Historical Behavior

Meta's investment strategy in AI infrastructure has shifted from "self-built and self-used" to "joint venture + leasing." In 2024, Meta's capital expenditure was $35 billion, all for self-built data centers. In 2025, it announced a $600 billion investment plan, but starting in 2026, it began to introduce external capital—Meta only contributed 20% (about $2.8 billion) to the El Paso project, while BlackRock contributed 80% (about $11.2 billion). In comparison, Google plans to self-build $120 billion in AI infrastructure by 2026, and Microsoft combines self-building with Azure customer sharing, making Meta's "light asset" strategy more aggressive.

Capital Path

The funding structure is cleverly designed: BlackRock participates with $12.5 billion in debt financing and $4.9 billion in equity, while Meta contributes land and ongoing projects ($2.3 billion) plus cash ($500 million). Meta receives a one-time allocation of $1 billion to align with the 80/20 equity ratio. The leasing terms are an initial period of 4 years with 4 extension options (up to 20 years), with annual rent undisclosed (estimated at $1.5-2 billion/year). Funding shifts from "on-balance sheet investment" to "off-balance sheet joint venture," reducing Meta's capital expenditure pressure but increasing long-term leasing obligations.

Analogy + Industry Position

Similar cases include Amazon's $10 billion data center project in partnership with Brookfield in 2023, where Amazon leases 100% of the capacity. In 2024, Microsoft partnered with BlackRock for a $5 billion AI infrastructure fund. Meta is currently in the mid-stage of "financialization of AI infrastructure"—compared to Google (100% self-built) and Microsoft (70% self-built + 30% joint venture), Meta has the highest joint venture ratio (80% external capital in the El Paso project). Its industry position prioritizes "capital efficiency"—leveraging $1 of its own capital to attract $4 of external capital, but at the cost of long-term leasing obligations.

Structural Judgment

This essentially represents capital concentration leading to asset lightness. Traditional data center investments are "heavy assets"—self-built, owned, and used, with a capital return rate of 10-15%. Meta's joint venture model is "light assets"—with external capital holding 80%, Meta only leases, increasing the capital return rate to 25-30%. This explains why Zuckerberg refers to it as "aggressive investment"—not using its own capital, but leveraging "joint venture + leasing" to attract external capital. As AI infrastructure scales to over $100 billion, no single company can bear all capital expenditures, necessitating the introduction of insurance funds, pensions, and sovereign funds. This drives the "REIT-ization of AI infrastructure"—data centers transition from "tech assets" to "infrastructure assets," with valuation logic shifting from "growth" to "stable cash flow."

ABAB News · Cognitive Laws

Joint ventures are a light asset solution for heavy asset industries.
Leasing obligations are invisible liabilities of off-balance sheet financing.
The ultimate goal of AI infrastructure is asset securitization.