2 hours ago
9
Asia’s banks have found their favorite customer, and it runs on GPUs.
Taiwanese lenders extended a record NT$4 trillion (approximately $126 billion) in loans during the first seven months of 2026, with much of that money flowing toward the AI supply chain.
Follow the money to the fabs
The standout figure involves working capital loans. These are the short-term credit lines companies use to keep operations running, such as paying suppliers, buying inventory and covering payroll while orders get filled.
For AI supply chain firms linked to TSMC, Nvidia and Microsoft, those loans grew nearly eight times faster than mortgage lending over the same period.
The boom is not limited to Taiwan. Across Asia-Pacific, G3 syndicated lending, excluding Japan, reached $171 billion in the first nine months of 2026. That marks a six-year high and an increase of more than 6% year over year.
Syndicated loans are big loans split among a group of banks, so no single lender has to carry the entire risk.
Data center financing reached about $56 billion year-to-date in 2026. For context, the total for all of 2025 was $26 billion.
The Asian data center debt market has climbed to nearly $29 billion since early 2025. That includes a record $15 billion in 2026 alone, a 27% year-over-year increase.
Why the hyperscalers are borrowing now
Hyperscalers and AI firms are increasingly turning to debt to fund enormous capital spending on chips and data centers. According to Morgan Stanley, hyperscaler debt leverage rose from roughly 0.9x to 1.8x within six months.
Global AI-related debt issuance could reach around $570 billion in 2026, based on projections cited in the research.
Corporate loan rates have risen by around 0.9 percentage points as this sector absorbs significant bank capacity.
The concentration problem regulators are watching
Authorities in Japan and Taiwan have stepped up scrutiny as lending to AI-linked companies balloons. Regulatory warnings have flagged concentration risk, meaning too much of the financial system’s exposure piling up in one place.
Southeast Asia has been singled out as particularly vulnerable. Economies that have leaned into data center construction and AI-related investment could feel a downturn sharply if the AI market stumbles.
Credit appears to be getting rationed for non-tech borrowers as banks prioritize AI-related lending.
What this means for borrowers, banks and investors
Rising leverage at hyperscalers means more of the AI story now depends on borrowed money being repaid out of future AI revenue. Companies carrying 1.8x leverage have less room to absorb a slowdown than those sitting at 0.9x did six months earlier.
Investors watching this space should keep an eye on regulatory action in Japan and Taiwan, including any limits on sector exposure. Hyperscaler earnings will show whether AI revenue is keeping pace with the debt being raised to chase it. And data center lending volumes for the rest of 2026 will reveal whether the current pace is sustainable or a peak.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·