Five years of borrowing 350 billion dollars to bet on AI! Are the top five US tech giants making history, or are they repeating Intel's mistakes?

The Zhitong Finance App learned that in order to seize the high ground of artificial intelligence data centers, the five major US tech giants have doubled their debt size over the past five years, paving the way for what they call “economic transformation” investments through large-scale borrowing. According to data compiled by the market, Alphabet (GOOGL.US), Amazon (AMZN.US), Meta (META.US), Microsoft (MSFT.US), and Oracle (ORCL.US) — the five largest US companies investing in new data centers — have added a cumulative total of about 350 billion US dollars in debt over five years.

These companies are convinced that cutting-edge AI services will bring in huge revenue in the future. Investors have also responded enthusiastically in the past, actively subscribing to the multi-currency denominated bonds it has issued. However, according to people familiar with the matter, Amazon's $25 billion bond issue this week was met with rare cold treatment in the market, indicating that the market's willingness to fund investment by tech giants is not endless.

For most companies, the cost of debt is still relatively manageable, and profitability remains strong. The total interest expenses of the five companies surpassed $10 billion last year, more than double that of 2019, but it is still insignificant compared to the free cash flow of any of them. By the end of March, Google's operating cash flow, after deducting capital expenses, had reached 64 billion US dollars in free cash flow.

However, some companies' balance sheets are already under pressure. Amazon's free cash flow turned negative for the quarter ending March 31; Oracle's 2025 debt is about 2.5 times revenue, and cash consumption is expected to accelerate. S&P Global Ratings downgraded Oracle to its lowest investment grade rating on Thursday because its AI spending continues to rise.

Software companies are generally high-margin businesses, and regular capital expenditure requirements are low. But for industry giants, the advent of cloud computing has changed this pattern — server farms require large-scale investments. AI data centers are generally larger than previous facilities, and chips are also more expensive, further driving up capital expenses.

DA Davidson & Co. analyst Gil Luria said, “The nature of these businesses is undergoing drastic and abrupt changes, which is why their cash flow is currently so sluggish.”

He argues that the companies argue that the expected return on investment from the new AI service — particularly in relation to the lower interest rate levels on their new bonds — is enough to make the expansion worthwhile. “That's what they said to us,” he said, “but you can see that investors aren't at ease.”

Amazon CEO Andy Jassi said in April that he “has a high level of confidence that it will be monetized” and cited customer promises to add data center capacity to its cloud business AWS. Meta CEO Mark Zuckerberg said in an interview recently that demand for AI computing power continues to be in short supply. “This gives us great confidence that continuing to build infrastructure will be a good investment.”

Equity investors are increasingly cautious about how and when large cloud computing companies (commonly known as “hyperscale enterprises”) reap rewards from huge expenses. Since this year, Alphabet's stock price alone has outperformed the S&P 500 index, and both Microsoft and Oracle's stock prices have fallen by more than 20%.

As these companies release quarterly earnings reports starting late this month, investors in the bond market will keep a close eye on their spending plans — which has become a weather vane for the health of the AI boom. Market concerns have expanded from “whether the company can keep up with the pace of peer expansion” to “how to finance” and “when the return will arrive”.

Jason Pompeii, a corporate debt analyst at Fitch Ratings, said, “I'm not sure we know if Amazon, Google, Microsoft, and Meta can actually get a return on their investment. At present, it seems that the massive demand hype is still largely a vision.”

Five large-scale enterprises have promised capital expenditure of up to 725 billion US dollars this year, mainly for data centers and the Nvidia chips that drive their operation. Sources of funding include own cash, additional loans, and partial financing through off-balance sheet arrangements by companies such as Meta.

Rising debt burdens may reduce a company's ability to respond to future crises or technological changes. Even decades of dominance in the technology sector do not save companies from the heavy burden of debt — once the position loosens, the risks become apparent.

Intel CEO Chen Liwu said when he took office in 2025 that his top priority was to clean up the balance sheet to put an end to questions from the outside world about whether the company can survive. The company, which remained the world's largest chip maker until 2022, borrowed heavily during its predecessor's tenure to support investor returns, acquisitions, and ambitious capacity expansion.

However, as debt continued to accumulate, Intel was unable to launch competitive AI chips, missing out on a huge opportunity for Nvidia to become the company with the highest market capitalization in the world. The wrong choice of a manufacturing technology route led to loss of product market share, revenue began to shrink, and losses were soon reported that were once unimaginable.

Wall Street expects 2026 to be its third consecutive year of losses. Ultimately, it was the US government's bailout and investment from Nvidia — a company that had been suppressed by Intel for decades — that allowed the established Silicon Valley giant to maintain operations.

DA Davidson analyst Luria said that currently hyperscale cloud companies are far from reaching that point. Referring to the total debt of these companies, he said, “This doesn't look bad. What if they borrow an order of magnitude? That would be bad.”