Why Meta's $26B Leaseback Rewrote AI Infrastructure Financing
Meta-Apollo Private Credit Structure, The Leaseback Model Defined, Why Hyperscalers Are Moving Off-Balance Sheet, What This Means for the Infrastructure Capital Stack
Welcome to Global Data Center Hub. Join investors, operators, and innovators reading to stay ahead of the latest trends in the data center sector in developed and emerging markets globally.
The Question Meta’s Finance Team Was Asking
Meta’s finance team faced a specific problem in 2023.
The company had committed to more than $35 billion in capital expenditure in 2024.
The AI infrastructure buildout required land, power, cooling systems, and facilities at a scale and speed the balance sheet could absorb but at a cost that would suppress free cash flow for years.
The question was not whether to build.
The question was who should own the buildings.
The answer produced the largest private credit transaction in the history of digital infrastructure.
How AI Infrastructure Has Always Been Financed
The Compute Factory series established the ownership transfer that defines each compute era.
IBM Built the Factory. The Market Built a Different One showed that factory ownership determines where returns accumulate.
Why AWS EC2 Rewrote the Economics of Compute Ownership traced how EC2 concentrated factory ownership inside a small number of hyperscale operators.
The financing model that followed that concentration was straightforward.
Hyperscalers built and owned the infrastructure they operated.
The capital expenditure sat on-balance sheet, funded through operating cash flow and corporate debt at investment-grade rates.
The return on that capital accumulated to the hyperscaler’s shareholders.
By 2023, the scale of the AI infrastructure buildout had outgrown that model.
The capital required to build the next generation of GPU-dense compute factories exceeded what any single hyperscaler’s free cash flow could fund without compressing returns across the rest of the business.
Meta found the answer before the rest of the market understood the question.
The Mechanics of the Deal
Meta closed a transaction with Apollo, Brookfield, KKR, Carlyle AlpInvest, and PIMCO in 2024.
The structure was a leaseback. The private credit consortium would finance the construction of data center capacity. The investors would own the assets. Meta would lease them back under long-term agreements.
The debt component was $26 billion. The equity contribution from the private credit partners was $3 billion. Total capital deployed: approximately $29 billion.
The economics were precise on both sides. Meta secured the infrastructure it needed without adding the asset to its balance sheet.
The lenders secured long-duration, investment-grade cash flows backed by one of the strongest corporate counterparties in the world.
The lease payments from Meta provided the debt service. The data center itself served as the collateral.
The deal closed without Meta deploying its $70 billion cash pile. The company’s investment-grade debt capacity remained available for other uses. The balance sheet stayed intact while the infrastructure scaled.
What the Leaseback Model Actually Is
Here is the analytical frame this capital series is building toward.
The leaseback model is an application of industrial financing logic to compute infrastructure. Every major industrial sector (pipelines, power generation, contracted manufacturing) has used this structure for decades.
A producer who needs output contracts with an operator who builds and owns the facility. The producer’s lease payments fund the construction cost and return. The asset serves as collateral for the debt.
What is new is the application of this structure to the compute factory at hyperscale. The Meta deal established three things simultaneously.
Private credit can price AI infrastructure risk. Apollo, Brookfield, KKR, Carlyle AlpInvest, and PIMCO underwrote $29 billion against a single counterparty’s lease commitment.
The willingness to do so at that scale established a market-clearing price for compute factory debt.
Hyperscaler operational control does not require hyperscaler asset ownership. Meta runs the infrastructure. Meta does not own it.
The separation of operation from ownership is the foundational logic of the leaseback and its arrival in AI infrastructure changes the capital stack structure permanently.
The lease payment is the offtake agreement. The private credit firms underwrote Meta’s obligation to pay, not the data center.
The asset was the contract. The collateral was the commitment. That distinction is the subject of the next article in this series.
What This Changes for the Market
The Meta deal is a template. Every hyperscaler with a large AI infrastructure commitment and constrained free cash flow has now seen the structure demonstrated at scale.
The private credit market has seen the risk priced and transacted. The conditions for replication are in place.
For independent data center developers, the leaseback model creates a new customer profile. A hyperscaler structured around leased capacity is a tenant, not a competitor.
The developer who can source private credit, build to hyperscaler specification, and structure a long-term lease is not competing against the hyperscaler. The developer is serving it and capturing the development margin the hyperscaler chose to outsource.
For private credit investors, the Meta deal established the asset class. AI infrastructure debt long-duration, investment-grade counterparty, collateralized by purpose-built assets fits institutional mandates across insurance, pension, and sovereign capital.
The deal size demonstrated the market can absorb large single transactions. The counterparty quality demonstrated the credit risk is manageable at investment-grade terms.
For public equity investors, the leaseback changes the capex read. A hyperscaler deploying $20 billion in AI infrastructure through a leaseback structure is committing $20 billion in future lease obligations appearing on the operating expense line rather than the capital expenditure line.
The finacial statements all read differently. The investment decision changes accordingly, and most equity frameworks are not yet calibrated for the distinction.
The Question the Deal Left Open
Meta’s finance team answered the ownership question. The leaseback model settles who owns the factory. Private credit firms build it. Meta leases it.
The credit instrument raises a deeper question. The Meta deal was underwritten against Meta’s corporate balance sheet, an investment-grade counterparty whose creditworthiness exists independent of the data centers being financed.
The next compute factory deal will not always have that counterparty. The next transaction will be underwritten against the offtake agreement alone, the lease commitment from an operator whose creditworthiness is the lease itself.
That is when the leaseback model’s true analytical requirements become visible. The asset is the contract. Understanding exactly what that means and how to underwrite it when the counterparty is not Meta is where this series goes next.
Great piece. Clearly, Meta's IG credit is the foundation, but it may start the trend for the financing structure to be used more widely. One question: post-ASC 842, I'm not sure Meta will get this entirely off-balance sheet, though the capex line is definitely lighter. Also, do you have a sense who is depreciating this asset for tax purposes?
Which deal is this? Are you referring to Meta’s Hyperion buildout? Because if so, that’s with Blue Owl and Pimco (and maybe 1/2 others).
Not 100% sure it’s a leaseback deal though I could be wrong. I’ve analysed the deal here and explained the use of SPVs and how ASC 842 plays a role.
https://lesbarclays.substack.com/p/the-mechanics-of-conduit-debt-financing?r=rq26d&utm_medium=ios