At the end of October, Meta submitted a contradictory report card: its third quarter revenue was approximately $51.2 billion, a year-on-year increase of 26%, setting a new historical high; But at the same time, the capital expenditure for 2025 will be raised to 70-72 billion US dollars, and it will be clearly stated that the absolute amount of capital expenditure in 2026 will be larger, and the growth rate of total expenses will also significantly accelerate. These investments mainly flow towards the construction of AI infrastructure, incremental cloud spending and depreciation, as well as AI talent compensation.
Even worse, a one-time tax shock of approximately $15.9 billion resulted in GAAP earnings per share of only $1.05, far below market expectations. After the release of the financial report, the stock price experienced a significant pullback, falling by over 13% in two trading days. The main question in the market is “when will we see returns on our massive investments”.
In fact, prior to the release of its Q3 financial report, Meta announced a $27 billion “Hyperion” data center joint venture project in mid October. The transaction structure is: Blue Owl contributes about 80%, Meta contributes about 20%, and issues approximately $27 billion in A+rated bonds through SPV, with approximately $2.5 billion in equity, anchoring long-term institutional funds such as PIMCO and BlackRock.
The key design is that these assets are not included in Meta’s balance sheet, and after the data center is built, Meta will lease it back for a long time and retain operational control.
Throughout the development history of various industries, there is a clear pattern: when assets are financialized and securitized, the industry often experiences explosive growth.
The demand for capital in the AI industry far exceeds that of most other fields in the past, making it one of the “most capital intensive” industries. Faced with this challenge, the industry has been actively exploring solutions, including innovative models that combine RWA (Real World Assets) with the AI industry.
The stones from other mountains can be used to attack jade.
Meta’s design is a typical case of such attempts.
Financing stratification and risk pricing: a structured path for the $27 billion intelligent computing center
The core mechanism of the entire transaction can be divided into four levels:
The first step is to reduce on balance sheet pressure through equity stratification. Meta only invests 20%, while Blue Owl invests 80%. It’s equivalent to Meta providing technology and branding, with Blue Owl paying for it. In this way, Meta does not have to withdraw a huge amount of cash at once, nor does it include this investment in the balance sheet, to avoid excessive financial leverage.
The second step is to raise funds through SPV structure financing. Morgan Stanley helped establish a specialized company, SPV (Special Purpose Vehicle), which can be understood as a specialized project company. By issuing bonds through SPV, the assets and risks of this financing are isolated from the parent company. This SPV issued $27 billion in A+rated bonds and $2.5 billion in equity, packaging the future cash flow of the data center into a financial product – temporarily referred to as the “AI infrastructure bond” – and selling it to qualified investors in a contractual manner.
Step three, Meta rents back the intelligent computing center. After the data center is built, Meta does not directly own it, but signs a long-term lease agreement. From an accounting perspective, this money becomes an annual operating expense OpEx, rather than a one-time large capital expenditure CapEx. In this way, Meta can not only alleviate the pressure on the balance sheet, maintain a high credit rating, but also continue to operate this intelligent computing center.
Step four, introduce residual value guarantee RVG mechanism. Meta promises that the value of this data center will still be guaranteed after 16 years. If the actual selling price of future related assets is lower than a certain guaranteed price, Meta, as the “backstop”, is responsible for making up the difference. This is equivalent to adding a layer of “insurance” to the bonds, thereby giving them an A+rating.
Despite the RVG guarantee and high rating, investors still require an interest of 225 basis points, or 2.25 percentage points, higher than the (US) treasury bond as risk compensation. This interest rate is even higher than the corporate bonds directly issued by Meta. Why is that? Because this type of structured bond has more uncertainty: it relies on the cash flow performance of underlying assets, depends on whether RVG counterparties can fulfill commitments, and also faces the complexity of legal structure and relatively poor liquidity. These additional risks naturally require a thicker risk premium.
Through this design, the $27 billion data center investment is structured into A+rated debt and a small amount of equity, held by long-term capital such as insurance companies and pension funds. Meta retains operational control but converts capital expenditures into operational expenditures, thereby reducing financial leverage and maintaining credit ratings.
Balancing the interests of Meta and bond buyers
Specifically, what are the benefits and risks of this design for Meta and bond investors?
For Meta, the first step is to optimize the balance sheet. Meta will convert its $27 billion capital expenditures into off balance sheet assets and long-term leases, undertaking construction and holding through SPVs, while paying rent on schedule, which will be recorded as operating expenses.
