Japanese Banks Confront the Risks Behind AI Lending

The financial institutions helping to fund the artificial intelligence boom are beginning to examine what might happen if the pace of investment slows. Japan's largest banks have warned that weaker spending on data centres could affect economic growth and credit conditions, even as they report stronger earnings from their traditional lending businesses. Their concerns reflect the extraordinary amount of capital being committed to AI infrastructure, where developers are constructing facilities that depend on expensive processors, reliable electricity supplies and expectations of sustained demand for computing services. The question facing lenders is whether those investments will generate sufficient commercial returns to support the financing arrangements behind them.

Mitsubishi UFJ Financial Group has disclosed approximately ¥3.3 trillion in data-centre credit exposure, compared with total credit exposure of roughly ¥180 trillion. The figures suggest that direct lending to the sector represents a relatively small proportion of the group's overall portfolio. Sumitomo Mitsui Financial Group has meanwhile tightened lending standards for data-centre projects, placing greater emphasis on highly rated hyperscale technology companies. Such measures reflect a preference for borrowers with stronger financial resources and a greater ability to withstand changes in demand or operating costs. They also demonstrate that banks are distinguishing between the broader promise of artificial intelligence and the specific financial risks associated with individual infrastructure projects.

Data-centre financing presents several challenges. Facilities require substantial investment before they begin producing revenue, and their economic performance depends on utilisation, electricity prices and the commercial arrangements established with customers. Rapid technological change introduces another uncertainty because equipment may lose competitiveness before its expected economic life has ended. Large technology companies can sometimes absorb those risks through diversified operations and substantial cash reserves. Smaller developers may depend more heavily on external financing or contracts with customers whose future requirements remain uncertain. For banks, assessing creditworthiness therefore involves examining not only the value of the physical infrastructure but also the durability of the demand it is intended to serve.

The Japanese warnings do not establish that the AI infrastructure industry is approaching a financial crisis. Nor do they indicate that the country's largest banks face immediate losses from their existing commitments. They do, however, reveal a growing awareness that investment in advanced computing cannot be evaluated solely through expectations of technological progress. Projects must ultimately produce cash flows sufficient to justify their construction and financing costs. A slowdown could affect equipment suppliers, electricity providers, construction companies and lenders, creating consequences beyond the technology sector itself. Japanese banks are responding by attempting to preserve access to a potentially important growth market while limiting the exposure that could arise if expectations prove too optimistic.