
"Implicit Guarantee" but "Not Consolidated"! Can Meta's Data Center "Off-Balance Sheet Financing" Hold Up?

Meta successfully placed a massive asset and liability off-balance sheet by spinning off a $27 billion data center project into a joint venture with Blue Owl. This complex financial engineering relies on three contradictory accounting assumptions: denying substantive control over the joint venture, asserting that a four-year lease may not be renewed, and that the residual value guarantee provided is unlikely to be triggered, thereby meeting the funding needs for AI infrastructure while maintaining a perfect credit rating
Meta is utilizing a typical structured financing approach, attempting to achieve seemingly irreconcilable financial reporting goals: to leverage external funding to build a giant AI data center worth tens of billions of dollars while striving to avoid having the associated massive debt and assets reflected on its own balance sheet. This operation, referred to as "AI meets artificial accounting," is pushing complex financial engineering to its limits.
According to The Wall Street Journal, Meta announced a new financing deal last month, spinning off its $27 billion "Hyperion" data center project in Louisiana into a joint venture established with investment management company Blue Owl Capital. Meta holds a 20% stake in the joint venture, while a fund under Blue Owl holds 80%. The holding company Beignet Investor, which holds Blue Owl's shares, subsequently sold a record $27.3 billion in bonds to investors, primarily Pimco.
Meta stated that it would not consolidate the financial statements of the joint venture, meaning that this massive asset and corresponding liabilities will remain off Meta's balance sheet. Meta claims it does not control the activities that most significantly impact the economic benefits of the joint venture, and that lease renewals do not have "reasonable certainty." This arrangement aims to maintain Meta's nearly perfect credit rating while meeting the enormous funding needs arising from AI infrastructure construction.
However, this favorable accounting treatment for Meta is based on a series of convenient yet seemingly contradictory assumptions. From the attribution of operational control to the determination of lease terms, to the implicit guarantees provided to bondholders, a series of accounting decisions by Meta is facing close scrutiny. The core of market skepticism lies in whether Meta is essentially bearing risks and obligations that should be reflected on the balance sheet.
Accounting Challenges of "Off-Balance Sheet" Treatment
Meta is using an accounting structure known as a Variable Interest Entity (VIE) to handle the joint venture. According to accounting standards, if Meta is the "primary beneficiary" of the entity, it must consolidate it in its financial statements. To be the primary beneficiary, two conditions must be met: first, having the power to direct the activities that most significantly impact the VIE's economic performance, and second, having the obligation to absorb significant losses or the right to receive significant benefits.
In terms of the determination of "control," Meta's claims contradict its actual capabilities. Although Blue Owl controls the joint venture's board, this voting power and legal form are not decisive factors in accounting determinations.
The key lies in substantive power versus economic influence. As a hyperscale computing company, Meta clearly possesses more expertise and decision-making ability in operating data centers than financier Blue Owl. The economic success or failure of the project ultimately depends on Meta's decisions and technology, yet Meta insists in its disclosures that it does not "direct the activities that most significantly impact the economic performance of the joint venture."
Risk Exposure and "Invisible" Guarantees
In terms of economic benefits and obligations, Meta's actual risk exposure is more apparent. Meta has operational control over the data center and its construction, bearing the risks of cost overruns and project delays More critically, Meta has provided a so-called "residual value guarantee."
This guarantee clause stipulates that if Meta does not renew the lease or terminates it early, and there is no new tenant to take over, Meta will cover the full amount owed to bondholders such as Pimco. This guarantee significantly reduces the risk for bond investors while also indicating that Meta is deeply involved economically. This arrangement provides a strong incentive for Meta to renew the lease, as once Meta defaults and leaves, it still needs to repay the debt through the guarantee.
The Game of "Operating Lease" and Renewal
The transaction structure is set up as a leasing model, with Meta leasing the data center starting in 2029, an initial term of only four years, and an option to renew every four years, for a maximum of 20 years. This short initial leasing structure allows Meta to adopt "operating lease" accounting rather than "finance lease" accounting, thereby minimizing the lease liabilities and related assets that need to be recognized. If a finance lease were used, Meta would appear more like an asset owner on the books and bear the associated debt.
To maintain this favorable accounting outcome, Meta must assert under accounting rules that future renewals are not "reasonably assured." This is a highly subjective threshold. However, considering the long-term nature and substantial investment in AI infrastructure, using only four years seems contrary to business logic. There is a clear tension here: if the guarantee incentivizes Meta to renew, then the renewal should be viewed as more certain, thereby increasing on-balance-sheet liabilities.
Logical Paradox: Wanting, Needing, and Expecting
Meta's accounting treatment ultimately falls into a logical paradox. According to an analysis by The Wall Street Journal, if, as per Meta's assumptions, renewal is uncertain, then the likelihood of triggering the "residual value guarantee" increases. According to accounting rules, if payment is "probable," Meta should recognize this guarantee liability. Conversely, if the guarantee makes renewal certain, then the scale of lease liabilities should significantly increase.
For Meta's current accounts to hold, investors must simultaneously accept three conflicting assumptions: Meta lacks decision-making authority over key project matters; the lease term for this core asset may only last four years; and Meta is unlikely to fulfill its substantial guarantee. This series of judgments based on a specific factual pattern reflects not so much economic substance as a display of sophisticated financial engineering
