Treasury yields have climbed to their highest levels since 2007, raising borrowing costs for companies financing the historic AI infrastructure buildout. JPMorgan Chase estimated in June that $4.1 trillion in AI-related debt will be issued through 2030. With the 10-year Treasury yield near 5.17%, up about 1 percentage point since the start of the year, issuers must offer higher returns to attract investors.
Market reactions vary among key players. CoreWeave shares rose almost 8% this week, while Oracle fell 7% for the week and about 30% this year. SoftBank raised $11.1 billion in junk bonds, with yields reaching 9.75% for a seven-year tranche. Hyperscalers like Amazon, Google, Meta, and Microsoft maintain investment-grade ratings, but neocloud providers face tighter lending standards. CoreWeave disclosed that a 100-basis-point rate increase could raise its interest expense by $30 million. Oracle reportedly sent a force majeure notice regarding its New Mexico data center project, though it stated the project remains on schedule.
The surge in Treasury yields introduces significant operational risk for the AI sector, particularly for entities lacking investment-grade credit ratings. While hyperscalers can absorb higher capital costs due to strong balance sheets, neocloud providers and smaller infrastructure firms face a more challenging financing environment. The divergence in stock performance between CoreWeave and Oracle highlights how market participants are assessing individual exposure to rising interest rates and potential project delays. Lenders are becoming selective, reducing the pool of viable borrowers from dozens to approximately twenty, which may consolidate the market around larger, better-capitalized players.
Despite the increased cost of capital, demand for AI services remains robust, suggesting that price sensitivity is currently secondary to capacity acquisition. Companies like SoftBank and OpenAI-backed ventures appear willing to pay premium yields to secure funding, driven by insatiable demand for compute resources. However, regulatory and political headwinds, such as local opposition to data centers and environmental permit halts in Texas, add layers of uncertainty. Investors should monitor whether sustained high yields eventually throttle new project launches or if the urgency to meet AI demand continues to override traditional financial caution.
