The AI competition among tech giants shifts to off-balance-sheet financing, is the trillion-dollar debt time bomb a precursor to a crisis?

In the fierce arms race of artificial intelligence, tech giants are turning to a more complex and controversial financing method to raise the massive capital needed to drive their ambitions.

On October 31, media reports indicated that tech giants are establishing special purpose vehicles (SPVs) and joint ventures, placing tens of billions of dollars in debt off their balance sheets, a maneuver that meets enormous funding needs while avoiding direct impacts on their financial status and credit ratings.

The reports noted that Meta has secured approximately $60 billion in funding for its data center construction this month. Half of this, or $30 billion, was raised through off-balance-sheet transactions structured by Morgan Stanley, with this debt held by an SPV associated with Blue Owl, and it will not appear as a liability on Meta's balance sheet.

Bankers stated that off-balance-sheet financing through SPVs or joint ventures is becoming the preferred choice for artificial intelligence data center transactions. Musk's xAI is also seeking to raise $20 billion through an SPV to lease NVIDIA chips.

UBS strategist Matthew Mish remarked in an interview that for anyone who has experienced a credit cycle, this situation is "striking." These financing tools were once associated with major scandals like the Enron bankruptcy, and their resurgence has analysts concerned about potential hidden liabilities and risks.

A More Concealed Financing Path

Wall Street has provided the latest solution for tech giants in the AI race: by transferring most of the risk to third-party investors, companies can obtain the necessary funding with limited impact on their balance sheets and ratings.

This strategy draws on the long-standing model used by energy companies to finance renewable energy projects. Banks establish an SPV or joint venture that holds assets such as chips or data centers and brings in equity investments from asset management firms or venture capital companies.

The newly formed entity can then issue bonds—typically investment-grade because it is associated with a rapidly growing AI company—to raise more funds. In return, tech companies pay rent or other fees to the entity, allowing them to enjoy the benefits of the project while limiting their financial risk exposure to the lease agreements.

Anish Shah, head of global debt capital markets at Morgan Stanley, stated:

The market value and strength of hyperscalers have brought these transactions into a whole new dimension.

He pointed out:

Blue-chip tenants like Meta, with a market capitalization of $2 trillion, have opened up possibilities for raising capital far exceeding previous project financing amounts.

Musk's xAI is a typical case. To bypass the limits on its secured debt, xAI has turned to the SPV model.

According to media reports citing informed sources, a new financing entity led by Antonio Gracias's Valor Equity Partners and Apollo Global Management is raising $20 billion through an SPV independent of xAI The entity will purchase NVIDIA chips and then lease them exclusively to xAI. xAI's risk exposure is limited to a five-year lease agreement.

Limitations of Traditional Debt

Of course, most blue-chip companies can still directly access the corporate bond market for financing.

Oracle Corporation issued $18 billion in publicly traded bonds in a single day last September to fund its cloud infrastructure deal. However, the downside of this practice is that it increases the company's liabilities, affecting future borrowing capacity, and may even lead to a downgrade of its credit rating from investment grade to junk status, thereby raising financing costs.

More importantly, in rapidly evolving fields like artificial intelligence, tech companies are reluctant to be tied down by long-term corporate bonds. S&P Global Ratings analyst Naveen Sarma stated:

These tech giants do not know what the AI landscape will look like in five years, which is part of the reason they are not solely issuing corporate bonds; they want to retain flexibility in case a certain data center is no longer needed in the future.

Although data center owners (such as Switch Inc.) can finance by packaging their receivables into asset-backed securities, this type of debt still typically appears on the company's balance sheet.

Historical Risks and Market Concerns

The history of off-balance-sheet financing tools is not glorious.

In 2001, Enron used off-balance-sheet entities to hide massive debts, ultimately leading to the company's collapse.

Before the 2008 financial crisis, banks also commonly transferred debts such as mortgages to off-balance-sheet tools, and when these liabilities were forced back on the books, the crisis ensued.

Although accounting and rating standards have tightened significantly since then, the return of "financial engineering" still raises concerns among some analysts about whether all commitments are easily traceable.

Current risks are equally significant.

First, investors face the risk of lease agreements being terminated early or terms not providing sufficient protection.

Second, the obsolescence of assets may occur faster than expected. Most cloud service providers estimate the lifespan of chips to be five to six years, but in reality, their effectiveness may diminish within three years, and an entire data center could become technologically outdated within five years.

The thirst for debt has begun to raise concerns among regulators, including the Bank of England. UBS's Mish pointed out a key difference from the dot-com bubble era:

During the dot-com bubble, most growth was financed by equity rather than debt. Therefore, when the bubble burst, the impact on the economy was manageable. Now, the capital expenditure growth of AI companies is driven by debt and is starting to be placed off-balance-sheet.

Capital Frenzy

In this financing feast, private credit funds play a crucial role.

These funds have raised billions of dollars from insurance companies and pension funds, always eager to find investable high-yield, investment-grade debt projects, and the explosion of AI infrastructure just happens to meet their needs.

According to Morgan Stanley's estimates, the entire AI ecosystem requires about $1.5 trillion in external financing. About $350 billion will come from equity investments by venture capital, sovereign wealth funds, and private equity, such as the approximately $13 billion in equity financing that Anthropic received from institutions like Iconiq Capital and the Qatar Investment Authority.

However, the remaining—over $1.15 trillion massive gap—will appear in the form of debt.

As tech giants compete fiercely in the AI field, this capital frenzy driven by off-balance-sheet financing has just begun