ING's Auction Fails to Clear at Acceptable Prices
Dutch banking group ING has encountered resistance from credit investors in its attempt to sell approximately $500 million of loans linked to US data centre projects, adding to concerns about the pricing and liquidity of debt financing the rapid expansion of artificial intelligence infrastructure.

According to Bloomberg's reporting on 20 November 2026, the bank's competitive auction failed to generate sufficient interest at acceptable prices before the 19 November bidding deadline. ING is now exploring individual loan sales, with prospective buyers reportedly engaging directly with the bank after the formal auction window closed. The development marks a significant change from the original disposal strategy, which sought to attract competitive bids for approximately a dozen loan exposures through a bids-wanted-in-competition process.

The portfolio includes financing associated with several prominent data centre operators, including Vantage Data Centers and CyrusOne, alongside a $25 million loan participation linked to Project Jupiter in New Mexico. The latter forms part of the broader Stargate AI infrastructure initiative involving OpenAI, Oracle and SoftBank. The reported sale process does not establish that these loans are non-performing or that ING has classified them as impaired. It does, however, indicate that prospective buyers are unwilling to acquire at least some of the exposures at prices the bank considers acceptable.

The distinction is particularly important. A bank may seek to sell performing loans to reduce concentration, release regulatory capital, manage liquidity or create capacity for new lending. A failed auction does not necessarily indicate deterioration in contractual loan performance. Nevertheless, when a seller cannot achieve acceptable bids in a competitive process, the difference between the bank's carrying value, its reservation price and the market's assessment of risk becomes an increasingly relevant consideration.

Project Jupiter and the Emerging Price of Construction Risk
One of the more closely watched exposures in ING's proposed sale is the $25 million participation associated with Project Jupiter, a large data centre development in New Mexico. The project has faced difficulties involving power infrastructure, and Oracle previously issued a notice invoking force majeure in connection with the development. Bloomberg reported that bids for ING's Jupiter-related loan were particularly low, while other market reporting has indicated that comparable exposure was being discussed at discounts to face value.

These developments illustrate the complexity of underwriting data centre projects whose commercial prospects depend on the timely completion of substantial physical infrastructure. Unlike conventional corporate lending, financing for a development-stage data centre can involve several interconnected risks, including construction completion, electricity availability, grid connection, cooling infrastructure, customer commitments and the financial strength of counterparties responsible for long-term capacity payments.

A delay in securing electricity supply, for example, may postpone the commencement of contracted revenue even when the underlying facility is otherwise substantially complete. That can increase capitalised interest, extend construction financing requirements and reduce the margin available to absorb further cost overruns. The resulting credit exposure cannot be assessed solely by reference to the reputation of the technology companies associated with the project.

There is also a difference between financing a completed, operational data centre with established customers and lending against a project still dependent on construction milestones and infrastructure delivery. Both may fall under the broad description of AI infrastructure financing, but their credit risk characteristics, expected cash flows and potential recovery values can differ substantially.

For banks and investors evaluating these exposures, the key considerations include the contractual allocation of construction risk, the enforceability of completion guarantees, the availability of contingency funding and the extent to which projected revenue depends on a small number of counterparties. Those factors can materially influence the price at which a loan can be transferred, even before any payment default occurs.

European Banks Reassess Their Exposure to AI Infrastructure
ING is not the only European lender seeking to reduce or redistribute exposure to the sector. Recent Financial Times reporting indicates that Société Générale has also explored sales of approximately $300 million in financing linked to Oracle-related data centre projects in Wisconsin and Texas. The same reporting described increasing use of loan syndication, credit insurance and risk-transfer structures as banks attempt to manage concentrations arising from the scale of AI infrastructure investment.

The growing volume of financing required for data centres presents a challenge for traditional bank balance sheets. Individual projects can require billions of dollars in construction and equipment expenditure, while the same financial institutions may have commitments across several developers, operators, technology companies and related infrastructure providers. Although the underlying borrowers and contractual structures may differ, their performance can remain exposed to common factors such as electricity costs, construction inflation, technological change and the pace of AI adoption.

Banks consequently face a balance-sheet management problem that extends beyond individual borrower creditworthiness. Even where loans remain performing, increasing sector concentration can influence internal limits, capital allocation and the willingness to retain long-dated exposures. Selling loan participations or transferring credit risk may help manage those constraints, but the economic benefit depends on the price at which risk can actually be transferred.

The market's response to ING's auction suggests that the availability of prospective buyers should not be confused with the ability to execute a disposal at the seller's preferred valuation. Some investors may be willing to acquire the debt, but only at discounts that compensate them for illiquidity, project-specific uncertainty and the possibility of further deterioration in financing conditions.

For ING, the reported decision to pursue individual transactions may provide greater flexibility in negotiating with buyers whose risk appetite differs across projects. It may also allow the bank to avoid accepting a single portfolio-level discount that does not adequately reflect differences in the underlying loans. Whether that approach produces better execution remains to be seen.

From HSBC's Project Jupiter Disposal to ING's Stalled Auction
The latest development also follows reports that HSBC sold approximately $300 million of exposure associated with financing for Project Jupiter. That transaction, previously examined by Restructuring.Insights, provides useful context for understanding the evolving secondary market for AI infrastructure debt.

