Higher bond yields raise the cost of financing AI data centers

In short: Rising US government bond yields are making it more expensive to borrow money for AI data centers and chips, especially for smaller players.

What's going on

The yield on the US 10-year Treasury, a widely watched government interest rate, has climbed to about 5.17%. That is roughly one percentage point higher than at the start of the year. It briefly hit 5.34% on October 1, 2026, the highest level since 2002, before easing.

These Treasury yields matter because they act like a baseline for many other loans. When the government has to pay more to borrow, companies usually have to pay more too. It is similar to how a store might raise prices when its own suppliers charge more.

AI infrastructure needs large upfront spending. That includes data centers, power connections, cooling equipment, servers, and advanced chips. Big cloud companies have committed hundreds of billions of dollars to these build-outs, and many projects across the industry rely on debt, meaning borrowed money.

Higher rates can raise interest bills on new loans and on some existing loans with floating rates (rates that move up and down). They can also make some projects harder to justify if the expected profits do not rise along with borrowing costs. CNBC cited one estimate that AI-related debt issuance could reach $4.1 trillion through 2030, which is a projection, not money already borrowed.

What to watch

The impact will not be the same for everyone. Large tech firms often have more cash and stronger finances, so higher rates may slow or reshape plans rather than stop them. Watch for stress to show up first among more heavily indebted data-center operators and newer cloud providers, especially if AI demand and revenue do not grow fast enough to cover higher financing costs.

Source: NYTimes