Credit Concerns Mount as AI Infrastructure Debt Grows and Risk Transfer Becomes Harder
Key Facts
The credit market is showing early signs that it may be nearing its capacity to absorb more debt tied to AI infrastructure, according to Bloomberg. This reflects growing concern about lenders’ ability to keep financing projects and transfer their risks to other investors. The news agency said bankers are finding it harder to transfer the risks of data center loans to insurers and other investors. That difficulty points to tighter risk distribution in the market, not a broad freeze in lending. In the latest sign of the pressure, Bloomberg reported on November 5 that ING is seeking to sell about $500 million in data center loans.
One prominent case involves a data center project in Texas linked to Anthropic. On October 28, Bloomberg reported that banks had delayed plans to sell part of a $15 billion loan to finance the facility. The delay came after Texas paused new projects to assess risks to the power grid. This shows how regulatory and infrastructure factors can affect the timing of transferring loans to other investors. The delayed plans concerned part of the financing, not cancellation of the loan or a halt to funding for the entire sector.
On November 5, Bloomberg reported that ING is exploring the sale of about $500 million in data center loans amid concerns that election results could disrupt projects. Putting the loans up for sale means the bank is seeking to transfer part of its exposure to potential buyers at a time when risk distribution is becoming more difficult. The move comes within the broader picture Bloomberg described of the market’s declining capacity to absorb additional debt for AI infrastructure. But it alone does not establish that ING or other lenders are withdrawing from financing all projects. The available information also does not specify the sale terms or potential buyers.
Concerns are growing alongside the scale of expected capital investment by major cloud computing companies. S&P Global Ratings forecast that their combined capital expenditure would exceed $1.3 trillion by 2027. The agency also forecast that the six largest companies would record negative free operating cash flow in 2026 and 2027. This outlook puts substantial financing needs at the center of the debate over companies’ ability to continue expanding. These figures remain forecasts by S&P Global Ratings, not realized financial results.
S&P Global Ratings expects its models to reach an inflection point in 2028. This outlook depends on faster revenue growth and more moderate capital expenditure growth as returns on investment improve. That means the timing of stronger returns remains central to assessing the financing outlook, alongside the scale of spending required in the preceding years. The agency’s forecast does not guarantee that this inflection point will be reached; it describes the broad assumption underpinning its models. Current financing pressure remains tied to the market’s ability to absorb debt before investments translate into revenue under these models.
In Europe, European Central Bank officials have discussed internally and with lenders the idea of reviewing data center loans. Bloomberg reported on November 6 that the proposed aim is to assess risks related to shifts in sentiment and valuations tied to AI. The agency said senior officials at the bank raised the idea within the institution and in discussions with the entities they supervise. This indicates that banks’ exposure to this type of lending is being monitored by Europe’s supervisory authority. The available information did not say that a formal review had been approved or begun.