The all-consuming AI craze is squeezing what many have considered, for decades, the world’s safest asset: US sovereign debt. And the spillover will be worldwide.

Things got out of hand so fast that the world’s biggest economy had to, in a way, bail itself out by promising to buy back long-term debt to tame the yields. It worked for a few hours, but the yields on the benchmark US 10-year bonds resumed their move back towards 5%.

Simply put, the US government now has to compete with hyperscalers borrowing insane amounts of money, leaving less and less on the table every day even for Washington, D.C.

Imagine you’re a large pension fund deciding where to invest billions of dollars. You can lend money to the US government. Or perhaps you can lend to one of the world’s richest technology companies at a more attractive yield. Reports suggest some Google-linked financing has offered returns as high as 9.3% to attract the capital required.

Meanwhile, the US government debt has crossed $40 trillion, making many around the world nervous.

The dilemma is so real that the world’s biggest and safest borrower has seen its cost swell by 27% from a year earlier, and this week, it hit the highest level since 2007. That’s bad news for almost everyone else, including those holding the US dollar. Except if you’re holding gold and Bitcoin.

The hyperscalers’ unhinged appetite for capital can also be a double-edged sword. A rise in yields can push up interest rates, making the AI boom more expensive to finance. A delicious irony!

All valuations rest on the expected earnings years from now. Investors effectively translate those future dollars into what they’re worth today using a discount rate heavily influenced by government bond yields. The higher that rate climbs, the less valuable distant profits become today.

So AI potentially gets squeezed from both directions: the infrastructure becomes more expensive to finance, while the future profits expected to justify that infrastructure become less valuable in today’s money.

And by some accounts, the math doesn’t back even the current prices. Peter Berezin, chief economist at Montreal, Quebec-based BCA Research, estimates that the hyperscalers may need $10 trillion in revenue to justify the data centre investments.

$10 Trillion In Annual AI Revenue May Be Necessary To Monetize All The Capex Being Plowed Into Data Centers

Hyperscaler capex is expected to reach $1 trillion in 2027, most of which will be AI-related. Let us assume that a capex bust is avoided and capital spending remains at $1… pic.twitter.com/KBkTzMWsXW

— Peter Berezin (@PeterBerezinBCA) August 21, 2026

“Keep in mind that the foregoing calculation does not even include revenue from SpaceX, the neoclouds, or Chinese AI companies,” Berezin added in his post.

Global AI sales, excluding China, reached $25 billion for hyperscalers and neoclouds in the first quarter of 2026, according to data from Exponential View, an independent research organisation founded by tech entrepreneur and author Azeem Azhar in 2015.

That leaves about $4 billion on the table after covering the estimated depreciation costs of $21 billion per quarter. The narrow margin will shrink further if financing costs keep rising.

The central question surrounding AI was whether the technology would work. Then came another question: could companies make money from it?

We’re now beginning to encounter a third: who finances the infrastructure required to make all of this possible—and at what price?

That question reaches far beyond Silicon Valley because the price of capital doesn’t stay inside a data centre. Government borrowing costs influence corporate bonds. Corporate borrowing costs influence investment. Bond yields influence mortgages, valuations and currencies. Eventually, the price of money works its way through almost everything.

Share this with your friends and ask how they think this will play out.

For more such priceless insights, become a member of the CNBC-TV18 Access community where you can find the opinions from some of the influential voices, and sometimes get to meet them in person too. Check out our upcoming events here.

Happy Reading, and Stay Ahead of the Curve!