The massive investment required to build artificial intelligence infrastructure is creating financial risks that could spread beyond technology companies, India’s Chief Economic Adviser Anantha Nageswaran warned, pointing to rising borrowing, complex financing arrangements and circular investment relationships.
Nageswaran told the 2026 Per Jacobsson Lecture in Bangkok that the AI investment boom deserved scrutiny even if the technology ultimately delivers the productivity gains its supporters anticipate.
He cited estimates suggesting that the United States may require more than $10 trillion in data-centre investment by 2032, equivalent to about 3.6% of gross domestic product annually.
According to Nageswaran, major cloud companies were expected to spend more on capital expenditure than they generated in cash for the first time in 2026, increasing the importance of external financing.
He warned that debt was increasingly being raised through structures that could make the underlying financial exposure less visible on corporate balance sheets.
As an example, he cited Meta’s Hyperion data-centre project, which he said was about 90% debt-financed through a joint venture whose bonds did not appear on Meta’s balance sheet.
Nageswaran also highlighted financial links between AI developers and the cloud companies supplying their computing infrastructure. Large technology companies invest in AI laboratories, which in turn sign substantial computing contracts with those same providers.
Such relationships can complicate assessments of underlying demand and profitability, particularly when investments and commercial contracts reinforce each other’s reported value.
He said about a third of the earnings of the five largest cloud companies in the first quarter came from marking up their stakes in other businesses, according to the figures he presented.
The risks could extend into financial markets through leveraged exchange-traded funds concentrated in AI stocks and increased hedge-fund borrowing from prime brokers. Nageswaran said the latter had doubled since 2023.
The concerns are significant because the financial system may become exposed to a technology investment cycle whose returns remain uncertain. If AI adoption delivers weaker profits than expected, heavily indebted projects and investors could face pressure. Even if the technology succeeds, the distribution of its economic gains may not support the valuations currently attached to AI companies.
Nageswaran urged financial supervisors to treat debt embedded in leases, residual-value guarantees and joint ventures as genuine financial exposure when assessing risks. He also called for the Financial Stability Board and other authorities to map these exposures across borders.
He further warned governments against incorporating expected AI productivity gains into fiscal forecasts before those gains materialise in economic data.
The lecture raised questions about whether the financing structure supporting AI infrastructure could amplify losses if investment expectations prove too optimistic.
