IMF warns of AI-linked economic risks

The International Monetary Fund (IMF) has issued a warning about the growing financial risks tied to the massive debt taken on by major technology companies to support the global expansion of artificial intelligence (AI) infrastructure. This includes the construction of large data centers that power AI systems.

Hyperscalers face cash flow challenges

The largest data center operators, known as "hyperscalers," are generating substantial profits but are struggling with declining cash flow as they invest heavily in AI infrastructure. To meet these financial demands, they are increasingly relying on debt financing, which the IMF views as a growing concern.

Circular financing raises risks

The IMF report highlights the use of circular financing agreements, where one company funds another's AI projects in exchange for future product purchases. These arrangements could create complex financial interdependencies, increasing the risk of a financial crisis if one company fails.

Hyperscalers hold massive debt

The five major hyperscalers — Oracle, Microsoft, Meta, Amazon, and Alphabet — collectively carry over $1.3 trillion in debt and around $3 trillion in off-balance-sheet commitments. Much of this infrastructure was built to support AI research firms like OpenAI and Anthropic, which are currently unprofitable.

AI's productivity promises unfulfilled

Economist Dean Baker pointed out that while AI is expected to drive a productivity boom, there is little evidence that it has delivered the promised gains. If AI does not significantly boost productivity, the investments in the sector could lead to a stock market crash and a deep recession.

AI bubble compared to dotcom era

Torsten Slok, chief economist at Apollo Global Management, warned that the current AI investment boom is even larger than the dotcom bubble of the late 1990s. He suggested that a collapse in the AI sector could lead to a major economic downturn, similar to what happened in the past.

AI development faces real-world constraints

In March 2026, Edward Zitron criticized the AI bubble, pointing out that the rapid sales of processors have not matched the slow pace of data center construction. This is due to energy and logistical challenges, and he accused tech companies of hiding these delays and finding ways to bypass export restrictions to China.

Nvidia CEO remains optimistic

Despite concerns about the AI bubble, Nvidia CEO Jensen Huang believes the sector is still on a strong trajectory. He argues that the growth in AI spending is driven by a fundamental shift in computing infrastructure, which could sustain the industry for some time.