Global Banks and Investors Curb Lending to AI Companies as Debt Risks Rise

Global banks and institutional investors are becoming increasingly cautious about lending to artificial intelligence companies as concerns grow over the enormous debt financing behind the AI infrastructure boom. After months of strong enthusiasm for funding data centres, advanced computing infrastructure and AI cloud providers, lenders are beginning to question whether the sector can generate enough revenue to justify its rapidly increasing borrowing.

The shift follows a sharp market sell-off and growing uncertainty about the future growth of artificial intelligence businesses. According to a Reuters report published on November 9, 2026, some banks are now reluctant to finance additional AI-related deals, even when major technology companies provide guarantees. Investors are also examining borrowers’ cash flows, customer contracts, construction schedules and repayment capacity more carefully.

Key Takeaways

• Global banks and institutional investors are becoming more selective about lending to AI companies.
• Lenders are increasingly concerned about the amount of debt used to finance AI data centres and computing infrastructure.
• Some banks are reportedly reluctant to fund new transactions even when major technology companies provide guarantees.
• Goldman Sachs estimates that global hyperscaler debt issuance could reach $420 billion in 2027, compared with an estimated $250 billion in 2026.
• Credit spreads on 10-year hyperscaler debt have widened by approximately 50 basis points year-to-date.
• Riskier borrowers, including neocloud companies and independent data-centre operators, face greater financing pressure.
• AI cloud company Nebius reportedly abandoned plans for a high-yield bond offering.
• CoreWeave’s shares have fallen approximately 35% year-to-date, reflecting growing investor concerns.
• Lenders are increasingly examining customer commitments, cash flows and the conditions under which contracts can be terminated.

Global Banks Become More Cautious About AI Lending

The rapid expansion of artificial intelligence has created enormous demand for computing infrastructure. Companies developing AI models and applications need advanced chips, servers, electricity and large data centres to train and operate their systems.

For much of the past year, banks and private credit investors were eager to finance these projects. Major technology companies and AI infrastructure providers attracted billions of dollars in loans and other financing arrangements.

However, the lending environment is changing.

According to Reuters, several banking and credit-market sources said some lenders have recently become reluctant to increase their exposure to AI-related borrowers.

The shift has occurred despite the involvement of well-known technology companies that can provide financial guarantees or long-term commercial commitments.

Banks are now questioning whether the size of their existing exposure is becoming too large relative to the potential returns and risks.

The change could make it more difficult for AI infrastructure companies to secure funding for new data centres and computing facilities.

Why Are Banks Pulling Back?

Risk factor	Why lenders are concerned
High borrowing	AI infrastructure projects require enormous amounts of capital
Uncertain revenue	Future AI demand may not grow quickly enough to justify investment
Customer concentration	Some infrastructure providers depend on a small number of large customers
Construction delays	Delayed projects can postpone revenue and debt repayment
Technology obsolescence	Advanced chips and computing equipment can lose value quickly
Market volatility	Falling asset prices can increase financing risks

The change does not mean banks have stopped financing artificial intelligence companies altogether. Instead, lenders are becoming more selective about borrowers, project economics and the structure of individual transactions.

AI Companies Need Hundreds of Billions of Dollars in Debt

One of the biggest reasons behind the financing concerns is the enormous amount of money required to build AI infrastructure.

Major cloud providers, including companies operating large-scale computing networks, are spending heavily on data centres, servers and advanced processors.

Much of this expansion requires external financing because infrastructure investments are too large to be funded entirely through operating cash flows.

Goldman Sachs Research estimates that global hyperscaler debt issuance could reach approximately $420 billion in 2027, compared with an estimated $250 billion in 2026.

Financial year	Estimated hyperscaler debt issuance
2026	$250 billion
2027	$420 billion

Figures are estimates cited by Reuters, not completed borrowing totals.

The projected increase represents more than 60% growth in annual debt issuance.

Hyperscalers are large technology companies that operate extensive cloud computing and digital infrastructure networks.

They are investing in AI infrastructure because demand for computing power is increasing as businesses adopt generative AI, AI agents and other advanced applications.

However, lenders are beginning to ask whether the revenue generated by these investments will be sufficient to support the borrowing required.

Investors Are Becoming More Concerned About AI Debt

The growing financing requirements are also changing investor behaviour.

Credit investors previously showed strong interest in AI-related bonds and loans, partly because large technology companies offered financial support or guarantees for certain infrastructure projects.

Now, investors are paying greater attention to the actual cash flows behind these transactions.

According to Reuters, the spread on 10-year hyperscaler debt has widened by approximately 50 basis points year-to-date.

By comparison, spreads on debt issued by high-quality industrial companies not directly linked to AI have widened by around 20 basis points.

