The AI financing boom has heightened concerns about credit risk. Why hasn't soaring US bond yields boosted corporate bond spreads?

The Zhitong Finance App notes that despite the recent sharp rise in US Treasury yields, the investment-grade corporate bond market has not simultaneously released clear credit risk alerts. The yield on 10-year US bonds rose from 3.96% in February 2026 to 5.31% in early October. At one point, it hit a high level since 2002, but the credit spread on investment-grade corporate bonds did not rise but fell during the same period. This contrast shows that bond investors currently do not generally require companies to pay more additional compensation for credit risk.

However, in a context where artificial intelligence (AI) infrastructure construction continues to expand and the financing needs of technology companies continue to increase, credit spreads operating at low levels do not mean that corporate financing risks have disappeared. As the scale of investment in AI data centers, chips, and supporting infrastructure expands, more and more technology companies are beginning to use bonds, bank loans, and other financing arrangements to support capital expenses. When the high interest rate environment is combined with the capital requirements for AI investment, whether enterprises can turn huge capital expenses into sufficient cash flow will become an important variable affecting future credit risk.

US bond yields have risen sharply, so why are interest spreads on corporate bonds narrowing instead?

Judging from recent market data, the rise in US Treasury yields did not trigger a simultaneous increase in credit spreads on investment-grade corporate bonds.

According to ICE data, the options adjusted spread (OAS) of the US investment grade corporate bond index fell from 86 basis points in February 2026 to 84 basis points in September, and fell further to 82 basis points on October 8. Meanwhile, the 10-year US Treasury yield rose to 5.31% in early October and hit a higher level at one point. There is a clear contrast between the two sets of data: as an important benchmark for corporate financing pricing, US bond yields have risen sharply, yet the additional credit risk premium that investment-grade companies need to pay compared to US bonds has declined.

To understand this phenomenon, we first need to distinguish between market interest rates and credit spreads.

The financing cost of corporate bonds is usually determined by a combination of the base interest rate and the credit spread. The rise in US bond yields means an increase in the benchmark financing costs faced by companies when issuing new bonds; credit spreads reflect the additional compensation investors require when holding corporate bonds rather than comparable US bonds, covering expected default losses, liquidity, and other risk factors.

Therefore, a narrowing of credit spreads does not equal a reduction in corporate financing costs. Assuming that the yield on US bonds rises by 100 basis points and the credit spread on corporate bonds narrows by 20 basis points, the yield on newly issued corporate bonds will still rise by about 80 basis points without changing other conditions.

In other words, the recent performance of corporate bond spreads shows that investors are not asking investment-grade companies to pay more credit risk compensation, but this does not offset the pressure on financing costs brought about by rising benchmark interest rates.

So why have US bond yields continued to rise, yet credit spreads haven't widened?

One explanation is that rising interest rates don't necessarily mean that the economy is getting worse. If rising yields reflect improved economic growth expectations, strong investment activity, or rising inflation expectations, corporate profitability and solvency may remain resilient. In particular, in an environment where economic growth is relatively steady, investors do not necessarily need to drastically increase risk compensation for corporate bonds.

Inflation may also play a role. For enterprises holding fixed-rate debt, if income and nominal cash flow can rise with the price level, the actual repayment burden of existing debt may be reduced. However, this effect is not inevitable: if costs rise faster than income, or if high inflation forces interest rates to stay high for a long time, financial pressure on companies may still increase.

Furthermore, corporate debt often has different maturity times, and not all debts are immediately repriced as market interest rates change. For fixed interest rate bonds that have already been issued, the coupon interest rate will not automatically increase due to rising US bond yields. Therefore, the impact of changes in market interest rates on an enterprise's overall interest expenses often gradually becomes apparent as debt matures and refinancing.

The wave of AI infrastructure financing is changing the capital needs of enterprises

If the traditional investment-grade corporate bond market has yet to significantly increase the credit risk premium, then the financing needs brought about by AI infrastructure construction are adding new variables to the future credit market.

As large technology companies speed up the construction of AI data centers, procure high-performance chips, and expand cloud computing capabilities, AI investment is no longer just R&D expenses or software development expenses, but increasingly involves land, electricity, servers, network equipment, and long-term infrastructure construction. These projects often require large-scale, continuous capital investment, and the payback period may also be long.

Some media quoted Morgan Stanley data as reporting that the estimated AI-related debt financing scale for technology companies will reach 466 billion US dollars in 2026 billion, which is significantly higher than the 101 billion US dollars in 2025. At the same time, AI-related debt financing activities have cooled down after the previous sharp expansion: in September 2026, the issuance of relevant debt fell to about 23 billion US dollars, less than half of August; the US investment-grade market also did not issue AI-related bonds in that month, while the total number of AI-related bonds issued in this market from January to August reached 306 billion US dollars.

These figures reflect the deepening impact of AI infrastructure construction on the capital market, and also show that investors are beginning to pay more attention to financing scale, project returns, and potential leverage risks.

In the past, tech giants were able to rely more on operating cash flow to support cloud computing infrastructure construction, but as the scale of AI investment rapidly expands, even large enterprises with sufficient cash flow may need to make more use of bonds and other financing instruments. For AI companies and related projects that have not yet formed a stable profit model, the importance of external financing is even more prominent.

