The Global Economy Is Betting Heavily on AI, but Warning Signs and Rising Debt Are Straining Wall Street

Global financial leaders have warned that the world economy is becoming increasingly dependent on artificial intelligence (AI) investment, even as rising debt and mounting credit pressures strain Wall Street.

The International Monetary Fund (IMF) has cautioned that if AI fails to generate the revenue investors expect, the consequences could spread across the global economy.

Countries are increasingly linked through AI-related investment, infrastructure spending and trade. Spending on AI infrastructure is helping drive US economic growth and boosting hardware exports across Asia.

But analysts warn that the gap between investment and expected returns could expose more than $4 trillion (£3 trillion) in AI assets to write-downs by 2032.

Spending Outpaces Projected Revenue

AI infrastructure providers are projected to spend $7 trillion (£5.25 trillion) by 2030, while research and advisory company Gartner forecasts that AI revenue attributable to data centre over the period will reach no more than $3.6 trillion (£2.7 trillion).

Data centre owners reportedly need a cumulative $8.5 trillion (£6.375 trillion) to cover capital costs, debt financing and an assumed 12% return on investment.

The challenge is that businesses may need 10 to 20 years to redesign their operations around AI, according to industry estimates. That timeline raises questions about whether infrastructure providers can generate enough revenue to justify their current spending.

Bank of America analysts have described the boom as a classic capital cycle. High initial profit margins attract investment, eventually creating excess capacity and putting pressure on returns.

Debt now finances about one-third of AI capital expenditure, rather than companies relying solely on their cash reserves. But borrowing has become more expensive, with bond yields reaching 20-year highs.

AI-related debt issuance fell from $113 billion (£84.75 billion) in June to $23 billion (£17.25 billion) in September 2026. JPMorgan is among the financial institutions tightening credit to limit its exposure to concentrated risks.

Credit pressures are also emerging around major infrastructure projects. About $18 billion (£13.5 billion) in bank loans tied to an Oracle data centre campus recently moved into stressed territory. Meanwhile, Firmus, a neocloud company backed by Nvidia, shelved an initial public offering that had sought a valuation above $30 billion (£22.5 billion).

Neocloud Providers Face Structural Pressure

Neocloud providers rent graphics processing unit (GPU) capacity to companies such as Microsoft and OpenAI. Gartner projects that more than 80% will cease operating independently by 2030. These businesses face a particular challenge: processors lose value as newer chips arrive, potentially leaving operators servicing debt on ageing hardware.

CoreWeave has accumulated $50 billion (£37.5 billion) in debt and raised its annual capital expenditure forecast to between $35 billion (£26.25 billion) and $39 billion (£29.25 billion). Analysts expect annual losses of $2.13 billion (£1.5975 billion).

Yields on its junk bonds have risen to 16%, while credit default swap spreads have reached 1,000 basis points. Analysts have described CoreWeave as a 'canary in the coal mine' for debt-heavy AI infrastructure investments.

CoreWeave's management has pointed to a revenue backlog worth $104.2 billion (£78.15 billion) as evidence of future demand. Nvidia chief executive Jensen Huang has also said the chipmaker agreed to buy unsold CoreWeave capacity through April 2032.

The commitments offer a counterpoint to concerns about the company's debt and losses, although they do not eliminate the broader questions surrounding AI infrastructure spending. AI providers face another challenge as cheaper open-weight models, including those developed in China, and falling token prices put pressure on monetisation.