The artificial intelligence boom is creating enormous opportunities for the global economy, but the scale and financing of the investment behind it could also introduce new risks to the financial system, according to the head of the Bank for International Settlements.

BIS General Manager Pablo Hernández de Cos has warned that the rapid expansion of AI is becoming increasingly important to central banks because of its potential impact on productivity, employment, asset markets and financial stability.

The numbers explain why.

The BIS estimates that the world’s five largest technology companies are expected to invest more than $1 trillion in artificial intelligence across 2025 and 2026.

Global AI investment could reach around $4 trillion by 2030.

That puts artificial intelligence in the territory of a major global infrastructure cycle rather than an ordinary technology upgrade.

Where Is All the Money Going?

Training and operating modern AI systems requires enormous physical infrastructure.

Technology companies are building data centres containing thousands of advanced processors.

Those facilities require networking equipment, high-bandwidth memory, cooling systems and enormous electricity supplies.

The largest technology companies are consequently committing unprecedented amounts of capital to infrastructure.

But even companies with enormous cash flows cannot necessarily finance unlimited expansion entirely from operating profits.

Debt markets are becoming increasingly important.

AI’s Debt Machine

Hyperscalers have been issuing bonds at a remarkable pace.

Alphabet, Amazon, Meta, Microsoft and Oracle have collectively issued around $220 billion in bonds over the past year as data-centre investment accelerates.

Other companies in the AI infrastructure chain are borrowing heavily as well.

The BIS is particularly concerned about financing arrangements that may be less transparent than traditional corporate debt.

Private credit and special-purpose financing structures can make it harder to understand where risks ultimately sit.

That is manageable while AI investment produces the returns investors expect.

The problem emerges if it doesn’t.

What If AI Revenue Disappoints?

Today’s spending assumes enormous future demand.

Technology companies expect AI to become embedded across business software, search, advertising, healthcare, science, entertainment and consumer technology.

If those expectations prove correct, enormous data-centre investments could generate strong returns for decades.

But investment booms do not always develop smoothly.

The railway boom transformed transportation.

The internet boom transformed communication.

Both also produced periods when investors committed far more capital than individual companies could economically justify.

Artificial intelligence could experience something similar.

The technology might transform the economy while some investments made during the boom still lose money.

Productivity Is the Bull Case

The BIS is not arguing that AI has no economic value.

Quite the opposite.

Research reviewed by the institution shows generative AI producing productivity improvements ranging from roughly 10% to 65% for particular tasks.

If those improvements spread throughout the economy, AI could increase output substantially.

Businesses could automate repetitive tasks.

Workers could complete knowledge work more quickly.

Scientific research could accelerate.

New products and industries could emerge.

That productivity growth is the economic argument supporting today’s enormous investment.

The uncertainty lies in how quickly those benefits will spread.

Financial Markets Are Becoming Connected to AI

Artificial intelligence already occupies an enormous position in equity markets.

Some of the world’s most valuable public companies are directly exposed to the AI boom.

Now the connection is spreading into credit markets.

That means an abrupt change in expectations about AI profitability could affect more than technology stocks.

Corporate bonds.

Private credit.

Infrastructure financing.

Utilities.

Real estate.

Power projects.

Semiconductor suppliers.

All could potentially feel the effects.

Why It Matters

For central banks, AI creates an unusual problem.

The technology can influence several parts of the economy simultaneously.

Massive investment stimulates demand.

Productivity improvements could expand supply.

Automation could affect employment.

Technology valuations influence financial markets.

Data-centre construction affects electricity demand.

The consequences can point in different directions at the same time.

That makes AI increasingly relevant to monetary and financial-stability policy.

The Bigger Picture

There is a useful distinction between an AI bubble and an AI investment boom.

They are not necessarily the same thing.

A technology can be genuinely transformative while financial markets still overpay for particular companies or projects.

The internet demonstrated exactly that.

Many dot-com companies disappeared.

The internet itself became vastly more important.

AI could follow a similarly complicated path.

What Happens Next

Investors will increasingly scrutinise returns rather than simply spending announcements.

How much revenue does a new data centre generate?

How quickly does it reach full utilisation?

What is the cost of producing AI inference?

How much debt is required?

Who ultimately carries the risk?

Those questions will determine whether today’s extraordinary investment becomes the foundation of a new economic era—or produces financial stress along the way.

Artificial intelligence is already changing technology.

The BIS warning highlights something bigger:

AI is becoming large enough to matter to the financial system itself.