Why Is Bank of America Warning About AI Companies’ Borrowing?

TL;DR

AI companies are borrowing heavily because infrastructure spending is outpacing the cash they can comfortably devote while meeting other obligations. Amazon, Google, Meta, Microsoft, and Oracle are projected to need $2.9 trillion for AI data spending through 2028, while investment could consume 80% of their cash flow next year. Read on to understand why Bank of America sees widening bond spreads as a warning signal.

Transcript

Look, Wall Street just warned the big investors while you and me, the retail investors are once again left uninformed. And it kind of pisses me off that there is this one set of data for the big money and then there's just noise for us retail investors. So, I just read this research report from Bank of America. I've got on the screener in front of ... Read More

Key Insights

• AI infrastructure is extremely expensive because it requires enormous data centers, millions of computer chips, and substantial electricity consumption. Morgan Stanley data cited in the discussion estimates that leading technology companies will need $2.9 trillion for AI data spending through 2028.
• Hyperscaler free cash flow is not expected to cover every planned AI investment comfortably. The five companies discussed reportedly have $580 billion in free cash flow this year, while data-center investments are projected to consume 68% of that total.
• AI spending is projected to absorb 80% of hyperscaler cash flow next year. Such a concentration would leave only 20% for other requirements mentioned in the discussion, including additional investments, dividends, and share buybacks.
• Technology companies are increasingly financing AI development through debt. Bank of America reportedly estimates that the companies may need to borrow $1.5 trillion through 2028 to remain competitive in the AI infrastructure race.
• Recent borrowing has accelerated sharply according to the figures presented. Meta reportedly borrowed $30 billion, while Oracle borrowed $18 billion through bonds plus $38 billion in loans, and another company borrowed $27 billion, contributing to more than $100 billion of financing in two months.
• A bond spread is the additional interest investors demand above the rate on a comparatively safe investment, identified here as US government bonds. A wider spread indicates that lenders perceive greater risk and therefore require more compensation for holding corporate debt.
• Hyperscaler bond spreads reportedly increased from about 0.5% to 0.8%. Although the absolute change appears small, the discussion characterizes it as a 60% increase in the risk premium demanded by investors and a signal that perceived credit risk is rising.
• The bond market can reveal concerns before the stock market because bond investors focus on cash flow and debt repayment. The discussion says US technology bond prices fell 8% during the year before the dot-com bubble peak, preceding the stock-market decline.

Questions & Answers

Q: Why are AI companies borrowing so much money?

Amazon, Google, Meta, Microsoft, and Oracle are pursuing costly AI infrastructure involving data centers, millions of computer chips, and massive electricity bills. Their AI investments are projected to consume 68% of cash flow this year and 80% next year, so they are using bonds and loans to supplement their available cash.

Q: What is Bank of America’s warning about AI borrowing?

Bank of America highlights rising hyperscaler bond spreads as a warning that debt investors perceive greater repayment risk. Those spreads reportedly increased from about 0.5% to 0.8%, which the discussion describes as a 60% rise in the risk premium.

Q: How much could AI infrastructure spending cost through 2028?

Morgan Stanley data cited in the discussion projects $2.9 trillion of AI data spending through 2028. The scale of that requirement means the companies are expected to combine their own cash flow with substantial borrowing.

Q: How much cash flow are hyperscalers spending on AI infrastructure?

The five companies discussed reportedly have $580 billion in free cash flow this year, with data-center investment expected to consume 68% of it. That share is projected to reach 80% next year, leaving less cash for other investments, dividends, and buybacks.

Q: How much might AI companies need to borrow through 2028?

Bank of America reportedly estimates that the companies may need to borrow $1.5 trillion through 2028 to keep up with the AI infrastructure race. The discussion also suggests they may need to raise more than $200 billion next year.

Q: How quickly has AI data-center borrowing increased?

The discussion says more than $100 billion was financed through bonds and loans in two months, including $75 billion in bonds and another $38 billion in loans. Two years earlier, the same companies reportedly borrowed $14 billion during the entire year.

Q: How much did Meta and Oracle reportedly borrow?

Meta reportedly borrowed $30 billion, while Oracle borrowed $18 billion through bonds and another $38 billion through loans. The discussion also mentions another company in the sector borrowing $27 billion.

Q: What is a hyperscaler bond spread, and why does it matter?

A hyperscaler bond spread is the extra interest investors demand for holding a large technology company’s debt rather than comparatively safe US government bonds. A wider spread means lenders want more compensation because they perceive greater risk, making it a potential warning metric for financial stress.

Summary & Key Takeaways

• Amazon, Google, Meta, Microsoft, and Oracle are competing to build costly AI infrastructure, including data centers, computer chips, and electricity capacity. Morgan Stanley data cited in the discussion projects $2.9 trillion of AI data spending through 2028, forcing even highly profitable technology companies to supplement their cash flow with substantial borrowing.

• The companies reportedly have $580 billion in free cash flow this year, but AI infrastructure investment is expected to consume 68% of it. That proportion could reach 80% next year, leaving less money for other investments, dividends, and buybacks while increasing dependence on bond issuance and loans.

• Bank of America highlights hyperscaler bond spreads as an AI bubble warning metric. These spreads reportedly rose from 50 to about 80 basis points, a 60% increase in the risk premium. The discussion argues that bond investors may detect financial stress before stock investors because they focus closely on cash flow and repayment capacity.