy.
Companies are spending enormous amounts today based partly on expectations of future efficiency.
But installing AI software does not automatically make an organisation more productive.
Businesses may need to redesign workflows, retrain employees, improve databases and integrate AI with existing systems before meaningful gains appear.
Some investments may work exceptionally well. Others may produce disappointing returns.
The ECB has noted that the economic impact of AI will depend heavily on how widely and effectively companies adopt the technology.
That uncertainty creates a difficult policy problem.
Central banks can observe today’s investment boom, but tomorrow’s productivity gains are much harder to measure.
Central Banks Could Misread Strong Economic Growth
Suppose an economy grows faster because companies are spending billions constructing AI infrastructure.
A central bank might interpret stronger growth as evidence that demand is becoming excessive and therefore keep interest rates high.
But suppose AI then produces a major productivity increase.
The economy could potentially grow faster without creating sustained inflation because its productive capacity has increased.
Keeping interest rates too high in that environment could unnecessarily slow investment and employment.
The opposite mistake is also possible.
Central banks might assume future productivity gains will prevent inflation and lower interest rates too quickly.
If those gains take much longer to arrive while AI investment continues boosting demand, inflation could remain stubbornly high.
The BIS has highlighted exactly this difficulty: AI simultaneously influences supply and demand, making it harder to determine how much economic growth represents overheating and how much reflects sustainable increases in productive capacity.
AI Could Change the Neutral Interest Rate
Artificial intelligence may even affect something central bankers call the neutral interest rate.
This is the theoretical interest rate that neither stimulates nor slows the economy when inflation is stable.
If businesses expect AI investments to generate substantial future profits, they may continue borrowing and investing even when interest rates are relatively high.
Federal Reserve Governor Lisa Cook noted in March 2026 that soaring investment in data centres and chips was already occurring despite interest rates remaining elevated compared with much of the previous two decades. She suggested that strong AI-related investment could mean the neutral interest rate is higher than it was before the pandemic.
That would have important implications.
Interest rates that once looked restrictive might no longer slow economic activity as much as expected.
Central banks could therefore need to reconsider how they judge whether monetary policy is truly tight.
Rising AI Stocks Can Boost Consumer Spending
The AI boom can influence inflation through financial markets as well.
Technology companies associated with artificial intelligence have helped drive major stock-market gains.
When households see the value of their investments rise, some feel wealthier and may increase spending.
Economists call this the wealth effect.
The BIS says optimism surrounding artificial intelligence helped sustain favourable global financial conditions during the recent investment boom.
That can reinforce demand even further.
AI investment boosts economic activity directly, while rising stock valuations may simultaneously encourage consumption among investors.
For central banks attempting to cool an economy, those financial effects can partly work against higher interest rates.
What Happens If the AI Boom Suddenly Reverses?
Central banks also have to consider the opposite risk.
AI investment expectations are extremely high.
If companies eventually discover that returns do not justify the enormous amount of capital being spent, investment could fall sharply.
The BIS has warned that intense competition may encourage technology companies to overcommit resources to projects whose eventual returns remain uncertain. A disappointment could turn today’s capital expenditure boom into an investment downturn and tighten financial conditions.
That would create an entirely different problem.
Instead of worrying that AI is generating too much demand, policymakers might suddenly face weaker investment, falling asset prices and slowing economic growth.
Interest-rate decisions therefore have to account not only for today’s AI boom but also for the possibility that financial enthusiasm eventually cools.
AI Could Create Different Inflation Experiences Around the World
The impact will not be identical everywhere.
Countries that manufacture advanced semiconductors may benefit from increased exports.
Economies hosting large concentrations of data centres may experience unusually strong electricity and construction demand.
Countries that adopt AI rapidly could enjoy greater productivity gains.
Others may primarily experience higher global prices for energy, chips and infrastructure without receiving equivalent productivity benefits.
The IMF has warned that in parts of Sub-Saharan Africa, for example, additional AI-related electricity demand could raise power prices when supply cannot expand sufficiently. Where electricity tariffs are regulated or subsidised, the pressure may instead appear in government finances.
This means AI could widen differences in inflation and growth between economies, making coordinated global monetary policy even more complicated.
Central Banks Face a Technology Paradox
The AI inflation debate ultimately comes down to a paradox.
Artificial intelligence could become one of the world’s most powerful tools for reducing costs and improving productivity.
But building the infrastructure needed to achieve those benefits requires an extraordinary amount of spending first.
Data centres must be constructed. Power generation must expand. Electricity grids must be upgraded. Semiconductor factories need investment. Skilled employees must be recruited.
Those activities create demand today.
The productivity payoff may come later.
This time gap is what makes AI unusual from a monetary-policy perspective.
Central bankers are accustomed to dealing with shocks that clearly reduce supply or increase demand. Artificial intelligence can do both — while also potentially increasing supply dramatically in the future.
The Inflation Fight Just Became More Complicated
Global inflation is still expected to ease over time, but the AI investment boom has introduced another variable into an already complicated economic environment.
For central banks, the key question is no longer simply whether artificial intelligence will boost productivity.
It is whether those productivity gains will arrive quickly enough to offset the demand, energy and infrastructure pressures being created by the AI build-out.
If productivity accelerates strongly, AI could eventually allow economies to grow faster while keeping inflation under control.
If infrastructure bottlenecks persist and productivity gains disappoint, the technology boom could instead keep pressure on electricity, wages and other costs for longer than policymakers expect.
The most likely outcome may contain elements of both.
AI inflation could be strongest during the construction and investment phase before gradually giving way to productivity-driven disinflation as the technology spreads through businesses.
That leaves central banks navigating an uncomfortable transition: they must set today’s interest rates based partly on a technological transformation whose ultimate economic impact is still unfolding.
Artificial intelligence may eventually make economies more efficient.
Before it does, however, the race to build the AI economy is making the inflation fight considerably harder to read.
