When bond yields get too high: The risk to AI and global markets
As federal deficits swell and hyperscalers borrow for AI, a decisive move above 5% on the US 10-year Treasury yield could halt projects and reshape markets worldwide.
As federal deficits grow and hyperscale companies borrow heavily for artificial intelligence, a 5% rise in the US 10-year Treasury yield could halt projects and reshape global markets. Two key dynamics are at play: swelling government debt and the impact of rising borrowing costs on the AI boom that has driven equity markets for two years. The meeting point of these forces is the yield on the US 10-year government bond.
Warnings about the US debt burden date back to the 1970s when the country abandoned the gold standard. However, a bull market in the 1980s and 1990s, along with a recovering economy, prevented a crisis. Deficits and debt have contributed significantly to corporate profitability, benefiting technology giants that are now among the most profitable companies.
Fiscal deficits above 6% of GDP, particularly during full employment, transfer income to the private sector, either directly or indirectly through tax cuts. This arrangement has been viable until the bond market's recent steep repricing, with the US 10-year Treasury nearing 4.8%. As a result, federal interest expenses have doubled as a share of GDP, signaling a shift from abstract fiscal concerns to immediate budgetary issues.
The AI boom is fueled by record borrowing by hyperscalers, who are increasingly relying on debt markets to fund their investments. This development may lead to a bubble if cheap money persists. The 10-year Treasury yield, currently around 4.8%, is nearing the top of post-dotcom era levels. A decisive move above 5% would have economic and psychological implications, potentially slowing borrowing costs and enabling AI investment to continue.
However, the combination of soaring sovereign borrowing and growing private sector demand for AI could pose challenges.
Written by urgent.news from Daily Maverick's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.