A credit spread is the additional interest rate investors demand over a comparable government bond yield to compensate for credit and other risks.

A wider spread generally means borrowers must pay more to access debt financing.

This could increase the cost of building AI infrastructure, particularly for companies that do not have the strong balance sheets of major technology companies.

Why Credit Spreads Matter

When investors become concerned about a company’s ability to repay its borrowing, they may demand higher interest rates.

For AI infrastructure companies, higher borrowing costs can create additional pressure on already expensive construction and computing projects.

If financing becomes too expensive, companies may delay new developments, reduce expansion plans or seek alternative sources of capital.

Reuters also reported that bonds issued by Meta and SpaceX earlier in the year were trading below their original issue prices.

This does not automatically indicate that either company is facing financial distress. It does, however, demonstrate that investors are reassessing the prices of some technology-related debt securities.

Neocloud Companies Face Greater Financing Pressure

The financing slowdown is particularly challenging for neocloud companies and independent data-centre operators.

Neocloud providers specialise in renting computing capacity, often using advanced graphics processing units (GPUs), to companies developing and operating AI models.

Unlike the largest technology companies, many of these businesses have smaller balance sheets and depend heavily on external financing to expand their infrastructure.

They also face risks associated with expensive equipment, customer concentration and changing technology requirements.

Two companies attracting attention in this market are CoreWeave and Nebius.

CoreWeave Faces Investor Concerns

CoreWeave is an AI cloud infrastructure provider that offers access to GPU computing capacity.

Its business has benefited from growing demand for AI computing, but expansion requires substantial spending on data centres, equipment and infrastructure.

Reuters reported that CoreWeave’s shares had fallen approximately 35% year-to-date as investors reassessed risks associated with AI infrastructure financing.

Its bonds have also experienced significant declines in the secondary market.

Construction delays or slower-than-expected demand could affect the company’s ability to generate sufficient cash flow to meet its financial commitments.

However, a falling share price alone does not establish that the company is unable to repay its debt.

Nebius Shelves Planned High-Yield Bond Offering

Nebius, another AI cloud infrastructure company, reportedly decided not to proceed with a planned high-yield bond offering that had been under consideration in September.

The decision reflects the more difficult financing environment facing companies that need substantial external capital to expand their computing infrastructure.

However, Nebius had also entered into a $3.5 billion senior secured loan facility with Citigroup in October.

The facility was intended to support data-centre development, GPU procurement and general corporate purposes.

This distinction is important: the company did not stop borrowing altogether. Rather, it reportedly abandoned one planned bond transaction while retaining access to a separate financing arrangement.

Banks Are Scrutinising Data-Centre Contracts More Closely

Another important development is the increased attention lenders are paying to the contractual arrangements behind AI infrastructure projects.

Many financing transactions involve complicated structures.

A bank may provide financing to an infrastructure company that purchases advanced chips and builds a data centre. A large technology customer may then agree to lease computing capacity for several years.

In some cases, chip manufacturers or major technology companies provide guarantees that support the financing.

These arrangements can reduce certain risks for lenders, but they do not eliminate them.

Banks are now examining several questions before committing additional money:

• Is the customer contract legally binding?
• Can a major customer reduce or terminate its commitment?
• Will the data centre become operational on schedule?
• Are the expected revenues sufficient to cover interest and principal repayments?
• What happens if demand for AI computing grows more slowly than expected?
• Can the equipment be sold or reused if the project fails?

These questions are particularly important because data centres require large upfront investments before they begin generating meaningful revenue.

A delay in construction or a change in customer requirements could affect the entire financing structure.

JPMorgan’s Data-Centre Exposure Highlights the Scale of Lending

The growth of AI infrastructure financing is also changing banks’ commercial real estate exposure.

Reuters reported that JPMorgan Chase had approximately $29 billion in outstanding loans backed by other income-producing properties, a category that includes data centres, at the end of the third quarter.

These loans represented approximately 12% of the bank’s commercial real estate exposure.

Three years earlier, the category represented around 8% of its commercial real estate loans.

This shows how quickly data centres and other income-producing properties have become a more significant part of banks’ lending portfolios.

However, the $29 billion figure covers a broader category of properties and should not be interpreted as JPMorgan’s direct exposure to AI data centres alone.

The growing concentration of lending in the sector is one reason banks are reassessing how much additional risk they want to accept.

Regulators Are Also Watching AI Infrastructure Financing

The concerns extend beyond individual banks and investors.

Central banks and financial regulators are increasingly examining how the AI investment boom is being financed.

The Reserve Bank of Australia highlighted several risks in its October 2026 Financial Stability Review.

These include the growing use of debt, complex financing structures, off-balance-sheet arrangements and the possibility that AI infrastructure investments could fail to generate sufficient returns.

