It’s interesting how concentrated AI bond issuance is in 2026. Among investment-grade bonds with maturities of 10 years and above, companies related to AI account for nearly 40%, totaling a little over $70 billion. Back in 2022–2025, that proportion was only 5%–10%. This year, a handful of companies—Amazon, Alphabet, SpaceX, Meta, Oracle, and Nvidia—collectively issued $220 billion in debt.
This kind of concentrated supply puts direct pressure on the long-end yields along the yield curve. Historically, when a single sector issues a large amount of debt in the short term, it often raises the term premium—especially when the market is sensitive to duration risk. During the telecom bubble in 2000, a similar phenomenon occurred. A group of companies such as WorldCom and AT&T issued debt aggressively; long-end interest rates were pushed higher, which in turn accelerated the rise in financing costs and the bursting of the bubble.
Now AI companies have huge capital expenditures. Data centers, chips, and power infrastructure all burn cash. Issuing long-dated debt to lock in costs is a rational choice, but concentrated issuance can create a supply shock. If the Fed doesn’t cooperate—or if inflation expectations pick up again—long-end yields could rise further, ultimately turning around to hurt these companies’ valuations and their ability to refinance.
The cycle is straightforward: a large amount of capital floods into a theme → supply surges → yields rise → financing costs increase → profit pressure grows → valuations are repriced. Will this AI cycle end up replaying the telecom bubble script? At the very least, the bond market is already sending signals.
This kind of concentrated supply puts direct pressure on the long-end yields along the yield curve. Historically, when a single sector issues a large amount of debt in the short term, it often raises the term premium—especially when the market is sensitive to duration risk. During the telecom bubble in 2000, a similar phenomenon occurred. A group of companies such as WorldCom and AT&T issued debt aggressively; long-end interest rates were pushed higher, which in turn accelerated the rise in financing costs and the bursting of the bubble.
Now AI companies have huge capital expenditures. Data centers, chips, and power infrastructure all burn cash. Issuing long-dated debt to lock in costs is a rational choice, but concentrated issuance can create a supply shock. If the Fed doesn’t cooperate—or if inflation expectations pick up again—long-end yields could rise further, ultimately turning around to hurt these companies’ valuations and their ability to refinance.
The cycle is straightforward: a large amount of capital floods into a theme → supply surges → yields rise → financing costs increase → profit pressure grows → valuations are repriced. Will this AI cycle end up replaying the telecom bubble script? At the very least, the bond market is already sending signals.