Opinion – The Meta/Blue Owl deal broken down: off-balance-sheet gymnastics 24 years after Enron
Prasad Gollakota
Prior to the 2008 financial crisis, the US government insisted that mortgage giants Fannie Mae and Freddie Mac were not part of the federal balance sheet. Their debt did not count towards official public debt metrics, and their bonds and mortgage-backed securities carried explicit language saying they were not obligations of the US government.
Yet markets priced their bonds almost as if they were Treasuries, on the assumption the state would never let its housing agencies fail. There was an implicit government guarantee that allowed Fannie and Freddie to fund 20bp–40bp inside their standalone credit.
Investors were looking at privileges bestowed under their federal charters, including a standing line of credit with the US Department of the Treasury, and precedents of Washington bailing out government-sponsored enterprises.
When stress hit in 2008, that belief was vindicated. Both were taken into conservatorship and taxpayers backstopped more than US$200bn of losses. The obligations had been economically public all along, despite what government accounts showed.
The US$27.3bn Meta Platforms/Blue Owl project bond financing that funds the buildout of the Hyperion data centre for Meta-backed vehicle Beignet Investor is a corporate-scale version of the same accounting mirage.
Beignet is, at one level, a straightforward Rule 144A project bond issuer: a gigantic, long-dated fully amortising, single-asset issue funding a huge greenfield data centre campus in Louisiana. Infrastructure project bonds usually refinance operating assets; here investors are funding construction from day one, with proceeds parked in Treasuries and drawn down as the campus is built. The deal supports the view that institutional investors are comfortable taking large-scale data centre construction risk, provided the sponsor and structure are strong enough. Investors also accepted no make-whole on early repayment of the bonds. That's unusual, but not inexplicable, in a market desperate for long-dated, fixed-rate exposure to the AI revolution. Nothing in that alone demands the elaborate geometry of Beignet’s structure.
The contradiction
The contradiction at the heart of Meta’s Hyperion financing is between the bond ratings and Meta's proposed accounting treatment. If you follow the cash, Beignet's bond is almost pure Meta exposure. Meta funds the construction risk; Meta’s rent and Meta’s residual value guarantee service the debt; and Meta is the party that suffers if demand for AI computing disappoints.
If Meta ever walks away from the project, the RVG obliges it to write a cheque that, together with whatever the campus (or part thereof) can be sold for, makes bondholders whole. The RVG runs for 16 years and is triggered if Meta fails to renew a lease at expiry, terminates early or defaults under the lease. Any one of these leads to a payment of a “guaranteed minimum value” that is sized so the bond can be repaid, either from the underlying sale proceeds or directly by Meta. The final four years of the bond are not covered by the RVG but by a termination fee equal to the outstanding balance at that point.
It is at this point that S&P, the only agency to rate the deal, essentially stops its analysis. Beignet is Meta credit in project-finance clothing and S&P’s rating work makes that explicit. During construction, S&P applies a “Meta-adjusted” lens and lands at Meta minus one (from Meta’s AA– to A+) for structural wrinkles (no direct recourse to project assets or contracts, and legal separation between the JV and bond issuing vehicle). Once the campus is operating, S&P applies its project finance methodology and based on a strong debt service coverage ratio of 1.12x, and assuming low variability in lease revenues given Meta’s profile and the RVG, S&P again arrives at A+. Effectively, investors get Meta-linked risk at a spread pickup to Meta’s own long bonds.
US GAAP standards, by contrast, are still prepared to be persuaded that form trumps substance. The deal structure shifts 80% of the equity and the formal governance of the joint venture, JV Co, between Meta and its joint venture partner, private credit firm Blue Owl Capital, into the latter’s hands. Do that, while capping and distancing the RVG, slicing the obligation into four-year lease blocks with options, and you can argue that Meta neither controls the relevant special purpose vehicles, including JV Co, nor has a long-dated liability.
The result: a transaction the ratings agency treats as “Meta minus one”, but where the proposed accounting treatment keeps the debt off balance-sheet because the leases are considered operating (not finance) obligations.
