Michael Hartnett argues AI spending has pushed hyperscaler bond yields high enough to beat the equity story
The most interesting trade in big tech right now might not involve buying a single share. According to Bank of America’s Michael Hartnett, the better opportunity sits in the debt that mega-cap tech companies are issuing to pay for artificial intelligence.
The reason is simple. Investors have grown nervous about how much these firms are borrowing for the AI buildout, and that nervousness has pushed yields up. Even investment-grade hyperscalers are now paying as much as 8%.
What Hartnett is actually saying
Hartnett, Bank of America’s chief investment strategist, laid out the case in the firm’s Flow Show weekly note. The research behind it is dated October 29, 2026.
His core point: long-duration bonds from companies like Oracle, Meta, and Alphabet now yield somewhere between 6% and 8%. That range, he argues, makes them more attractive than the same companies’ stocks.
The individual figures are striking. Oracle bonds were reported at 8.4%, Meta at 7.5%, and Google at 6.9%.
That is exactly what has happened. US investment-grade tech bonds have seen prices drop approximately 9% year-to-date. Their total returns relative to the broader index have hit a 30-year low.
So the sector has been one of the bond market’s worst neighborhoods this year. Hartnett is effectively arguing that the bad year has created a good entry point. He described these bonds as “diamonds in the rough.”
Why yields climbed: the AI spending bill
Hartnett’s projections put hyperscaler capital expenditure at around $240 billion in 2024. That figure is expected to rise to $1.2 trillion by 2027.
Spending at that pace eats into cash. Free cash flow for these firms is predicted to swing from +$240 billion to -$180 billion by 2027.
When spending outruns cash flow, the gap has to be financed somehow. Increasingly, the answer is the bond market.
Hyperscaler bond issuance is expected to reach $182 billion in 2026. In 2024, that figure was just $17 billion.
The case for bonds over stocks
Hartnett’s argument comes down to where the risk lands. Stockholders own the upside if AI pays off, but they also wear the downside if the spending fails to generate returns.
Bondholders sit higher in the pecking order. They get their coupons as long as the company stays solvent, and these are some of the most profitable companies on the planet.
With yields now in the 6% to 8% range, that steady payout looks more competitive.
Hartnett is also cautious on markets more broadly. He advises a risk-off stance through mid-November 2026, citing rising bond yields and potential geopolitical risks. Key risks around equity exposure, in his view, remain present.
He is not simply pitching tech bonds as a bargain. He is pitching them as a way to stay exposed to the AI theme while stepping back from the more volatile end of the capital structure.
What this means for investors and the AI trade
The second implication concerns the companies themselves. Borrowing at 7% or 8% is meaningfully more expensive than tapping cheap credit or internal cash. If issuance keeps climbing toward the projected $182 billion, financing costs become a bigger line item in the AI math.
Oracle stands out as the name to watch. Its 8.4% yield is the highest among the examples cited, which signals the market sees it as carrying more financing strain than Meta or Google.
The spread between those three companies tells its own story. Investors are not treating all hyperscalers as interchangeable credits, and they are differentiating based on how much each firm leans on debt.
Hartnett’s risk-off call runs to mid-November 2026, so the coming weeks will test whether higher yields attract buyers or prompt more selling.
For now, Bank of America’s message is clear enough. The AI buildout has made big tech’s balance sheets look more like those of capital-hungry industrial giants. Investors who want a piece of that story may find the debt pays them better, and more reliably, than the stock.
