AI债务对国债收益率的驱动局限性

The Limits of AI Debt as a Driver of TreasuryYields As long-term Treasury yields reach multiyear highs, a deluge of AI-related bond issuance has been cited as acontributing factor.The case is based on supply and demand: As more AI debt is issued, the overall supply ofdebt increases, and yields on all bonds must rise to attract sufficient demand. This effect is strongest if AI debtcompetes directly with Treasurys for investor dollars. This may be the case for some investors, since the spreadson the highest-rated AI issuers are so low. But since AI issuance accelerated in the fall of 2025, spreads onmany hyperscaler corporate bonds have started to creep toward more typical investment-grade levels. As AIdebt becomes less like Treasury debt, the direct-competition channel weakens.12 On the other hand, even when AI debt does not compete directly with Treasurys, the sheer scale of AIhyperscaler debt could still overwhelm the debt markets and lead to higher yields for all forms of debt.AIinfrastructure spending is projected to accelerate further in 2027 and continue at least through 2030.Twofactors could dampen future issuance, however. First, rising spreads make debt more costly for thehyperscalers. Second, banks may be approaching the limit of the amount of hyperscaler business they cansupport. Banks buy CDS protection to stay under issuer-level exposure limits, and the cost of buying CDSprotection on most hyperscalers has risen even more than bond spreads.345 These market dynamics point toward less direct supply pressure from AI debt and a more constrained path forfuture issuance, which may limit its contribution to Treasury yields going forward. The author thanks László Arany for his contribution to this post. Trends in AI hyperscaler bond and CDS may limit their impact onTreasury yields Loading chart...Please wait. Michael HayesExecutive Director, MSCI Research &Development Subscribe todayto have insights delivered to your inbox. Enter your email address Subscribe Fixed Income Offerings Our holistic approach — integrating indexes, analytics and a robust climate/sustainability framework — sets the stagefor portfolio differentiation. Learn more Scenario Analysis: Tracing Credit Stress Across Private and PublicMarkets Two stagflationary scenarios — private-credit contagion and hyperscaler repricing — each produce ~11% portfoliolosses but through different channels. Identifying which dominates matters more than the headline loss. Read more Mapping AI Exposure Across Global Markets AI's geography is more distributed than headlines suggest — we map which economies lead each layer of the valuechain and what that means for global equity investors. Explore more 1 Davide Barbuscia, Ye Xie, and Michael MacKenzie, “AI Is Driving Up Treasury Yields: ‘It Just Touches Everything,’”Bloomberg, Aug. 17, 2026. 2 For many investors, general investment-grade corporate debt does compete directly with Treasurys, such as those invested inan aggregate bond index. In these cases, there is reason to expect that AI debt has an especially strong effect on supply. A lotof investment-grade issuers, such as banks, prefer floating-rate debt, so they use swaps to convert their fixed-rate bonds tofloating. This means that their debt issuance is duration-neutral. In contrast, AI issuers, like utilities, prefer fixed-rate debt, sothey retain the fixed-rate exposure. Furthermore, if they issue private debt, which is typically floating, they use swaps to convertthe floating-rate debt to fixed. This means that AI issuers are creating (public) duration supply from both their private andpublic issuance. Hugo De Vere, Srini Ramaswamy and Seth Searls, “How AI debt financing impacts duration supply and interestrates,” Dallas Federal Reserve, Feb. 10, 2026. 3 In theory investors should invest in all available debt in proportion to its risk-adjusted yield, which would mean that newsupply of any new debt could draw investment dollars away from other forms of debt. Although there are many frictions thatlimit investment across the credit and duration spectrum, some supply effect is always expected, especially when the magnitudeof issuance is high. 4 Lucas Baynes, “The AI buildout comes to the bond market,” Vanguard.com, Aug. 19, 2026. 5 Caleb Mutua and Tasos Vossos, “Hyperscaler Debt Flood Brings Derivatives Bonanza,” Bloomberg.com, May 23, 2026. Herewe examine the five current hyperscaler constituents of the S&P Global CDX North America Investment Grade Index: Alphabet,Amazon, Meta, Microsoft and Oracle, which tend to be the most liquid. Notably Alphabet, Meta, and Microsoft were addedduring the latest index revision in March 2026, reflecting strong demand for protection on those issuers. The content of this page is for informational purposes only and is intended for institutional professionals with the analyticalresources and tools necessary to interpret any performance information. Nothing herein is intended to recommend anyproduc