Goldman Sachs Says AI Junk Bond Spreads Hit Record High as Investors Reassess Risk

Goldman Sachs has warned that borrowing conditions for companies associated with artificial intelligence are becoming increasingly challenging as investors reassess the risks of the AI investment boom. According to reporting published on October 17, 2026, spreads on US dollar-denominated, high-yield AI-related debt reached a record 360 basis points above US Treasury yields, following a sharp increase during the week. The widening reflects growing investor concern about the amount of debt being accumulated to finance data centres, computing infrastructure and advanced AI systems. Although demand for AI technology remains substantial, investors are increasingly questioning whether the revenues generated by some infrastructure projects will be sufficient to justify their financing costs. The development represents another sign that financial markets are becoming more selective about which AI-related businesses they are willing to support.

Bond spreads measure the additional yield investors demand for holding a company’s debt instead of a relatively safer government bond. A spread of 360 basis points means investors are demanding an additional 3.6 percentage points in annual yield over the relevant Treasury benchmark. When these spreads rise, companies generally face higher borrowing costs, making it more expensive to finance expansion, refinance existing obligations or fund new projects. The latest increase comes amid a broader rise in interest rates and a growing supply of corporate debt linked to AI infrastructure. Goldman Sachs previously reported that investors were becoming more cautious about AI-related borrowing as technology companies and data-centre operators brought large volumes of new bonds to the market. The concern is not necessarily that every AI company is close to default, but that the sheer amount of financing required could strain investor demand.

Data-centre operators and so-called neocloud companies have been particularly exposed to the recent market pressure. These businesses invest heavily in specialised chips, servers, electricity infrastructure and computing facilities that support AI training and inference. CoreWeave, a major provider of cloud computing infrastructure, has experienced a sharp deterioration in investor sentiment as its shares and debt came under pressure. Reuters also reported that credit-default-swap spreads linked to CoreWeave moved above 1,000 basis points, while spreads for Oracle and SoftBank rose to record or near-record levels. Credit-default swaps are financial contracts used to hedge against the risk of a borrower failing to meet its debt obligations; rising premiums indicate that the market is assigning greater risk to that possibility. These indicators do not prove that a company will default, but they show that investors are demanding substantially more protection against potential losses.

One of the main questions facing the AI industry is whether the enormous spending on computing infrastructure will generate sufficient returns over time. Technology companies are investing billions of dollars in data centres and AI chips, while developers continue to build increasingly capable models and AI-powered services. However, the financial benefits may take years to materialise, and some facilities depend on optimistic assumptions about future demand, pricing and utilisation. Rising interest rates add to the pressure because debt becomes more expensive to service and refinance. Analysts have also raised concerns about complex financing arrangements and the extent to which companies depend on a continuing expansion of AI spending. If demand grows more slowly than expected, some heavily indebted operators could face difficulties meeting their obligations or attracting fresh capital.

Despite these concerns, the widening spreads do not mean that the entire AI industry is facing an imminent financial collapse. Large technology companies such as Amazon and Meta retain substantial businesses and stronger credit profiles than many smaller infrastructure providers. The current market adjustment appears to be placing greater pressure on companies with high borrowing requirements, uncertain future cash flows and limited financial flexibility. For investors, the distinction between profitable AI businesses and companies relying heavily on debt-financed expansion is becoming increasingly important. For the industry, the challenge will be to demonstrate that investments in chips, data centres and AI services can produce sustainable returns rather than depending indefinitely on rising spending. As bond markets continue to reassess risk, financing costs may become a more significant factor in determining which AI companies can expand successfully and which may be forced to slow their plans.