Japan's FSA puts AI data center loans under scrutiny
Japan's financial regulator has started looking past the AI data centers and at the balance sheets paying for them.
The Financial Services Agency named project finance — data centers included — a priority area for monitoring in its policy priorities for the 2026 fiscal year, and officials say the near-term focus is how the biggest banks and life insurers underwrite and then watch those exposures. Japan's lenders have become some of the largest creditors to the global AI buildout, largely because a decade of near-zero rates left them with cheap deposits and few domestic places to put them. Sanae Takaichi's government has encouraged exactly that: a growth plan that targets roughly ¥250 trillion in private investment by FY2040, with AI data centers as a named destination.
That is why the scrutiny is worth reading carefully. This is not a clampdown, and the FSA has said as much — it treats the sector as a growth industry and does not intend to stop the lending. What it is doing is checking the plumbing: whether banks' lending policies, deal screening, post-disbursement monitoring and concentration-risk management can actually hold up against loans whose repayment depends on hyperscalers earning a return on enormous capex. That return is the live argument in the industry right now, and regulators are no longer willing to take it on faith.
The uncomfortable detail is the direction of travel. Japan spent July loosening its guardrails, letting banks exceed existing large-exposure ceilings for big M&A and AI data center financing, and reports this week said the FSA plans to relax those limits further for regional banks doing AI and renewables deals. So the same regulator is simultaneously opening the tap and tightening the meter. In practice that means Japan is not betting against the buildout — it is trying to make sure it is lending against contracted cash flows rather than against the story. With no banks and no projects named, and the exposure concentrated in US-based projects, there is not much to size yet. But the FSA's own priorities document asks a question the rest of the AI financing market should be answering too: when the concentration sits at a handful of institutions, who is watching the total?
The FSA's document covers domestic real estate lending, large exposures, overseas fund lending and project finance in the same breath — a reminder that data centers are being underwritten with the same muscle as office towers. Elsewhere, other regulators have already made the same move: the ECB reviewed AI-sector risks to euro-area banks earlier this year, and Singapore's Monetary Authority has warned that the sustainability of large-scale AI investment is now a global financial-stability question, not just a technology one.
What to watch: whether the FSA publishes exposure figures, and whether the large-exposure relaxation lands as scheduled — loosening concentration limits and tightening concentration monitoring in the same quarter is a contradiction the regulator will eventually have to resolve out loud. If hyperscaler lease commitments keep growing faster than the return picture firms up, expect Japan to be the first regulator to say so in numbers.
Japan wants to grow an AI sector and police the loans funding it at the same time. Can any regulator pull both off? Tell us in the comments.
