AI in action · Markets
Morgan Stanley Says AI Giants' Bonds Now Look Cheap
Morgan Stanley sees the best risk-reward of 2026 in top hyperscaler bonds after $67 billion of issuance since September 30, while neocloud debt stays volatile.
While AI stocks swung through the week, the bond market quietly asked a different question: who can afford to keep paying for the build-out? Morgan Stanley's answer, in a note by strategist Fernanda Lima on November 10 reported by CNBC, is that the strongest borrowers now look like a buying opportunity, and the weakest still do not.
The supply has been heavy. Since September 30, Microsoft, Meta Platforms and Oracle have issued new bonds in the US, and Alphabet and Meta have also borrowed in Europe, about $67 billion across both markets, according to Lima. Spreads on most of those names have tightened since the deals priced; Oracle was the exception, with its spreads widening. Hyperscaler bonds trade near their widest levels of the year, at prices consistent with A and BBB ratings, and Lima called the risk-reward "as good as it's been all year."
The other end of the market looks very different. CoreWeave, the neocloud that rents out Nvidia chips, saw yields on its bonds top 16% earlier in November, making it the market's main example of AI credit stress, even before a third-quarter beat lifted its shares 25% on November 11. Morgan Stanley expects more spread volatility for data centre bonds and neoclouds, and a widening gap between the financing terms available to high-quality and lower-quality borrowers.
The bank's forecasts explain why this matters beyond 2026. It expects little or no new AI-related issuance for the rest of this year, but projects AI capital spending of $1.66 trillion in 2027, up 75% from 2026, and $1.77 trillion in 2028. Some projects may slip after Democrats won Congress and voters backed local limits on data centres, but Lima said the leading hyperscalers "will remain well capitalized" and could use more equity issuance to bridge the gap between spending and revenue.
The message for anyone holding AI exposure through credit funds is that the label hides very different risks. A bond from a cash-rich platform company and a bond from a leveraged data centre builder can both be called AI debt, but the first is priced like a solid investment-grade issuer and the second like a high-yield bet on future rental income. The Federal Reserve made a similar point in its Financial Stability Report on November 13.
Sources: CNBC — Stick with high-quality AI bonds as debt worries persist, says Morgan Stanley (Nov 11, 2026)