BNP Paribas Warns AI Debt Could End the Corporate Credit Bull Market

AI’s infrastructure buildout is increasingly becoming a bond-market story. BNP Paribas warns that the scale of borrowing by major cloud and technology companies could break the multi-year rally in corporate credit—not because hyperscalers are suddenly weak borrowers, but because their debt issuance may overwhelm the supply conditions that supported unusually tight credit spreads. 3

The warning: a shift from bond scarcity to bond abundance

BNP Paribas expects hyperscalers to sell about $400 billion of bonds next year. It also projects a record $3.7 trillion in net fixed-income supply across the market. The concern is that corporate bonds can no longer command scarcity-driven valuations when investors must absorb a much larger pool of government and corporate debt. 10

In this scenario, investors need more compensation to buy the additional paper. That compensation appears as higher bond yields and wider credit spreads—the extra yield companies pay above comparable government bonds. BNP expects euro investment-grade spreads to widen by about 6 basis points and U.S.-dollar investment-grade spreads by about 7 basis points by the end of 2026.

Why AI capex changes the credit-market equation

Large technology companies historically funded much of their investment from operating cash flow. The AI and data-center buildout has made bond financing a more important part of the funding mix. JPMorgan Asset Management estimates that hyperscalers could issue $300 billion of investment-grade bonds over the next year, while the broader data-center buildout could require roughly $2 trillion of investment-grade financing through 2030. 813

That matters because the issuers are large enough to influence the broader investment-grade market. JPMorgan estimates hyperscaler issuance could rise to 9% of U.S. dollar investment-grade issuance in 2026, up from 2% between 2022 and 2024.

The result is a market driven less by a shortage of high-quality corporate paper and more by a question of price: what yield and spread will persuade investors to absorb the additional duration and credit exposure?

Evidence that investors are becoming more selective

The repricing BNP anticipates is not purely theoretical. The European Central Bank reported that strong demand for euro-denominated hyperscaler bonds weakened after mid-2026: deal cover ratios began to decline and investors sought more compensation for the risks associated with higher expected AI capital expenditure and issuance.

In the U.S., AI hyperscaler debt issuance had reached $220 billion by August 10, according to BNP Paribas data cited by Reuters. Reuters also reported signs of investor fatigue as the wave of AI-related borrowing tested demand. 2

JPMorgan Asset Management found that hyperscaler spreads had widened by around 30 basis points year to date, compared with 2 basis points for the broader U.S. investment-grade index. On a duration-adjusted basis, it put hyperscaler spreads at roughly 105 basis points, versus about 80 basis points for the broader market.

Other forecasts point to the same supply challenge

The precise estimates differ, but other market participants see an unusually large financing cycle ahead:

• Morgan Stanley estimates total AI- and hyperscaler-related issuance could reach as much as $400 billion in investment-grade debt, plus $65 billion in high-yield debt and loans. 11
• Goldman Sachs expects major hyperscalers’ bond issuance to keep rising as companies fund AI investment, while Reuters reported that investors were becoming more selective about purchasing the growing supply. 6
• JPMorgan Asset Management estimates the broader data-center ecosystem could require roughly $5 trillion through 2030, including approximately $2 trillion financed in investment-grade credit markets. 13

These estimates do not establish that a credit downturn is inevitable. They do show why issuance volume, rather than only company fundamentals, has become central to the outlook for corporate bonds.

What a “late-cycle” credit market would mean

BNP’s thesis can be understood as a potential transition from a mid-cycle environment—where limited supply and tight spreads support corporate bonds—to a later-cycle environment, where heavy borrowing requires higher yields and larger risk premiums to attract buyers.

That pressure can extend beyond technology. If investors devote more capital to very large, highly rated hyperscaler deals, smaller or lower-rated borrowers may need to offer more attractive pricing to compete. The effect would be wider financing costs across parts of the corporate market, even if the original supply surge begins with AI infrastructure.

The key caveat: this is a valuation and absorption risk

The warning is not that hyperscalers are automatically poor credits. Their scale, profitability and access to capital may still allow markets to absorb substantial issuance. BNP itself said investors can absorb supply in the short term, though it expects it to become a larger issue as borrowing continues. 3

The decisive variables are the pace of new issuance, the durability of AI-related earnings and capital spending, government borrowing needs, and whether investor demand rises alongside supply. If demand remains strong, the market may adjust without a severe disruption. If it does not, AI debt could mark the end of the unusually tight spreads and supply scarcity that defined the corporate-credit bull market. 3