Unpacking accounting consolidation
After the demise of Enron, US GAAP (under ASC 810) introduced the concept of a “variable interest entity” to capture special purpose entities or project vehicles and ask a simple question about them: should this really be part of another group’s balance sheet or not? If it should be consolidated, all the assets and liabilities of that vehicle are treated as if they belong to the group. The framework was introduced to stop companies moving economic risk into boxes they did not consolidate, only to be revealed in distress and magnifying a death spiral.
The Hyperion structure is complicated by the clever separation of Beignet, the issuer of the liabilities, and JV Co as the holder of the asset. Simply consolidating JV Co would not result in the 144A bond coming on to Meta's balance sheet. There would need to be some sort of upstream look-through by accountants to arrive at this result.
Nonetheless, ASC 810 still asks a central question: if Meta keeps Beignet and JV Co off its books, who is the primary beneficiary of the VIE? The primary beneficiary must both have power over the activities that most significantly affect the VIE’s economic performance and have exposure to losses or rights to benefits that could be “potentially significant”.
Meta is proposing to assert that Blue Owl, as 80% equity owner, has greater power over the main activities that most affect the JV’s economic performance and that Meta’s exposure is not “potentially significant”, relative to the JV as a whole.
The structure is engineered to land just on the right side of this qualitative question. Beignet is designed so that, under ASC 810’s qualitative “power plus economics” test, the finger points at Blue Owl for consolidation, not Meta.
That is where Blue Owl’s role starts to look more than just financial. If it really is in charge of the JV, it is curious that S&P rates the bond off Meta, not Blue Owl, which is rated BBB by S&P. If Meta is really wearing the risk, it is curious that Meta does not consolidate it. Both cannot be fully true at once.
Unpacking finance lease rules
There is another GAAP standard that Meta has to navigate around. Under ASC 842, if a lease looks like debt, the lessee must report it as a finance lease measured as the present value of future payments. A lease is classified a finance lease if it satisfies any one of five tests, including that the lease term covers most of the asset’s life, or the present value of lease payments (plus RVG) is “substantially all” of the asset’s fair value.
Hyperion’s leases are written as four-year non-cancellable blocks with options that can take total occupancy to around 20 years. To avoid finance lease treatment, Meta proposes to assert that: a) it is not “reasonably certain” that renewal will take place beyond the initial four-year term, which is not most of the data centre’s (25 to 30-year) economic life; and b) the present value of four years of lease payments, plus the guaranteed amount under the RVG, is less than “substantially all” of the campus’s fair value.
This way, the lease can stay in the operating bucket, even though bondholders expect Meta is committing either to a long series of renewals or to writing a large residual cheque.
Accounting gymnastics
S&P's analysis supporting the bond rating is sound. Its analysis of Meta’s credit metrics is open to debate, but not materially consequential for now. S&P does not “consolidate” the structure in any accounting sense and initially only brings eight years of lease payments into its base case. It has been explicit that it may revisit that stance if it becomes clear that Meta is effectively on the hook for longer. Importantly, the agency stresses that even if Hyperion were fully reflected on Meta’s balance sheet from day one, its AA– rating would be unchanged, because the main constraints are regulatory risks, not leverage.
As long as stakeholders focus on the substance of who pays and who bleeds, they will mostly be digging in the right place. A credible version of the consolidation test would also ask where JV Co actually generates cashflows. If they are overwhelmingly generated by one party that also holds variable interests in it, that should be a strong indicator that this party is, in substance, the one that controls and supports the VIE, whatever the legal wrappers say.
On the lease side, ignoring economic incentive features feels like we are back in 2001. If Meta fails to renew the Hyperion lease, it faces having to write a large residual value cheque. That is an economic incentive to keep using Hyperion. It is not a neutral optionality; it is an economic commitment that should carry real weight in deciding whether a series of four-year leases coupled with the RVG are, in substance, long-term financing. Bond investors are behaving as if this is true. It is time for accountants to catch up.
Prasad Gollakota is a former FIG banker and co-head of the global capital solutions group at UBS. He was later chief content and operating officer at edtech company xUnlocked and specialises in financial institutions, banking regulation, capital markets and complex capital and funding solutions.