When will AI stocks be unlocked?

👆Click the blue text to follow us

BofA's Hartnett

There was once a boss who used to have so much cash that he couldn't spend it all. Every year he would invite shareholders to the annual meeting and give out red envelopes. Later, he spotted a new opportunity and invested more each year, quickly running out of money and having to borrow. But nowadays, borrowing costs have been rising year after year, and yet investment can't be stopped, so the boss's pressure keeps growing. This story is now happening to the most valuable companies on the U.S. stock market.

BofA Chief Strategist Hartnett points out in his latest report that the conflict between AI and the bond market is intensifying, and that the real period of relief for tech stocks may not come until the first half of next year.

According to BofA data, the capital expenditures of cloud giants are expected to reach $1.2 trillion by 2027, five times what they were two years ago. In contrast, free cash flow will turn negative, projected at negative $180 billion by 2027. This gap can only be closed by borrowing, so these companies are issuing a large amount of debt. In 2024, they only issued $17 billion in bonds, but this year, it has soared to $182 billion, plus $89 billion in new stock offerings. All the money goes into investments, so buybacks have plummeted, dropping from $105 billion to around $14 billion this year. Hartnett says the era of massive buybacks is over.

Ironically, the bond market is the last place investors want to be in recent years. Hartnett summarizes asset allocation in the 2020s with several super trends, one of which is ABB—Anything but Bonds.

Hartnett believes the main reason for ABB is the government. As the wealth gap widens, voters are demanding more government spending to subsidize living costs, and the government can only keep issuing more debt. But persistent spending expansion has led investors to structurally underweight bonds. Data shows that since 2020, the annualized return of 10-year U.S. Treasuries is minus 4%, while gold is up 15% and the Nasdaq 17% in the same period.

The rise in yields is not only because of ABB. Over the past six months, global central banks have hiked rates 42 times. U.S. Treasury yields have been rising, and the dollar has followed suit. Borrowing is getting more expensive for big tech, and the cost of doing AI is heavier. Hartnett sees the bond market as playing the role of enforcer for AI capital expenditures: the tech rally has clearly stalled, and SOX has already fallen 20% from its high.

Will yields just keep rising? Hartnett’s answer is no. When yields get too high and threaten the equities market and banking system, policymakers will not stand by. As long as central banks signal debt buying, such as through QE or yield curve control, yields won’t spiral out of control.

But “not out of control” doesn’t mean “relief.” Hartnett argues that for a rebound in AI-related stocks, yields must come down materially, and that requires two conditions: first, progress toward peace in ongoing conflicts, driving oil prices down sharply; second, the end of the global rate-hiking cycle. This window is most likely to open in the first half of next year.

Before AI stocks bounce back, BofA favors tech bonds. Because big tech is issuing so much debt, it has suppressed tech bond performance. U.S. investment-grade tech bonds now yield less compared to overall investment-grade bonds than at any time in the last 30 years. Meta, Amazon, Google, Nvidia—all their long-term bonds yield between 6% and 8%. BofA calls these “overlooked pockets” and suggests investors pay attention to this opportunity.

Alright, that’s the event. Now let's hear from U.S. investment analyst Justin.

Justin does not entirely agree with Hartnett’s premise that AI stocks will only rebound if yields fall materially. We can simplify stock price as a fraction: the numerator is the company’s future earnings, the denominator is the interest rate. The higher the rate, the less valuable the discounted earnings.

Looking at the denominator first, U.S. Treasury yields surged in September, the Fed hiked once more, and even Treasury Secretary Yellen had to step in to calm markets, but the stock market remained resilient— the Nasdaq actually rose 1.9% in September. But in October, things reversed. The 10-year yield fell about 40 basis points from its October 7 high, dropped back toward 5%, and today dipped another 6 basis points, now approaching 4.9%. Valuation pressure isn’t as high as before, yet AI stocks are weaker than before.

This comparison shows both numerator and denominator matter, but the AI monetization in the numerator is more important.

So how do you assess AI monetization? The most direct indicators are the ARR of the two OA labs and the core revenue growth of the three major cloud providers.

On October 9, OpenAI’s annualized revenue as of the end of September was well below market expectations, and the Nasdaq dropped 1.25% that day. Anthropic’s IPO prospectus still only provides detailed info up to the end of June, and media reports show the Q3 ARR growth rate is also clearly slowing, below expectations. For both OA companies, we’re not seeing growth signals in line with market hopes.

Regarding the three major clouds, this quarter's cloud business performance was generally lukewarm compared to expectations, but not bad by historical comparison. Regardless, how much is actually earned and what is the ROI of AI investment? Financials don’t break it out in detail, so investors can’t get a clear read. What’s more troubling is that the core businesses previously sustaining AI investment are themselves starting to slow. So where is the evidence for robust AI monetization? It’s unclear.

Therefore, I believe the market can absorb some rate pressure, but the main focus is AI’s growth and returns. For a real rebound in AI, the numerator side must give clear signals—such as a reacceleration in OA’s revenues, or clear ROI numbers for cloud providers, or a stabilization of core businesses. Otherwise, even if yields drop or the Fed cuts rates again, the benefit to AI may only be temporary. So, to sum it up: AI monetization needs clear, data-driven proof that beats market expectations. We remain optimistic but will approach future developments with caution.