Urgent.News

What's breaking now, across thousands of outlets.

AI

Banks are now trying to offload the AI data center debt they just underwrote

Six months after lenders rushed to underwrite data center loans tied to Oracle and AI companies like Meta, JPMorgan, MUFG, Morgan Stanley, and SMBC are scrambling to unload pieces of the $38 billion in construction debt they helped create, according to the Financial Times. The banks are exploring risk transfers that allow them to keep loans on paper while shifting the riskiest portions to outside investors.

This move is necessary as they approach internal limits on the amount of AI-linked exposure any single lender can hold. Oracle's credit default swap costs have skyrocketed to 151 basis points, the highest level seen since 2009, while its credit rating has been downgraded to BBB-, just above junk status. The demand for AI-related bonds has thinned out, with bond offerings from Alphabet, Amazon, Meta, Microsoft, and Oracle trading at a premium to non-AI corporate debt.

The Federal Reserve's September meeting minutes even cited AI-related corporate borrowing as a factor contributing to the rise in the 10-year Treasury yield, now at 5.304%, its highest level since 2002. While some argue that inflation and government deficits are the primary drivers, the reality is that banks are uncomfortable holding such concentrated exposure to a single sector.

This situation highlights the risks of financing AI infrastructure projects and the potential for a more challenging lending environment in the future.

Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at startupfortune.com →

More in AI

Freshworks to buy Torii to track SaaS and AI spending

Freshworks has agreed to acquire Torii, a startup that shows companies which software and AI tools their staff use and what those tools cost.

  • Freshworks acquires Torii for $ undisclosed on Oct 12, 2026
  • Integration of Torii's software into Freshworks' IT service management
  • Torii aids in tracking AI tool spending, addressing shadow IT challenges

Stop trusting a 99% uptime claim: the LLM API relay failures we hit, and how we test stability

*Originally published on the Folkbench blog: https://folkbench.com/en/blog/ai-api-relay-stability-guide?utm_source=luntan&utm_campaign=dev * A relay site may promise "99.9% fast and stable" and a…

  • LLM API relay services may have unreliable performance despite 99.9% uptime claims.
  • Developed three testing approaches: blind trial, concurrency test, and stream state machine test.
  • Key metrics for relay performance: latency, throughput, and reliability.

More from Wednesday 14 October →