What Key Concerns Drive Global Central Banks' Collective Assessment of AI?

A technological revolution is changing how central banks study the economy.

Artificial intelligence (AI) is developing quickly, from the money invested in it to its use in industry and the financial system. These changes are happening at different speeds. Individuals, businesses, markets and entire economies do not all experience them at the same time.

On October 19, 2026, Pan Gongsheng, Governor of the People's Bank of China, spoke about AI at the Annual Meeting of the Financial Street Forum. He said that the effects of AI on demand and supply may appear at different times. In the short term, AI may help individual businesses work more efficiently, encourage investment and push inflation higher. Over the medium and long term, it may increase productivity across the economy and reduce inflation. He also noted that the AI boom has become one of the main forces driving global stock markets higher. Global financial markets are closely connected and can experience large swings. He identified three priorities for the next stage: closely study and track AI's effects on the wider economy, support the AI industry through financial services, and carefully expand the use of AI in the financial sector.

Since the start of this year, the Federal Reserve, European Central Bank, Bank of Japan, Bank of Canada, Reserve Bank of Australia and Bank of England have also studied how AI may affect the economy. Their work covers investment, prices, jobs, interest rates and financial stability.

01

The Puzzle of How to Respond With Policy

Central banks around the world are watching different things. The Bank of Japan and the Reserve Bank of Australia have noticed how demand for AI is affecting the prices of technology products. The European Central Bank is looking at the gap between growing investment, the use of new technology and higher productivity. The Bank of Canada is studying the difficult choices that may arise as the types of jobs available change. The Federal Reserve is discussing future economic growth and the workforce, while the Bank of England is including AI-related financing in its monitoring of financial stability. These studies address questions that have not yet been settled. They do not mean that the countries involved have changed their policies as a result.

One important feature is that AI is creating an economy in which different things move at different speeds. As AI technology continues to improve, businesses are already spending money on equipment and infrastructure. Financial markets are pricing in expected future profits before they arrive. Meanwhile, productivity improvements, changes in employment and adjustments to rules and institutions are happening on different schedules.

This is not just the familiar delay between the invention of a technology and its widespread use. The rapid development of new technology also makes it harder for people to understand events and make decisions. More investment does not automatically mean that the economy can produce more goods and services at the same time. Nor does a market price prove that a business will actually earn the money it expects.

This is the main challenge for central banks. They must understand what is happening in the economy now, separate short-term increases in demand from lasting improvements in the economy's ability to produce goods and services, and work out how expectations about the future are already affecting financial conditions today.

Consider how AI investment begins. Companies buy chips and servers, build data centers, and invest in power grids, cooling systems and other supporting facilities. They place orders, organize construction, arrange energy supplies and obtain financing. This spending quickly creates demand in the economy. But it takes time to find out whether the completed equipment will be used fully and how much new income it will generate.

On July 28, 2026, the Bank for International Settlements (BIS) published a report titled AI and the Global Economy: Implications for Central Banks. It said that AI is driving a large increase in investment, with businesses relying more and more on borrowing to pay for it. AI is also affecting the global economy through trade, financial markets and changes in wealth. AI could bring large productivity gains, but the results remain uncertain and are likely to differ across countries and industries.

In its July Outlook for Economic Activity and Prices, the Bank of Japan included AI-related demand in its official economic forecast. This demand supports Japan's economic growth but also pushes up the prices of goods such as semiconductors. Because of a combination of energy prices, a weaker yen and AI-related demand, the Bank of Japan expects the year-on-year rise in the consumer price index in the second half of fiscal 2026 to be well above 2%.

This does not mean that AI has become the main cause of inflation in Japan. It does show that a technology expected to lower costs over the long term can also affect actual prices while it is still being developed and adopted.

In its monetary policy statement on September 29, the Reserve Bank of Australia said that demand related to AI had helped push global technology product prices up quickly.

