Pablo Hernandez de Cos, the head of the Bank for International Settlements (BIS), told delegates at a conference hosted by India’s central bank that the swift expansion of artificial intelligence poses fresh risks to financial stability. He said investment in AI-related infrastructure has already reached a magnitude that can alter global economic conditions.
Hernandez de Cos underscored that, for central banks, the advent of AI does not alter monetary policy mandates. However, he warned that AI complicates macroeconomic analysis by exerting simultaneous effects on demand, supply and financial markets - factors that make economies harder to interpret.
The BIS has calculated that the world’s five largest technology firms will invest more than $1 trillion in AI between 2025 and 2026. Industry forecasts referenced by Hernandez de Cos indicate that global investment in AI could expand from roughly $500 billion at present to as much as $4 trillion by 2030. "The promise of AI is real," he said, while stressing that long-term outcomes will hinge on policy choices, investment in skills and infrastructure and on the breadth of benefit-sharing.
One notable concern the BIS chief flagged is the changing makeup of AI financing. He said the current boom is increasingly being financed through debt and private credit rather than through corporate earnings, and that much of this funding remains "opaque and interconnected." That opacity, in his view, merits close scrutiny because it can conceal linkages and concentrations that elevate systemic risk.
Hernandez de Cos also drew attention to shifts in global trade flows linked to AI. Economies that are closely integrated with the technology supply chain - including South Korea, Singapore, Malaysia and Taiwan - have seen benefits through stronger export prices for AI chips and related equipment. Those trade effects are part of how AI is already influencing economic outcomes beyond domestic investment.
On the productivity front, Hernandez de Cos cited evidence that generative AI can deliver significant gains in particular tasks. Studies he referenced report productivity improvements ranging from 10% to 65% in activities such as coding, consulting and professional writing. The broader question, he said, is how much of those task-level improvements translate into economy-wide productivity growth.
Current estimates presented by Hernandez de Cos suggest AI could raise total factor productivity growth by about half a percentage point per year, depending on the pace of adoption and the effectiveness of labour and capital reallocation. Advanced economies are expected to see the earliest gains because of their larger service sectors and greater readiness to deploy AI. By contrast, prospects for emerging economies are more varied. Hernandez de Cos said India has a "genuine opportunity" to narrow gaps, helped by its digital public infrastructure.
Workforce effects are already visible in certain segments, according to the BIS chief. While overall job losses to date have been limited, there are early signs of displacement in customer service, programming and administrative roles. Hernandez de Cos highlighted the growing importance of retraining and reskilling as these changes unfold.
He also warned that lofty market valuations, increasing concentration in a few large firms and opaque financing structures could create vulnerabilities if corporate profits fall short of expectations. "I do not say that this is where the AI boom must lead," Hernandez de Cos said. "But the scale and speed of the current investment boom, and the weight of expected commercial returns, do warrant some caution," he added, drawing parallels in tone with earlier historical booms such as the railway expansion and the dotcom surge.
Implications for markets and policy
Hernandez de Cos framed AI as a force that simultaneously reshapes demand patterns, supply capabilities and the structure of financial intermediation. That intersection - large-scale capital deployment, altered trade dynamics and new financing channels - is where financial-stability risks may arise, he argued.