In this way, Meta’s balance sheet will not directly bear huge project debts, and the presentation effect of surface leverage ratio and free cash flow will be better, which will help maintain credit ratings and retain space for stock repurchases and dividends.
Secondly, accelerate the process of AI expansion. A large-scale AI data center requires significant capital investment, and the introduction of Blue Owl provides 80% of the capital, making the construction speed no longer limited by the pace of self owned capital expenditure. At the same time, Meta leads the construction and operation management to ensure that the project progresses according to plan and meets technical standards.
In terms of control, although it is a joint venture model, Meta is responsible for construction management and long-term leasing of these facilities, still mastering the technical roadmap and operational standards of the intelligent computing center. When necessary, strategic flexibility can be maintained by adjusting leases or repurchases.
Financing costs have also been optimized. By providing a 16 year residual value guarantee, Meta helped SPV issue bonds with an A+rating, pushing the interest rate of this ultra long term fund into the investment grade range. Although still higher than the cost of Meta’s direct bond issuance, it is lower than the typical asking price for pure project financing, and the overall efficiency may be better than relying solely on on on balance sheet financing.
In addition, risks are accurately priced and isolated. The uncertainty of technical depreciation and residual value is transformed into the provision of residual value guarantee, and Meta only needs to make up for the price difference in extreme cases, without having to bear higher comprehensive capital costs throughout the entire construction cycle.
For Blue Owl and bond buyers, including insurance funds and pensions, this is an extremely attractive investment portfolio. It has a triple guarantee of high rating, physical asset support, and long-term leases for large tenants: stable rental cash flow paid by Meta, heavy assets such as data centers as collateral or disposables, and a 16 year residual value guarantee provided by Meta as downside protection. The overall risk return structure is very suitable for long-term fund allocation.
The yield level is significantly higher than that of bonds of the same grade. This is a structured private credit product based on new underlying assets. In the face of liquidity uncertainty and the risk premium space brought by the ultra long project cycle, the coupon is about 225 basis points higher than treasury bond. For buyers, they have achieved more substantial interest rate spread returns within an investment grade framework.
This transaction also allows investors to seize the growth opportunity in the AI infrastructure field. Blue Owl has obtained 80% equity and management say in the project, accumulating significant digital infrastructure assets. For bond investors, this is equivalent to gaining core asset exposure with predictable cash flow in the field of AI computing power new infrastructure.
The transaction terms are clear and the risks can be layered. The asset disposal during the construction period, operation period, expiration or non renewal of lease, as well as the triggering path for residual value guarantee, are all pre written in the contract. Investors at different levels can choose different investment levels such as equity, mezzanine, or senior bonds based on their own risk preferences.
However, both sides also have their own costs to bear.
Meta provides residual value assurance to bring the tail risks of technical depreciation and residual value back to itself. If the actual value of the asset is lower than expected after the expiration of the contract, or if there is a significant technological iteration during the contract period that causes a sharp decline in asset prices, Meta needs to make up the difference according to the terms, which constitutes a contingent liability.
On the investor side, despite the dual protection of A+rating and residual value guarantee, there are still multiple uncertainties: Meta’s credit redemption ability as a counterparty, the actual progress of construction and grid connection, and the possibility of drastic changes in leasing prices due to long-term technological iterations. Especially for a high-tech enterprise, during a period of intense industry change, 16 years is a very long time. 16 years ago, portal websites such as Yahoo, AOL, and MSN still existed, but with the further evolution of Web 2.0, they disappeared in just a few years.
The long-term significance of “AI REITs”
In the financial field, there is a term called asset backed securitization (ABS). This new model can be seen that AI platform manufacturers are trying to shift AI infrastructure from the traditional enterprise capital expenditure (CapEx) model to a financial innovation model centered on private equity credit and ABS, which can be called AI infrastructure backed securities (AIBS).
The AI computing industry and real estate have many similarities, and this new AIBS is similar to REITs. In this way, investors can view the debt of the intelligent computing center as a new type of fixed income product with “high rating and low volatility”, while the operator can quickly expand through the SPV model without affecting the balance sheet.
The constraint on the scale of AI infrastructure is not the amount of capital, but the capital structure.
If this model is successful, the computing industry can use financial power to drive a larger scale infrastructure frenzy. The internal circulation driven by Nvidia OpenAI, which is widely discussed in the industry, has been found by capital to inject liquidity continuously.
Next, the industry only needs to allocate electricity, which is green and centralized.
暂无评论内容