Investment in Europe has also reached a noticeable level. On October 13, Philip Lane, a member of the European Central Bank's Executive Board, said that companies in the euro area planned to spend about 10% of their investment budgets for 2026 on AI. AI-related loans also made up about a quarter of the recent increase in loans to businesses.

Lane had also explained that if households and businesses make spending and investment decisions based on expected future productivity gains before those gains actually arrive, AI could push inflation higher by increasing demand in the early stages of the transition.

Why might a technology that could lower production costs in the future make some economic activities more expensive while it is still developing?

In her 2002 book Technological Revolutions and Financial Capital, technology economist Carlota Perez divided major technological changes into two stages: the "installation period" and the "deployment period." In the first stage, investment drives the building of infrastructure and the growth of industries. The technology can deliver wider productivity gains only after businesses change how they are organized, how they operate and how institutions support them.

Railways, electricity and information technology all took different amounts of time to spread widely after investment began. Perez's theory offers a historical way to understand today's AI boom, but it does not prove that AI will follow the same financial cycle as earlier technologies.

Central banks need to work out whether today's growth mainly comes from building new infrastructure or whether the conditions are in place for the economy to keep increasing its ability to supply goods and services. This distinction affects how they judge inflation and how long it may last.

02

When Will Productivity Gains Arrive?

Changes in orders, investment and prices can often be seen fairly quickly. Improvements in productivity usually take much longer to appear.

Results from a European Central Bank survey shared by Lane on October 13 showed that 64% of the euro area businesses surveyed had tried AI or were using it to a moderate extent. Only 7% said their use of AI was "significant." Meanwhile, 52% of the workers surveyed said they used AI at work, compared with 26% in 2024.

People using AI have reported saving a lot of time, but the benefits are mostly found in specific tasks, such as programming. The European Central Bank has not yet found a clear, significant effect of AI on productivity across the economy as a whole.

AI tools are becoming part of more people's working lives. Why, then, have they not made a clear difference in overall economic productivity figures?

In a paper published in March this year, Bank of Canada researchers Joshua Brault, Maryam Haghighi and Jing Yang called this the "AI productivity paradox." Some studies of individual tasks have found that AI can raise productivity by 15%-60%. However, these gains have not yet fully appeared in economy-wide productivity statistics.

These figures come from different tasks and groups of people. They cannot be used directly to estimate how much more efficient the entire economy has become.

Earlier research offers another explanation. In 2021, economists Erik Brynjolfsson, Daniel Rock and Chad Syverson published a study on the "Productivity J-Curve." It explains that when businesses adopt technologies that can be used across many industries, they usually need to change work processes, train employees, reorganize their businesses and invest in intangible assets. These assets are difficult to measure fully.

These early investments do not always lead to an immediate, visible increase in output. Productivity gains may become clearer as the technology and the organizations supporting it develop and mature.

So, more people can use a technology even while overall productivity figures show little growth. But a delay in the statistics does not guarantee that productivity will suddenly rise sharply in the future.

Technological progress can also change how many workers businesses need and what kinds of work they do.

In a model published in July this year, the same Bank of Canada research team described two ways AI can affect workers: it can help them work more efficiently, or it can automate tasks that people previously did. The model finds that the effects may differ greatly between industries. If a central bank tries to protect jobs in industries affected by AI by boosting overall demand, this could also add to inflationary pressure in other industries.

This is a difficult policy choice highlighted by a theoretical model. It is not a statement by the Bank of Canada that AI has already caused unemployment.

In August, Kevin Warsh, a former Chair of the Federal Reserve, raised several unanswered questions in a speech at the Jackson Hole Economic Symposium. When will AI bring lasting productivity gains across the whole economy? Will computing power and workers complement each other, or will they compete? Who will ultimately receive the returns from investment in the technology?

These questions matter for the economy's potential growth. They also affect how central banks assess employment, wages and inflation.

