AI investment boom faces financing concerns as banks rethink lending risks

Close-up of high-density AI computing server racks, cooling connections and network equipment inside a data centre during the AI investment boom.
High-density computing infrastructure inside an AI data centre. Representative Image

New York, November 10: Banks and credit investors are becoming more cautious about financing artificial intelligence infrastructure projects, with some lenders reluctant to increase their exposure to data centres and AI computing companies following a sharp sell-off in technology stocks.

The shift comes as borrowing requirements for AI infrastructure continue to grow. According to Goldman Sachs Research projections cited by Reuters on November 9, global debt issuance by major cloud computing companies, known as hyperscalers, could rise from an estimated $250 billion in 2026 to $420 billion in 2027.

The projected increase highlights a growing challenge for the AI industry: companies need substantial funding to build data centres and purchase advanced chips, but lenders are becoming more selective about the risks they are willing to accept.

Reuters reported, citing banking and credit industry sources, that some banks that had recently been competing to finance AI projects were becoming reluctant to take on additional exposure.

In some cases, lenders were reportedly hesitant even when financing arrangements included commitments or guarantees involving major technology companies such as Google and Anthropic.

Neither company confirmed that its financing had been refused. Both declined to comment on the borrowing arrangements discussed in the Reuters report.

The shift in sentiment follows substantial losses in technology shares. On November 5, the Nasdaq 100 lost approximately $2.5 trillion in market value during a sharp sell-off that intensified concerns about the pace and profitability of AI investment.

Investors have also been assessing the possibility of increased regulatory scrutiny of data centre development following the US midterm elections.

According to Reuters, lenders are examining the financial arrangements behind proposed AI infrastructure projects more closely, particularly borrowers’ ability to generate sufficient revenue to repay their debts.

Banks and investors are stress-testing projected cash flows under weaker economic conditions and reviewing construction deadlines, service obligations and contractual guarantees.

They are also paying closer attention to provisions that allow major technology customers to reduce or terminate their commitments.

These conditions matter because many AI infrastructure projects require substantial investment before they begin generating significant revenue.

A data centre developer may borrow money to construct a facility and purchase expensive computing equipment, expecting technology companies to rent its capacity over several years.

If construction is delayed or a major customer reduces its commitment, the developer could face difficulties meeting its repayment obligations.

The risks are particularly relevant to specialised AI cloud providers, often called neocloud companies, which rent advanced computing capacity to businesses developing and operating artificial intelligence systems.

Reuters identified CoreWeave and Nebius among the companies facing heightened scrutiny in the sector.

CoreWeave’s shares had fallen approximately 35% since the beginning of the year, while its bonds had also declined in the secondary market.

Nebius, meanwhile, had explored a high-yield bond offering in September but decided not to proceed with the proposed issuance in October, according to a source cited by Reuters.

However, the company also secured a $3.5 billion senior secured loan facility from Citigroup in October for data centre development, graphics processing unit procurement and other corporate purposes.

These contrasting developments illustrate that financing remains available for some AI businesses, even as investors become more cautious about particular deals.

The shift is also reflected in borrowing costs.

According to Bank of America credit analyst Yuri Seliger, the spread on 10-year hyperscaler debt had widened by approximately 50 basis points since the beginning of 2026, compared with around 20 basis points for high-quality industrial companies outside the AI sector.

A wider credit spread generally means investors demand additional compensation for lending to a borrower, increasing the cost of raising debt.

Large banks are also monitoring their exposure to infrastructure-related lending.

At JPMorgan Chase, loans backed by a category of income-producing properties that includes data centres totalled approximately $29 billion at the end of the third quarter.

That category accounted for 12% of the bank’s commercial real estate lending exposure, compared with 8% three years earlier.

However, the $29 billion figure includes other property types and should not be interpreted as the bank’s data centre lending alone.

Despite growing caution, demand for AI infrastructure financing remains substantial.

Goldman Sachs Research expects global hyperscaler debt issuance to reach $420 billion in 2027, more than 60% above its $250 billion estimate for 2026.

These figures are projections of borrowing activity, not confirmed losses, defaults or outstanding debt.

For banks and investors, the central question is whether future revenue from AI services and computing infrastructure will be sufficient to justify the scale of borrowing required.

The Reuters investigation points to tighter scrutiny and greater selectivity rather than a complete withdrawal of financing from the AI industry.

It does not establish that lenders have imposed a universal credit freeze or that a widespread wave of data centre cancellations has begun.

For now, growing caution suggests that AI infrastructure developers may face more demanding financing conditions, particularly where projects depend heavily on debt, uncertain customer commitments or construction schedules that could be delayed.

Read also: Odisha’s Kalinga Ghat Reopens After 46 Days; Heavy Vehicles Still Restricted on Gajalbadi Route

Read also: Odisha’s Kalinga Ghat Reopens After 46 Days; Heavy Vehicles Still Restricted on Gajalbadi Route