Meta Pushes Its Largest Data Center Project Off Its Books With $27B JV

Meta has formed a $27B joint venture with Blue Owl Capital to finance the development of its planned AI data center megacampus in Louisiana.

The deal marks an inflection point in Big Tech’s use of special purpose vehicles to fund billions of dollars of artificial intelligence infrastructure with debt that will never officially appear on their balance sheets. 

Meta Platforms and Blue Owl’s JV, announced Tuesday, will develop and own the Hyperion data center campus, a 2,250-acre complex planned in northeast Louisiana. Funds managed by Blue Owl Capital will own an 80% interest in the JV, while Meta will own 20%.

Much of the capital raised by Blue Owl is debt funded through securities Pimco issued earlier this month.

While this SPV structure, which allows hyperscalers to effectively borrow billions of dollars without the debt appearing on their balance sheets, is increasingly being pitched by tech firms to finance unprecedented data center spending amid an AI arms race, the Hyperion deal is the largest such transaction to be completed so far. 

To some industry observers, the use of SPVs and the growing use of debt to fund AI build-out are indicators that the data center building boom is entering bubble territory. But leaders at Meta and Blue Owl framed their deal as a responsible path to pursuing what CEO Mark Zuckerberg has termed “superintelligence” in the near future.

“Our AI ambitions will be realized through our ability to deliver the infrastructure to support it,” Meta Chief Financial Officer Susan Li said in a statement. “Our partnership with Blue Owl Capital to develop the Hyperion Data Center is a bold step forward — combining Meta’s deep expertise in building and operating world-class data centers with Blue Owl’s strength in infrastructure investment.”

Blue Owl contributed around $7B to the JV, while Meta contributed the land and assets under construction connected to its Hyperion project. 

Located in Richland Parish, Louisiana, Hyperion will be Meta’s largest-ever campus, spanning nearly 4M SF of data center space across at least nine buildings, with construction continuing through 2030. The project is initially slated to host 2 gigawatts of AI computing, although Zuckerberg has said its capacity could eventually surpass 5 GW.

Some of that power will be generated by a new natural-gas power plant being developed next to the site by New Orleans-based energy firm Entergy. 

With the JV, Meta will lease all the facilities on the Hyperion campus once construction is complete. These lease agreements will have an initial four-year term with options to extend.

Four years is a relatively short term for a hyperscale lease, which are typically no shorter than 10 years. But to offset the added risk that comes with this optionality, Meta is providing the JV with a residual value guarantee for the first 16 years of operations — an agreement in which Meta would make a cash payment to the JV following a nonrenewal or termination of a lease. 

Meta’s leadership has previously said this optionality is central to the appeal of developing data centers through SPV structures. With uncertainty around the scale of the firm’s long-term needs for computing clusters devoted to AI training, these structures allow terms that limit risk compared to traditional leases or self-developed data centers.  

On Meta’s second-quarter earnings call, Li said that while the company will continue funding data center development mainly from its balance sheet, the need for flexibility will lead to increased use of off-balance-sheet financing.

“We generally believe that there will be models here that will attract significant external financing to support large-scale data center projects that are developed using our ability to build world-class infrastructure while providing us with flexibility should our infrastructure requirements change over time,” Li said.

But to some industry analysts, SPVs look less like savvy financing and more like evidence of an AI data center bubble. 

Tech giants like Meta, Amazon, Microsoft and Google are spending as much as 50% of their income on capital expenditures for AI infrastructure, well beyond the limits of what Wall Street would traditionally tolerate. This spending has been rewarded amid a wave of enthusiasm around AI’s potential.

But the prevalence of SPVs suggests the tide may be turning, according to Paul Kedrosky, a partner with venture capital firm SK Ventures and a fellow at the Massachusetts Institute of Technology's Initiative on the Digital Economy.

The use of SPVs like Meta and Blue Owl’s partnership shows that tech firms are effectively trying to hide their spending from the market, Kedrosky said on the Plain English podcast.

“That, for me, is a reflection of not wanting the credit rating agencies to look at what they’re spending,” Kedrosky said. “These are all, for me, the beginning of the sign that the bubble is becoming tired because at least there’s a perception that the market will punish me if I continue to keep this on my income statement.”

The debt attached to Meta’s Louisiana project also aligns with growing bubble fears. 

A portion of capital raised by Blue Owl will be funded through Pimco’s issuance of $27B in investment-grade bonds backed by the Louisiana project’s assets. The securitization hit the market earlier this month, with pricing for the debt soaring to more than $1.10 per dollar of face value.

For most of the AI boom, the tech giants leading the charge had funded capex with cash flow. But this has changed over the past eight months, as the building boom has suddenly become much more reliant on debt — particularly asset-backed securities. 

To skeptics, this growing use of debt shows that the players at the center of the AI infrastructure boom are overspending and trying to spread around the risk, leading analysts like TS Lombard's Dario Perkins to draw comparisons to the dot-com bubble's final days.

“I wouldn't touch this stuff now,” Perkins told Axios. “We're much closer to 2000 than 1995.”