Data center

Featured image by Josie Doan/PitchBook News

Private Equity

Meta, Blue Owl data center JV puts investor protections in the spotlight

Meta is unlocking billions of dollars for its Hyperion data center buildout through a joint venture with Blue Owl—but how are investors in the deal protected?
Data center

Josie Doan/PitchBook News

Tech giant Meta announced late Tuesday it is forming a joint venture with Blue Owl Capital, a private asset manager, to build one of the most expensive data centers ever.

The deal is another milestone in the use of private credit to fund massive infrastructure projects. It also sheds light on how lenders and tech companies are protecting themselves in the face of what many believe to be an artificial intelligence bubble.

The JV has committed to invest approximately $27 billion in the development of the Hyperion data center campus in Richland Parish, a sleepy corner of northeastern Louisiana. Blue Owl has an 80% ownership stake in the JV, with Meta accounting for the remaining 20%.

Blue Owl has put up an initial $7 billion in cash for the new entity, with Meta set to receive a $3 billion distribution upon closing. Blue Owl’s contribution is part-funded by fixed-income investors, including Pimco, through a private bond offering.

In a private lending market characterized by relatively low-risk investments in cash-generative businesses, Blue Owl has stood out for its embrace of large-scale, high-tech infrastructure investments.

In January, the firm lined up alongside Oracle, OpenAI and Softbank to back Project Stargate, a government-driven initiative that aims to invest $500 billion in AI infrastructure in the US.

While the true scale of the project remains to be seen, Blue Owl has already backed a $15 billion deal to help Oracle build a data center campus in Abilene, Texas, which will be the home of Project Stargate.

Much of the capital for this investment came through Blue Owl’s real estate funds, which specialize in making triple net lease loans, Co-CEO Marc Lipschultz said on the firm’s second-quarter earnings call.

These financings typically place the burden of capital expenditure and operating expenditure on the borrower, reducing the risk of spiraling costs for the lender.

On the call, Lipschultz said the thirst for capital among developers of AI infrastructure has, for Blue Owl’s real estate funds, created “the best risk-reward setup we’ve seen in our careers.”

Blue Owl declined to comment on which pools of capital were contributing to the Meta JV.

While the headline figures are eye-watering, the deal is an interesting case study in how structuring can mitigate risk for lenders and borrowers.

Meta in its recent earnings report said it expected the company’s 2025 capital expenditures to be in the range of $66 billion to $72 billion and indicated that spending will remain significant in the next year. Still, the deal gives Meta the option to walk away.

For Blue Owl, the transaction includes a residual value guarantee, which means that if Meta chooses not to renew its lease, the facility will be marketed for sale. And if the resale value of the data center falls below a pre-agreed threshold, Meta is liable to make up the shortfall to investors.

“Meta wants the optionality to be able to walk away from this facility if either their strategy changes or they decide they are going to take a different approach to AI training, or God forbid this ‘bubble cracks’ and the world goes a different direction,” said Naveen Sarma, a managing director at S&P, who covers credit ratings for companies in the telecommunication and media sector.

An exit might not be easy. The Hyperion data center is designed specifically for AI model training, and rural Richland Parish was chosen as the location because of its abundant space and easy access to the energy infrastructure required for such projects.

In contrast, traditional data centers, the most likely candidates for repurposing, are typically located near population hubs to minimize latency. As a result, many of them are built in regions like Northern Virginia or California, which have higher land prices and strained power capacity.

Even with the residual value guarantee, the facility could become a stranded asset, causing a significant impact on the local economy.

Not every deal includes similar downside safeguards, added Dhaval Shah, a director at S&P’s global project finance ratings team. If and when the AI bubble bursts, other investors will not be so well protected.

Learn about our editorial standards.