In September 2026, debt linked to the Project Jupiter data center in New Mexico, intended for Oracle, began trading at a discount of 89–91 cents on the dollar. Banks involved in the financing faced challenges selling these obligations to other investors, raising concerns about infrastructure construction limits and Oracle's debt load.
Oracle reported a 30% revenue increase to $19.3 billion in the first quarter of the 2027 fiscal year, with cloud infrastructure revenue growing by 121%. However, the company's free cash flow was negative at $5.4 billion, while capital expenditures reached $28.5 billion.
According to Columbia University researcher Stijn Van Nieuwerburgh, total AI investments could reach $10.3 trillion by 2032, requiring annual industry revenues of approximately $3.55 trillion to justify the costs. Meanwhile, new AI-related debt issuance dropped to about $23 billion in September 2026, less than half of the August level.
Broadcom CEO Hock Tan stated in October 2026 that capital, rather than equipment availability, has become the primary constraint for AI development. Rising interest rates increase the cost of debt for capital-intensive projects, pressuring project timelines and expected returns on investment.
The rapid evolution of AI models creates a risk of infrastructure obsolescence, as hardware may become outdated before the initial investment is recovered. Companies continue to invest heavily to remain competitive, despite the potential for creating excess capacity and the need for constant re-evaluation of project economics.
