Oracle has issued a force majeure notice regarding its Stargate project's key data center. The simultaneous reduction in OpenAI's infrastructure investment, the increase in U.S. interest rates, and the surge in Treasury yields have raised concerns about a potential bubble in debt-dependent AI infrastructure investments.
According to the IT industry on September 26, Oracle notified the developer of the Project Jupiter data center campus in New Mexico of the force majeure. Oracle is reportedly stepping back from its position as the core tenant of the facility to allow for delayed payments if the data center does not become operational as planned by 2028.
The reason for this delay is a power supply issue. The pipeline supplying gas to the 2.45 GW campus has faced regulatory setbacks, pushing the completion date to February 1, 2027, about six months later than expected. Additionally, air quality permits for the fuel cell power generation facilities have not yet been approved.
The market reacted immediately, with Oracle's stock closing down 3.47%, while the parent company of the developer, Blue Owl Capital, fell 3.6%. According to the Financial Times, Oracle's $18 billion loan linked to this data center is currently trading at around 89 to 91 cents on the dollar.
Oracle's financial burden is already significant. Analysis from London-based credit hedge fund Sonar Asset Management indicates that Oracle's debt reached $125.3 billion as of August 31, and the yet-to-be-launched data center lease agreements total $288 billion. Of the $664 billion in backlog, only about 13% is expected to convert to revenue within the next 12 months. A substantial portion of this backlog is reportedly tied to OpenAI, which has not been profitable on an adjusted basis. Recently, Oracle increased its estimated costs related to layoffs by about $700 million, raising the total restructuring plan cost to approximately $2.8 billion for 2026.
OpenAI, at the center of AI infrastructure funding, has also slowed its pace. On September 20, OpenAI revised its infrastructure investment target for 2030 from $1.4 trillion to $600 billion. Nvidia has conditionally pledged up to $100 billion to OpenAI but is reportedly considering reducing its participation to a maximum of $30 billion.
Warnings about a circular investment structure have also emerged. Sonar Asset Management analyzed $3.6 trillion in funding across 176 transactions involving 202 companies in the AI sector, classifying about 120 of these as highly circular. This means the same companies appear as investors, customers, and suppliers simultaneously. Given OpenAI's central role in many agreements, any slowdown in its spending or funding could quickly impact Oracle, CoreWeave, and their supply chains.
CoreWeave is also facing financial challenges. On September 17, it announced plans to issue $3 billion in convertible bonds maturing in 2033 (with an additional $500 million option), and its stock fell more than 4% immediately after the announcement. As of the end of June, CoreWeave's debt stood at $35.6 billion, with an S&P Global credit rating of B+. The premium on CoreWeave's five-year credit default swaps (CDS) has surged from about 500 basis points to 800 basis points in the past three months.
The interest rate environment is exacerbating these burdens. The U.S. Federal Reserve unanimously raised the benchmark interest rate by 0.25 percentage points on September 16, marking its first increase since 2023, bringing the target range to 3.75% to 4%. The yield on the U.S. 10-year Treasury note reached 5.12% on September 23, the highest level since 2007. The increase in corporate borrowing for AI infrastructure has led to a rise in bond supply competing for investment demand, contributing to the higher interest rates.
According to Goldman Sachs, about 24% of the total issuance of investment-grade corporate bonds in U.S. dollars this year has been attributed to AI-related issuers. Goldman Sachs Asset Management noted on September 24 that significant bond issuances from major cloud companies are expected, which is why it currently maintains an underweight position in the AI sector compared to the market.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.