BNP says AI borrowing will end the credit bull market, and it is lending to AI too

BNP Paribas expects the years-long corporate-bond rally to break as hyperscalers issue roughly $400B of bonds next year, contributing to a record $3.7T of net fixed-income supply. The ECB similarly found that US technology companies account for nearly 10% of new euro-denominated issuance, underscoring how AI-related infrastructure spending could pressure corporate credit spreads and bond valuations.

Analysis

The key transmission is not simply wider credit spreads: sustained AI-capex financing shifts the marginal buyer from duration-sensitive institutional accounts toward price-sensitive credit funds. That raises the all-in cost of capital for META, GOOGL, AMZN, MSFT and ORCL just as depreciation, power contracts and data-center lease commitments are increasing fixed-cost intensity. Even modest spread widening matters disproportionately for ORCL and AMZN, whose incremental infrastructure returns are less proven and whose valuation support relies more on forward free-cash-flow conversion than MSFT’s.

Near term (days to weeks), a broad selloff is unlikely without a weak Treasury auction, higher real yields, or an announced jumbo deal priced at a material concession. Over 1-3 months, monitor new-issue concessions on long-dated tech bonds and the CDX.NA.IG index: a persistent 10-15bp deterioration would indicate that issuance is no longer being absorbed cleanly. The second-order loser is lower-rated data-center infrastructure—Digital Realty (DLR), Equinix (EQIX), CoreWeave-linked financing, utilities with large capex plans—because hyperscaler borrowing can crowd out the same long-duration credit buyer base.

Consensus may overstate the immediate threat to the highest-quality issuers: MSFT, GOOGL and META retain substantial liquidity and can moderate buybacks rather than fund all capex with debt. The more actionable asymmetry is dispersion within technology credit and equity multiples, not a blanket short of investment grade. A reversal would be signaled by capex guidance moving toward internally funded levels, falling real yields, or repeat technology deals clearing with less than 5bp concessions.

For BNP Paribas, the direct effect is mixed rather than directional: larger primary-market activity supports underwriting and trading revenue, but a disorderly spread move raises mark-to-market and financing-risk sensitivity. The stock is therefore a poor pure expression of the thesis absent evidence that European credit-market stress is widening bank funding spreads.

Key Decisions for Investors

• Initiate a 3-6 month relative-value position: long MSFT / short ORCL, sized beta-neutral. MSFT has greater balance-sheet capacity to absorb AI capex; ORCL is more exposed to a higher financing-cost/FCF-conversion debate. Reassess if ORCL demonstrates accelerating cloud margins or if the relative spread widens by more than 15%.
• Buy 3-6 month CDX.NA.IG protection or use a modest short in LQD as the liquid macro hedge, but enter only after CDX.NA.IG widens 8-10bp from current levels or a major hyperscaler deal prices at a 10bp+ concession. Target a further 15-25bp widening; stop if spreads retrace below the post-entry level after successful deal absorption.
• Underweight DLR and EQIX versus the S&P 500 over the next quarter; their refinancing and development economics are more duration-sensitive than hyperscaler equities. Cover if long-end Treasury yields decline 40bp or if managements demonstrate unchanged development yields despite higher funding costs.
• Set an event-driven alert for debt issuance, capex, and buyback guidance from AMZN, GOOGL, META, MSFT and ORCL. A meaningful move from buybacks to debt-funded capex is the confirmation signal; without it, treat the supply thesis as a watch item rather than a broad risk-off trigger.