Ultimately, AI's long-term economic value depends on how it is used in practice. Improvements in AI models must lead to better business processes, more efficient production, and new products and services before their benefits can show up more widely in national economic statistics.

03

Different Answers About Interest Rates

While changes in productivity and employment are still developing, financial markets continue to set prices based on what they expect to happen in the future.

The Bank of England's Financial Stability Report, published in July, said that in 2025 some AI companies needed more investment than their operating cash flow could support. Their use of outside financing grew even faster in the first half of 2026. They obtained funds through several channels, including corporate bonds, bank loans, private credit and structured financing.

Market forecasts cited in the report show that expected capital spending by large AI technology companies in 2028 rose from less than 600 billion US dollars at the end of 2025 to more than 1 trillion US dollars.

Changes in how companies raise money are also changing the financial pressures on different businesses across the AI supply chain.

Data centers and their supporting facilities can remain in use for many years, but the chips and servers inside them may become outdated much faster. Whether a project can repay long-term borrowing from its cash flow depends on how often its equipment must be replaced, its customer contracts, how much its facilities are used and the financing available in the future.

The Bank of England is therefore watching how risks might spread through the wider credit market as AI construction projects move beyond companies' own financing and rely more on outside lenders. However, its report also clearly says that there is not enough evidence at present to show that AI financing has significantly reduced other businesses' or governments' access to capital.

In October, the International Monetary Fund's Global Financial Stability Report issued a further warning. If expected profits from AI investment fail to materialize, falling asset prices could make market swings worse because of borrowing, large concentrated investments and links between borrowers and lenders.

This is a possible risk that needs to be monitored. It does not mean that the cycle of investment in technology has already turned downward.

A less obvious change is happening in interest rates. On October 13, Lane said that the global expansion of AI investment and expectations of future productivity gains might be pushing long-term interest rates around the world higher. AI investment in Europe is relatively limited in size, but Europe is also affected when conditions in international bond markets change.

However, long-term interest rates are influenced by many other factors, including inflation, government borrowing, savings, the extra return investors demand for taking risks, and expectations about monetary policy. More investment in AI does not automatically mean that the natural interest rate will rise.

If AI keeps improving productivity, it may create more profitable investment opportunities. It could also raise incomes and savings and make it easier for businesses to repay their debts. The overall effect on the long-term equilibrium real interest rate still needs to be tested against actual data.

Perez's distinction between financial capital and productive capital is still useful here. Financial markets can put a value on the promise of new technology in advance, while businesses must go through a long process of building facilities, selling products and collecting payments.

Technology can continue to improve without asset prices rising at the same pace. Changes in financing costs can then affect which projects go ahead and which new ideas can obtain financial support.

This makes the relationship between AI and monetary policy more complicated. Central banks must watch not only for price increases caused by investment demand but also for the effects of AI on the economy's future ability to produce goods and services. They must also pay attention to new risks linked to growing credit, changes in how businesses obtain financing, and the use of AI by financial institutions themselves.

The policy direction set out by Pan Gongsheng is also important in practice. He said that, at an early stage, the People's Bank of China worked with relevant departments to study how to promote AI in the financial sector. Their work covered improving infrastructure, building the ability to use AI, expanding the situations in which it can be used, and preventing and managing security risks. They also developed and issued policy documents on "artificial intelligence + finance." The People's Bank of China will continue to explore ways to use AI to help analyze monetary policy conditions, identify financial risks, and support payments, credit reporting and anti-money-laundering supervision. At the same time, it will guide financial institutions to guard against risks involving data security and the use of AI models.

Major central banks around the world have begun to study and assess AI. They face several economic processes moving at different speeds: investment is already happening, AI is spreading into practical use, productivity gains still need to be confirmed, and interest rates and asset prices are changing every day.

Monetary policy must make decisions about current growth, prices and financial stability before the expected future benefits have fully arrived.

This article is from the WeChat Official Account "Economic Observer", written by OUYANG Xiaohong and published with authorization from 36Kr.