The two situations should not, however, be treated as directly comparable without examining the individual loan agreements, security packages, contractual maturities and sale prices. HSBC's reported disposal and ING's unsuccessful auction may involve exposures connected to the same broader development, but their economic characteristics and transaction outcomes are not necessarily identical.

What they demonstrate collectively is that banks are increasingly testing the market's willingness to absorb AI-related credit exposure. The results are beginning to reveal differences between the financing terms accepted when loans were originated and the returns now required by secondary-market investors.

A discounted sale of a performing loan can reflect a change in market yields, reduced liquidity or a reassessment of risk rather than an actual credit loss arising from borrower default. Conversely, a substantial discount may become an important warning signal if it reflects concerns about project completion, the reliability of expected cash flows or the adequacy of contractual protections.

For credit risk managers, the challenge is to distinguish between these explanations rather than treating every secondary-market discount as evidence of impairment. That distinction becomes particularly relevant under IFRS 9, where the assessment of a significant increase in credit risk and expected credit losses requires consideration of borrower-specific circumstances and forward-looking information. A market bid can provide useful evidence, but it does not automatically determine a loan's accounting classification.

What ING's Failed Auction Means for Credit Markets
The immediate significance of ING's stalled auction lies in the failure of a competitive disposal process to achieve acceptable pricing, rather than in any confirmed deterioration of the underlying loans. The bank has reportedly continued discussions with potential buyers, and the eventual outcome may involve partial sales, revised pricing or the retention of some exposures.

The wider implications are more substantial. AI infrastructure financing has expanded rapidly on expectations of sustained demand for computing capacity, but debt investors must ultimately assess whether individual projects can generate sufficient cash flow to service their obligations. That requires greater scrutiny of construction schedules, power availability, contractual revenue arrangements and the ability of project sponsors to absorb delays or additional investment requirements.

The growing secondary market for these loans may also expose valuation differences that were less visible when financing commitments were initially arranged. A bank's willingness to retain a loan at its existing carrying value does not guarantee that another investor will purchase the same exposure at par. As more transactions are tested in the market, the resulting prices may influence future underwriting standards, loan margins and the amount of leverage available to new projects.

For now, ING's experience should not be presented as evidence of a broader default cycle in AI infrastructure lending. The reported facts do not support that conclusion. It is, however, a reminder that credit quality, liquidity and market value are separate considerations, and that even performing loans can become difficult to distribute when investors demand greater compensation for uncertainty.

The next meaningful indicators will be whether ING completes individual disposals, the discounts ultimately accepted and whether comparable loans from other European banks encounter similar resistance. Those outcomes will provide a clearer indication of whether the current pricing difficulties are concentrated in particular projects or represent a more general adjustment in the market for AI infrastructure debt.

Editorial Disclaimer
This article is based on publicly available financial reporting and attributed market information. Details of ING's disposal process are based principally on Bloomberg reporting citing people familiar with private transactions and have not been independently confirmed by ING. The article does not imply that the loans concerned are non-performing, that the borrowers are in default or that any lender has recognised an impairment loss. References to credit risk, valuation and potential recovery outcomes are analytical observations rather than forecasts or allegations of misconduct.

This article does not constitute investment advice. As at the publication date, the author holds no direct or indirect financial interests in the banks discussed in this article.

Sources and References
1. Bloomberg News, 20 November 2026. ING's $500 Million Data Center Loan Auction Stalls on Low Bids. Reporting by Jeannine Amodeo, Reshmi Basu and Carmen Arroyo.
https://www.bloomberg.com/news/articles/2026-11-20/ing-s-500-million-data-center-loan-auction-stalls-on-low-bids

2. Bloomberg News, 5 November 2026. ING Said to Be Seller of $500 Million Data Center Loans. Initial reporting on the proposed auction and portfolio composition.
https://www.bloomberg.com/news/articles/2026-11-05/mystery-seller-puts-up-500-million-data-center-loans-backing-stargate-cyrusone

3. Reuters, 6 November 2026. ING Puts $500 Million of Data-Center Loans Up for Sale, Bloomberg News Reports. Independent publication confirming the initial reported sale process.
https://www.reuters.com/business/finance/ing-puts-500-million-data-center-loans-up-sale-bloomberg-news-reports-2026-11-06/

4. Financial Times, November 2026. Banks Seek to Curb AI Debt Exposure as Backlash Mounts. Reporting on loan disposals, risk transfers and European banks' exposure management.
https://www.ft.com/content/2351d9db-e938-4102-a449-318b1037765a

5. Reuters, 9 November 2026. Banks, Investors Grow Wary of AI Debt, Spooked by Market Selloff. Reporting on changing investor risk appetite and AI infrastructure financing conditions.
https://www.reuters.com/legal/transactional/banks-investors-grow-wary-ai-debt-spooked-by-market-selloff-2026-11-09/

6. Data Center Dynamics, 6 November 2026. ING Bank Looks to Sell $500m in Data Center Loans at Once, Including at Stargate Site. Project-specific background and financing context.
https://www.datacenterdynamics.com/en/news/ing-bank-looks-to-sell-500m-in-data-center-loans-at-once-including-at-stargate-site/