Oracle‘s $18 billion data center financing is under more pressure as worries about its growing debt load, financial prospects for its AI infrastructure plan and the resistance of one of the company’s biggest planned computing campuses in New Mexico mount. In the private debt markets, loans have reportedly depreciated from the project’s original value due to increased risk from the project, and caution on the part of banks and investors.
The debt is tied to Oracle’s planned data center campus in Doña Ana County, N.M., a large-scale initiative aimed at powering Oracle’s growing AI initiative. Project Jupiter is part of a larger deal between Oracle and OpenAI, covering approximately 1,400 acres, which will support the vast computing power needed to develop and operate sophisticated AI systems.
Late last year, a group of banks reportedly offered some $18 billion in financing for the project. The money was supposed to support the start of building and infrastructure for a very large scale operation of a data center. But, they have reportedly been having trouble allocating the debt to a broader set of investors.
Banks that have participated in the financing of the loans, such as Santander and Jefferies, have recently quoted around 89-91 cents on the dollar for the loans. If its price drops to that level, that means the debt is trading at a discount in the market. This doesn’t necessarily signal that a borrower will default, but it can signal that lenders and investors are expecting to be paid higher for the perceived financial or project related risk.
The situation is especially relevant, as Oracle is investing significant resources in growing its infrastructure for artificial intelligence. Data centers are some of the most capital-intensive sectors of the computer industry, as they are the most energy and electricity-consuming part of the services. Companies looking to build capacity in a hurry can thus be subjected to high initial expenses before the return on investment from the infrastructure can be realized.
Oracle has been aggressively pursuing an expansion strategy to keep pace with the demand for AI computing needs. While that approach has positioned the company for greater success in the ever-more-competitive landscape, it also has led to closer examination of the company’s balance sheet. The company has been an eye-catcher for investors given the billions of dollars it is spending on new infrastructure and technology projects.
The additional pressure comes from credit worries. Oracle’s corporate credit rating was lowered by S&P in July, leaving the company only one level above speculative or junk status. They may make financing more costly and the rating may be a factor in the bank and institutional investment decision making process for new debt on large projects.
This isn’t just a problem for Oracle. If loans made for a large-scale infrastructure scheme are proving hard to market to other investors then perhaps the banks that lent them the money will end up carrying more of this risk on their books. This may reduce the amount of capital available for other loans and make financial institutions more hesitant about other large-scale loans.
Project Jupiter isn’t only struggling in the financial markets. There has reportedly also been a growing local protest against the possibility of the data center affecting water and air quality. Water, depending on the cooling system used and the local climate, can be a significant resource in large data centers, as can be electricity. AI infrastructure’s environmental impact has become a critical concern for communities where new AI facilities are slated for construction.
The Jupiter debate is emblematic of a larger issue facing the tech industry. Artificial intelligence companies and their infrastructure partners are competing to develop computing power, and there is debate from local governments and citizens around the economic value of these capabilities to their electricity, water and other resource needs.
The development in New Mexico is significant for Oracle, given its broader AI initiatives. As generative AI continues to grow and evolve at a rapid pace, there is an increased demand for high-performance data centers that can handle the advanced processors and computing workloads associated with this technology. Acquiring sufficient capacity has become a key focus area for cloud providers and technology companies but building this capacity is an expensive commitment and takes time.
Project Jupiter has also come across an energy-related hurdle. The site was originally planned to be supplied with around 2.2 GW of gas turbines. But, the New Mexico state land office blocked a request to use a natural gas pipeline for the site, reports say. The project development schedule, operating costs and/or infrastructure requirements may be impacted by any changes to the planned energy supply.
The energy challenge is one example of the challenges of constructing modern AI data centers. It is not possible to approach these facilities in the same way as conventional commercial buildings. They need to have reliable and constant power supply for their computing machines, and they could have a large operation that demands a considerable amount of power from the local system. The challenge for developers is to ensure they have reliable energy supply, along with environmental regulations, land use planning and community issues.
The Oracle data center debt looms on an issue when the company is trying to juggle a number of priorities. It needs to sustain the efforts of heavy investment to gain a foothold in the rapidly expanding AI industry while continuing to have access to low-cost capital and retaining investors’ trust in its financial health.
It’s no simpler for the lender. While AI infrastructure is generally considered a potential growth area, each project has its own set of obstacles, including construction delays, regulatory hurdles, funding issues, and uncertainty regarding future demand. So a project may have a high strategic importance and at the same time be high risk financially.