On October 12, Reuters reported that Tencent plans to raise at least 4.5 billion US dollars through US dollars and offshore RMB bonds. The company's capital expenditure more than doubled year-on-year in the first half of 2026 to 96.3 billion yuan, reflecting that large technology companies are increasing their investment in AI. However, the financing plan still depended on market conditions and investor needs at the time, and cannot yet be considered a completed offering.

This also means that the impact of AI investment is extending from technology companies' capital expenditure budgets to the bond market. As the scale of financing expands, investors not only need to evaluate the company's current balance sheet, but also determine whether future investments can bring sufficient income and free cash flow to cover continuously increasing capital requirements.

The risk is not only how much money you borrow, but also whether the return on investment can be realized

The credit risk that may be brought about by the expansion of AI financing is first reflected in the mismatch between capital expenditure and cash flow.

Data center construction requires a large upfront investment, and servers and AI chips also require continuous updates, but whether related investments can generate sufficient returns depends on factors such as the progress of commercialization of AI applications, customer demand, computing power utilization, and service pricing. If a company's investment continues to grow faster than operating cash flow, it may be necessary to fill the capital gap through additional debt, equity financing, or other financing arrangements.

Second, some AI infrastructure investments involve complex leasing arrangements, project financing, and long-term procurement commitments. Simply looking at the debt figures on a company's balance sheet may not be enough to comprehensively measure its future financial obligations.

Regarding the potential leverage of AI hyperscale cloud service providers, Meta's future leasing commitments reached US$347 billion, while its balance sheet debt was approximately US$84 billion. Meanwhile, sell-side analysts expect the company's capital expenditure up to 2028 could exceed $500 billion.

Leasing commitments are not equivalent to debt already recorded on the balance sheet, and future capital expenditure projections are not equivalent to actual financing requirements. However, these figures suggest that when evaluating the financial risks of AI infrastructure construction, in addition to book loans, they also need to pay attention to leasing obligations, long-term contract commitments, and future cash expenses.

Third, high interest rates may gradually unfold the already manageable financing pressure. For companies that have already issued fixed-rate bonds, rising market interest rates will not immediately change their coupon interest, but financing costs may gradually rise as old bonds expire, new bonds are issued, or loans are repriced. If the company is still at the peak of capital expenditure at that time, or if the cash flow generated by AI projects falls short of expectations, solvency may be squeezed.

What is more noteworthy is that there are close financial and business ties between different enterprises in the AI industry chain. Chip vendors, cloud computing service providers, data center operators, and AI model developers may be linked through procurement contracts, cloud service agreements, leases, and financing arrangements. Once some projects are delayed, customer demand falls short of expectations, or financing conditions are tightened, pressure may spread along the industrial chain.

However, this does not mean that the expansion of AI investment will inevitably lead to a credit crisis. Large technology companies usually have stronger cash flow, financing channels, and market positions, which can cushion financial pressure to a certain extent. The key point is that investors need to distinguish between companies with abundant cash flow and mature business support, and projects that rely on continuous external financing and where there is still great uncertainty about revenue fulfillment.

Low credit spreads do not mean that the market has ruled out AI risks

Judging from the current interest spreads on investment-grade corporate bonds, the market has yet to generally require companies to pay more credit risk compensation. However, this indicator reflects overall market pricing and does not fully reflect the risks faced by all companies, industries, and financing structures.

On the one hand, the investment-grade corporate bond index is composed of different issuers, and overall interest spreads may mask changes in individual companies' credit conditions. On the other hand, credit risk is generally not immediately reflected in default data simply due to rising market interest rates. Businesses may respond to pressure by saving cash, reducing other expenses, or adjusting capital expenses before gradually facing the risk of rising refinancing costs or downgrading credit ratings.

Furthermore, there may also be differences in the risk performance of investment-grade bonds and high-yield bonds. Businesses with weak credit qualifications, unstable cash flow, or high levels of leverage are generally more vulnerable to tighter financing conditions. Therefore, to determine whether AI financing is becoming a source of risk in the credit market, it is necessary not only to look at interest spreads on overall investment-grade corporate bonds, but also to observe changes in interest spreads, debt issuance conditions, financing needs, and corporate cash flow performance of issuers with different ratings.

For AI-related companies, indicators worth focusing on in the future include whether capital expenditure continues to grow faster than operating cash flow growth, whether new debt and leasing commitments have increased significantly, whether AI-related revenue can cover infrastructure investment, and whether bond issuance needs to provide higher yields to attract investors.

If the return on AI investment falls short of expectations, financing demand continues to rise, and corporate bond issuance conditions deteriorate in the future, credit spreads may widen again. At that time, the pressure on companies will not only be increasing benchmark financing costs due to rising US bond yields, but may also include rising additional financing costs caused by investors demanding higher risk compensation.

At the end of the day, the spread between US bond yields and corporate bonds reflects different levels of risk. Interest spreads on current investment-grade corporate bonds remain low, indicating that the market has not generally raised the pricing of corporate credit risk for the time being; however, as the scale of AI infrastructure investment expands, the relationship between debt, leasing commitments, and future cash flow will become more important. For bond investors, what they really need to be wary of is not simply an increase in the scale of financing, but whether companies can turn increasing AI capital expenditure into a continuous cash flow sufficient to support debt repayment and reinvestment in an environment with high interest rates.