Kelsey Bero, a portfolio manager in JPMorgan Chase's asset management division, said Monday that while the global investment-grade bond market is bracing for a busy September issuance calendar, actual demand for AI-related corporate debt suggests fears of a supply glut may be significantly overstated. With anxiety over a potential "AI credit bubble" intensifying and long-dated Treasury yields climbing, JPMorgan's asset management arm is moving to cool market jitters, arguing that while AI capital expenditures are driving a surge in corporate financing, elevated coupon rates are simultaneously attracting retail and institutional capital, creating a dynamic where record supply is met with record demand.
Bero noted that market forecasts for September issuance range widely, from $175 billion to $250 billion. She added that credit portfolio managers generally believe even a $250 billion supply figure "would not be a real problem," but the uncertainty surrounding the final number is complicating preparations across global financial markets ahead of what is traditionally one of the busiest issuance months of the year. The wide spread between the highest and lowest projections—a $75 billion gap—makes it difficult for managers to position their portfolios in advance, and the mere propagation of such a wide forecast range is itself a source of market anxiety.
US blue-chip corporate bond issuance hit a monthly record for the third consecutive month in August, driven by borrowing for artificial intelligence infrastructure spending, extending the fastest issuance pace in market history. So far in August, with two weeks still remaining, issuance has already reached $157 billion, including a $25 billion deal from Google parent Alphabet, along with other large transactions from AbbVie and Advanced Micro Devices (AMD).
Thus, in the view of JPMorgan's asset management arm, even if September investment-grade issuance reaches $250 billion, the market can likely absorb it smoothly, potentially even triggering a rush from sidelined investors. However, this stance refutes the short-term "supply squeeze" risk but does not prove that AI project returns will be sufficient to cover long-term financing costs. The core thesis is that high yields continue to attract retail and off-exchange capital, with record demand capable of absorbing record supply, so the AI financing wave has not yet evolved into a liquidity crisis in the investment-grade bond market, though project profitability and longer-term credit risk remain unverified.
As the borrowing frenzy takes effect, investors are becoming more discerning about which bonds to buy and at what price. Some companies seeking to finance data center projects are increasingly turning to high-yield bond investors to raise billions of dollars, even when the debt itself is investment-grade. Despite concerns about supply pressure, demand has consistently kept pace with record issuance this year. Bero cited high-grade bond flow data showing that retail demand for investment-grade bonds in 2024 has already surpassed the full-year total for any year since 2010. "Despite record supply, demand is also at a record," she said, noting this dynamic is a key reason why September's heavy issuance calendar may be easier to digest than the market fears.
Bero warned that if September's bond supply is absorbed without market disruption, investors still on the sidelines could act quickly. "There will be a wave of capital inflow," she said, adding that any sign of market stability could attract buyers back in force.
According to the latest data compiled by institutions, companies have borrowed more than $410 billion this year for data centers and other AI investments. As of August 24, 2026, a "systemic AI credit bubble burst" is not the base case; a more accurate assessment is that the financing frenzy has moved past the indiscriminate easing phase into a period of credit spread repricing and issuer differentiation. Goldman Sachs, using a broader set of issuers and financing vehicles, estimates AI-related debt issuance is approaching $500 billion, with hyperscale cloud providers issuing approximately $194 billion—only about 40% of the total. While the two sets of figures use different methodologies, they both point to the same conclusion: AI has become the single largest capital theme in global credit markets.
JPMorgan Asset Management's Kelsey Bero believes that even if September investment-grade supply reaches $250 billion, record demand and a blended yield of roughly 5%–6% should still be sufficient to absorb it. However, this only proves the market currently has liquidity; it does not demonstrate that all AI projects will generate cash flows sufficient to cover their debt costs. What truly warrants vigilance is "bond market indigestion," rather than an imminent solvency crisis among big tech companies. Goldman's AI credit basket spread has widened to nearly double its level of about 74 basis points over the past year. AI-related companies account for approximately 40% of this year's issuance of investment-grade bonds with maturities of 15 years or more, and insurer participation in large 30-year orders has fallen by about half compared to the first quarter.
Institutional data shows that as of August 10, AI hyperscale cloud providers had issued approximately $220 billion in bonds this year, compared to just $12.5 billion in the same period last year. Tech bond spreads have reached 89 basis points, about 9 basis points wider than the overall investment-grade market, indicating that investors are reclaiming "blank checks" by demanding higher new-issue concessions, shortening duration, and limiting exposure to single issuers. Therefore, risk is concentrated first in highly leveraged new cloud providers, data center projects lacking committed customers, and special purpose vehicles relying on future token revenue for debt repayment—rather than in cash-rich, low-leverage companies like Alphabet, Microsoft, or Amazon.
For institutions like JPMorgan that hold a more optimistic view of the credit market's prospects, the AI computing infrastructure boom can continue, but it will shift from "expand as long as financing is available" to "structured construction constrained by contracts, cash flow, and capital returns." Hyperscale cloud providers still possess strong operating cash flows and robust AI revenue prospects driven by their cloud dominance. Combined with multiple layers of capital sources—including investment-grade bonds, private credit, infrastructure funds, and project financing—orders across the entire AI computing supply chain, from AI GPUs/TPUs, HBM/DRAM/NAND storage components, optical interconnects, power equipment, data center CPUs, and high-performance networking infrastructure to liquid cooling equipment, are unlikely to suddenly contract sharply over the next two years. However, rising long-term rates and widening credit spreads will force marginal projects to be delayed, scaled back, or require customer prepayments. Selecting infrastructure platforms with investment-grade customers, long-term contracts, prepayment mechanisms, secured power, and verifiable utilization rates will be crucial.
The real early warning signs of an "AI credit bubble" bursting would be persistently shrinking order books, GPU rental prices falling below capital costs, project financing failures, and hyperscalers simultaneously cutting capital expenditures—conditions that have not yet appeared all at once.