The Bank of England also warned in its July 2026 Financial Stability Report that rapidly growing AI-related borrowing could create risks for financial stability.

One concern is that AI infrastructure companies may borrow through special-purpose vehicles and other arrangements that make the full amount of financial exposure difficult to identify.

Another risk involves the difference between the lifespan of debt and the useful economic life of technology equipment.

For example, a data-centre building may remain operational for decades, but the GPUs and other specialised equipment inside it can become outdated much sooner.

If the equipment loses value before the associated debt is repaid, lenders may face greater losses if a borrower defaults.

These concerns do not mean a financial crisis is inevitable. They highlight the importance of understanding how risks are distributed among banks, investors, technology companies and infrastructure operators.

What Does the Lending Slowdown Mean for the AI Industry?

A more selective lending environment could affect the speed and cost of AI infrastructure expansion.

Companies with strong balance sheets, reliable customers and established revenue streams may continue to access financing.

Smaller infrastructure providers and companies with uncertain cash flows could face higher borrowing costs or difficulty securing new loans.

The impact could extend across several parts of the AI ecosystem.

Industry segment	Potential impact
Hyperscalers	Greater scrutiny of borrowing and infrastructure spending
Neocloud providers	Higher financing costs and more selective lending
Data-centre operators	Greater pressure to prove project profitability
GPU suppliers	Possible changes in equipment financing demand
Private credit funds	More detailed risk assessment before funding projects
AI startups	Potentially more difficult access to infrastructure capacity

The financing slowdown could also encourage AI companies to rely more heavily on internal cash flows, equity investment and partnerships with established technology firms.

For the largest technology companies, the ability to finance AI infrastructure from their existing businesses may provide a significant advantage over smaller competitors.

The Bigger Picture

The AI boom is increasingly becoming a test of the global financial system’s ability to finance large-scale technological infrastructure. Banks and investors initially supported the expansion because of expectations that AI demand would continue growing rapidly. However, rising debt levels, uncertain returns and complicated financing structures are now prompting a more cautious approach.

The lending shift does not signal that investment in AI is ending. Instead, it suggests that access to capital may increasingly depend on a company’s financial strength, customer commitments and ability to demonstrate sustainable cash flows. The companies that can prove their infrastructure investments will generate reliable returns may continue attracting financing, while weaker borrowers could struggle to fund expansion.

Frequently Asked Questions

1. Why are global banks reducing lending to AI companies?

Banks are becoming more cautious because AI infrastructure requires enormous borrowing, while the future revenue and profitability of some projects remain uncertain. Market volatility and concerns about concentrated exposure are also influencing lending decisions.

2. How much debt could AI hyperscalers issue in 2027?

Goldman Sachs estimates that global hyperscaler debt issuance could reach approximately $420 billion in 2027, compared with an estimated $250 billion in 2026.

3. What are neocloud companies?

Neocloud companies provide specialised cloud computing services, often using advanced GPUs to supply computing power to AI developers and businesses.

4. Why are CoreWeave and Nebius facing financing concerns?

Both companies operate in the capital-intensive AI infrastructure market. Their businesses require significant investment in GPUs and data centres, making them sensitive to borrowing costs, customer demand and financing availability.

5. What is a credit spread?

A credit spread is the additional yield investors demand over a comparable government bond to compensate for the risk of lending to a company. Wider spreads generally indicate that investors want more compensation for risk.

6. Are banks completely stopping AI lending?

No. Banks are becoming more selective rather than ending AI financing altogether. Companies with strong cash flows, reliable contracts and credible financial backing may continue to secure loans.

7. What is the biggest risk in AI infrastructure lending?

One major risk is that expensive data centres and computing equipment may not generate enough revenue to repay the debt used to finance them. Rapid technological changes can also reduce the value of specialised equipment.

8. Will the lending slowdown affect AI development?

It could make infrastructure expansion more expensive or slower for companies that depend heavily on external borrowing. However, major technology companies with strong balance sheets may continue investing in AI using internal cash flows and other funding sources.

Looking Ahead

Global banks and investors are likely to remain selective as AI companies seek increasingly large amounts of financing for data centres, advanced chips and computing infrastructure. Lenders will pay closer attention to customer contracts, project completion schedules, debt repayment capacity and the financial guarantees supporting individual transactions. Companies with established customers and stronger balance sheets may continue attracting capital, while riskier infrastructure providers could face more expensive financing.

The long-term direction of AI investment will depend on whether the revenue generated by AI services can justify the enormous spending on infrastructure. If demand and earnings grow as expected, today’s financing concerns could ease over time. However, if returns fall short of expectations, companies with heavy debt burdens could face refinancing pressure, delayed projects and lower valuations. The next phase of the AI boom may therefore be determined not only by technological progress but also by the industry’s ability to generate sustainable financial returns.