BIS Warns AI Capex Arms Race Fuels Systemic Risk

BIS chief warns AI capex arms race relies on opaque debt, posing systemic risks

The head of the Bank for International Settlements (BIS), Pablo Hernández de Cos, warned on Thursday that the breakneck surge in artificial intelligence investment could become a financial-stability risk if the expected returns fail to materialize.

In remarks tied to the BIS Annual Economic Report 2026 and delivered at a conference hosted by India’s central bank, Hernández de Cos said the scale and speed of the AI infrastructure buildout—particularly data centres and related capacity—warrant “some caution,” even as he emphasized that AI’s economic promise is real.

The BIS highlighted that the five largest technology “hyperscalers” are expected to spend more than $1 trillion on AI-related capital expenditure across 2025 and 2026. According to the BIS, these commitments now exceed both earnings and free cash flow for some firms, leading them to issue debt to keep funding expansion.

The bank also flagged the broader trajectory of global AI financing, noting estimates that worldwide AI investment could reach $3 trillion to $4 trillion by 2030.

A core concern, Hernández de Cos said, is that the boom is increasingly being funded through debt and private credit rather than internally generated cash, with parts of the financing described as “opaque and interconnected.” The BIS also pointed to the growth of private credit loans to AI-related companies—from about $3 billion in 2010 to more than $40 billion in 2025—and warned that circular structures mixing equity, debt, and supplier contracts can create exposures that may be “pledged multiple times.”

To underline the risks of investment cycles driven more by expectations than by near-term profits, Hernández de Cos referenced historical episodes including the railway and dot-com bubbles, where heavy spending and optimistic assumptions contributed to painful corrections.

The BIS framed the AI capex boom as one of several “pressure points” facing the global economy, alongside returning inflation, strained public finances, and growing financial vulnerabilities—stressing that these risks can interact rather than remaining isolated.

At the same time, the BIS chief said evidence already suggests AI can boost productivity in areas such as coding, consulting, and professional writing. However, he argued that the eventual economic impact will depend on how widely benefits are shared and whether policymakers invest in skills, infrastructure, and competition.

For central banks, Hernández de Cos added, AI does not change monetary policy mandates but can make economies harder to interpret by affecting demand, supply, and financial markets at the same time. He also noted that job losses have been limited so far, while early signs are emerging in customer service, programming, and administrative roles—reinforcing the importance of retraining and reskilling.