Summary: Artificial-intelligence companies and their infrastructure partners are raising unprecedented amounts of debt to fund data centres, chips, power capacity, and cloud computing. The borrowing wave is now colliding with US Treasury yields near levels last seen in 2007. Strong balance sheets protect the largest technology groups, but higher funding costs and increasingly selective bond investors could expose weaker projects and heavily leveraged operators.
AI moves from a cash-flow story to a credit-market story
For much of the artificial-intelligence investment boom, equity markets supplied the dominant narrative. Investors focused on semiconductor demand, cloud revenue, model adoption, and the capital expenditure of the largest technology companies.
Credit markets are now becoming just as important.
Global bond issuance by AI-linked companies has exceeded $400 billion in 2026 and is running at an annualised pace above $500 billion, according to Institute of International Finance figures reported Sunday. US borrowers account for approximately 90% of that total.
Meanwhile, US companies issued roughly $1.9 trillion of bonds through August, about 30% more than during the same period in 2025. AI investment has become a significant contributor to that supply. Yahoo Finance
The scale could grow further. Goldman Sachs estimates that Alphabet, Amazon, Meta, Microsoft, and Oracle will collectively spend around $800 billion on capital expenditure this year and $1.2 trillion next year. Not all of that spending is AI-specific, but data centres, computing hardware, and supporting infrastructure represent a substantial share.
Why the financing requirement is accelerating
Building AI capacity is unusually capital intensive. Companies must fund land, buildings, advanced chips, cooling equipment, network connections, and long-term electricity arrangements well before the associated revenue is certain.
The largest hyperscalers can initially finance much of this investment from operating cash flow. But the scale and speed of construction increasingly make external financing attractive, even for companies that historically relied relatively little on debt.
S&P Global estimates that hyperscaler AI-related capital expenditure has risen from about $80 billion in 2019 to more than $700 billion in 2026. Technology companies now account for 27% of US investment-grade corporate bond issuance and 20% of high-yield issuance, according to its September analysis. S&P Global Market Intelligence
The financing extends beyond household-name technology groups. Data-centre owners, specialist cloud providers, utilities, private-equity vehicles, and project-finance structures are also borrowing against expected AI demand. That broadens the opportunity set for investors, but it also moves more execution and demand risk into credit portfolios.
Higher benchmark yields change the calculation
This borrowing wave is arriving during a difficult period for fixed-income markets. The 10-year US Treasury yield ended last week near 5.2%, around its highest level since 2007. Because Treasury yields form the base rate for much corporate borrowing, companies must generally offer investors an additional spread above an already elevated benchmark.
The Federal Reserve reinforced the higher-rate environment on September 16, raising its policy range by 25 basis points to 3.75%-4.00%. The central bank cited elevated inflation alongside resilient spending and robust capital investment. Federal Reserve
This does not necessarily stop well-capitalised companies from borrowing. It does, however, raise the return a new AI project must produce to justify its financing cost. Projects that looked attractive with cheaper capital may appear less compelling when debt has to be refinanced or newly issued at significantly higher yields.
The market is beginning to differentiate
Bond investors have not rejected AI debt as a category. Instead, they are becoming more selective.
Reuters reported last week that buyers were demanding larger concessions on some AI-related bonds while showing stronger demand for conventional industrial and financial issuers. Goldman Sachs expects gross hyperscaler debt issuance to reach a record $420 billion in 2027, approximately 60% above its 2026 estimate. Reuters
Credit quality varies considerably. The largest technology companies generally possess substantial cash generation, diversified businesses, and investment-grade ratings. Their central risk is less likely to be near-term default than weaker returns on enormous capital commitments or a gradual deterioration in credit metrics.
Specialist infrastructure businesses present a different profile. CoreWeave, for example, reported $35.6 billion of future principal payments on debt at June 30 and $640 million of net interest expense during the second quarter. Its revenue is expanding rapidly, but the figures demonstrate how capital intensity can translate into substantial fixed obligations. CoreWeave SEC filing
Some borrowers must also accept notably expensive funding. SoftBank’s recent $11.1 billion junk-bond transaction reportedly included a seven-year tranche yielding as much as 9.75%. Those costs place a high hurdle on the investments being financed.
What it means for investors
The first implication is that AI exposure is spreading beyond equity portfolios. Investors may own technology and AI risk through corporate-bond funds, private credit, infrastructure vehicles, utilities, and structured finance, even when their direct technology-equity allocation appears limited.
Second, bond supply itself matters. A sustained flood of high-quality technology debt can force issuers to offer more attractive terms, compete with other companies for capital, and increase the technology sector’s weight in corporate-bond benchmarks.
Third, company selection is becoming more important. Investors should distinguish between cash-rich platforms using debt opportunistically and borrowers that depend on continuous refinancing or aggressive future utilisation assumptions. Relevant measures include free cash flow after capital expenditure, fixed versus floating-rate debt, maturity schedules, customer concentration, contractual revenue, and interest coverage.
Finally, credit markets may provide an earlier signal than equity prices. Wider spreads, weak new-issue demand, or tougher loan terms could indicate that investors are becoming less confident in project economics before the concern appears fully in share valuations.
