JPMorgan and Mitsubishi UFJ Financial Group initiated a $38 billion syndicated loan in August to finance Oracle data center projects in Texas and Wisconsin as part of Oracle’s partnership with OpenAI, according to people familiar with the matter; however, as of March 2024, the deal is nearing completion with more than 20 banks and investors participating, though lenders continue seeking to place nearly $1 billion in remaining commitments, signaling a notable shift in market appetite for AI infrastructure financing.
The $38 billion facility is structured as a delayed draw term loan, allowing Oracle to draw funds in tranches over time, according to insiders. The loan comprises multiple components: a $23 billion tranche for the Texas data center hub, expected to close within weeks; a $15 billion tranche for the Wisconsin project; an $18 billion debt package for a New Mexico facility; and a $14 billion financing for a Michigan facility still in final negotiation stages. Silver Lake and DigitalBridge Group have jointly committed $30 billion in equity investment to the projects, per people familiar with the transaction.
The difficulty lenders faced in completing the syndication reflects a broader cooling in investor enthusiasm for mega-scale data center financing. According to insiders, lenders anticipated subscription completion in December and January but were forced to cast a wider net to attract investors, including insurance companies, infrastructure funds, and participants in Asia. The fact that some lenders retained debt holdings above expected levels as the deal neared completion indicates that data center projects with Oracle as the primary tenant are facing increased scrutiny, according to people familiar with the matter.
The $275 billion in mega data center borrowing completed in the prior year represents the previous market appetite threshold; the current syndication difficulties suggest that threshold is declining, according to market participants. Sean McDermott, a partner at consulting firm Arthur D. Little specializing in data center business, stated that “the market will experience a ‘flight to quality,’ but I believe this is not because AI demand is slowing—rather, reality is catching up to ambition. Demand remains, but the challenge now is execution.”
Oracle’s cost of debt insurance, measured by credit default swap spreads, surged during the syndication period and reached historic highs in March 2024, according to S&P Global Ratings data. This escalation in perceived default risk complicated the financing effort significantly. The company faces negative free cash flow in the coming years due to massive capital expenditures required to expand data center capacity for customer computing processes, per S&P Global Ratings analysis.
Oracle’s financial trajectory is increasingly intertwined with OpenAI’s fortunes. The Stargate contract originally envisioned $500 billion in investment over four years, though the exact deployment timeline and funding structure remain subject to negotiation. Oracle recently appointed a new chief financial officer to manage the large-scale data center development program, according to company announcements.
To increase appeal to smaller investors, lenders recently offered a 1 percent upfront fee waiver on the facility, which carries a spread of 2.5 percentage points over the Secured Overnight Financing Rate (SOFR), according to people familiar with the transaction. The project financing structure—in which lenders invest in the project itself rather than the company, allowing debt to be segregated from the corporate balance sheet—represents the market limit of this lending model for data center scale, per insiders.
The Michigan project achieved a breakthrough when Oracle agreed to tighten certain lease terms, and PIMCO agreed to provide anchor funding for the $14 billion debt tranche after Bank of America Securities spent months attempting to assemble the financing, according to people familiar with the matter. This development allowed the Michigan facility to advance toward final closing.
McDermott stated that “many participants are moving rapidly and aggressively to secure their position in the AI ecosystem, and people have been trying different approaches to achieve this.” The syndication challenges reflect not a collapse in AI infrastructure demand but rather a market transition toward more rigorous credit analysis and a preference for projects with established revenue streams or creditworthy sponsors, according to market observers.
The successful completion of the Oracle financing, despite syndication headwinds, will be closely watched by the financial community as an indicator of whether mega-scale AI infrastructure projects can continue to attract capital at acceptable pricing levels.
Q: What is the current status of Oracle’s $38 billion data center financing as of March 2024?
A: The deal is nearing completion with more than 20 banks and investors participating in the syndicate, according to people familiar with the matter. The $23 billion Texas tranche is expected to close within weeks, followed by the $15 billion Wisconsin tranche. The $14 billion Michigan facility advanced after PIMCO agreed to provide anchor funding, per insiders.
Q: Why are lenders having difficulty placing the full $38 billion syndicate?
A: Lenders continue seeking to place nearly $1 billion in remaining commitments because investor appetite for mega-scale data center financing has cooled following $275 billion in borrowing in the prior year, according to people familiar with the transaction. Rising credit default swap costs for Oracle, which reached historic highs in March 2024 per S&P Global Ratings, have also increased perceived risk among potential investors.
Q: What is the Stargate contract, and how does it relate to this financing?
A: The Stargate contract is a partnership between Oracle and OpenAI that originally envisioned $500 billion in AI infrastructure investment over four years, according to people familiar with the matter. The Texas, Wisconsin, New Mexico, and Michigan data center projects are components of this broader partnership and are designed to provide computing capacity for OpenAI’s operations.