Are banks also afraid that AI investment will go down the drain? ING urgently sells US$500 million in data center loans, including Stargate

As the scale of financing for artificial intelligence (AI) infrastructure continues to expand, the banking industry has begun to accelerate the transfer of related credit risks.

[Financial Channel/Comprehensive Report] As the scale of financing for artificial intelligence (AI) infrastructure continues to expand, the banking industry has begun to accelerate the transfer of related credit risks. ING is reportedly looking for buyers to sell a portfolio of 12 data center loans worth about US$500 million (approximately NT$000 million), which also includes a US$25 million loan position related to the New Mexico "Stargate" project jointly participated by OpenAI, SoftBank and Oracle.

According to "Bloomberg" citing people familiar with the matter on Friday (6th), in addition to "Stargate", the loan portfolio for sale also includes some loan assets involving Vantage Data Centers and CyrusOne. ING will conduct the auction through a "bidding solicitation" (BWIC) process. Investors must submit bids before November 19, with each bid being at least US$10 million. An ING spokesman declined to comment.

The deal highlights how banks are increasingly reselling data center loans to private credit companies, asset managers and insurance companies to diversify credit risks from investments in AI infrastructure.

Hank Calenti, global market strategist at Sumitomo Mitsui Banking Corporation (SMBC), pointed out that investment banks usually adopt a "lending and distribution" model, first providing loans and then reselling related claims to other investors. Buyers may include private credit investors, asset managers and insurance companies, mainly optimistic that such assets are expected to bring higher returns.

However, as the expansion of AI infrastructure has led to a rapid increase in debt scale, and investors have begun to worry that actual returns may not be as expected, related financing risks have gradually surfaced.

The Democratic Party's huge victory in the U.S. midterm elections further deepened market concerns. The Democratic Party may push some states to suspend the construction of data centers or impose stricter restrictions on the construction of data centers. The market is worried about the slowdown in project progress, which has triggered a sharp decline in AI concept stocks on Thursday (5th). The Nasdaq Composite Index, which has a relatively high proportion of technology stocks, closed down 5.5% that day, setting the largest single-day decline since April 2025 and entering a correction range.

Pruksa Iamthongthong, head of Asia Pacific equities at Aberdeen Investments, said that the Democratic Party's gaining more political influence has made the market more worried that data center development will face stricter scrutiny, and investors have already questioned the financing methods and investment returns of large-scale AI capital expenditures. She believes that short-term market fluctuations are likely to continue, and investors will continue to evaluate the risks of delays in U.S. data center construction and limited computing capacity.

Carenti said whether banks will sell further loans out of concern about data center construction moratoriums will depend on the discounts they must accept at the time of sale, as well as the benefits of exiting the exposure and freeing up capital to continue lending. In addition, banks must also consider the costs of communicating the exposure to their boards. He believes that this may just be the daily business of banks, or it may mean that financing players hope to exit before the market environment worsens.

In fact, Wall Street banks have already begun looking for ways to reduce their exposure to data center loans. Earlier this year, lenders such as JPMorgan Chase and Morgan Stanley were reported to have studied various ways to reduce related lending risks.

The Financial Times reported on May 3 that relevant practices include reselling part of the debt to non-bank lenders through private transactions, and using the "significant risk transfer" (SRT) mechanism to transfer part of the bank's credit risk to investors, thereby freeing up balance sheet space and allowing the bank to undertake more loans.