Investors demand tougher terms as Oracle borrows for AI data centers

In short: Big tech companies are borrowing more money to build AI data centers, and investors are starting to treat that borrowing as riskier.

What's going on

Oracle and other “hyperscalers” (the biggest cloud companies that run huge computer networks) are increasingly using debt markets to pay for massive data center projects. A data center is basically a warehouse full of computers that run online services and AI.

A recent example is Oracle’s $16.3 billion financing for a data center campus in Michigan. The deal reportedly took longer than expected to close, and it leaned heavily on PIMCO to buy about $10 billion of the bonds after banks pulled back. Bonds are like IOUs that companies sell to investors, with a promise to pay interest.

Investors are asking for higher premiums (higher interest), tighter deal structures, and sometimes more equity support (more cash put in by the company or partners). Reporting cited by Bloomberg says Oracle now has about $120 billion in notes included in a major U.S. corporate bond index. Since last year, at least $290 billion in debt financing has been raised for hyperscaler initiatives, showing how much of the AI buildout is being funded with borrowed money instead of business cash.

What to watch

Wall Street is watching whether AI spending will turn into steady cash flow fast enough to comfortably cover these growing debt loads. The risk is like building a large apartment complex with loans and assuming it will stay fully rented for years. If demand for AI services or data center leases comes in slower than expected, refinancing that debt could get harder and more expensive.

Source: NYTimes